Search

• A novel water transparency calculation algorithm is introduced, employing deep neural network and image processing technology. This algorithm obviates the need for a water gauge, allowing the calculation of water transparency solely through the analysisof Secchi disk videos. Notably, it demonstrates heightened detection efficiency and success rates. • An enhanced Siamese tracker, termed SiamDCFF (Siamese tracker based on dual correlation feature fusion), is proposed for Secchi disk recognition. Distinguishing features of SiamDCFF include a dual correlation module and a feature fusion module. On the Secchi disk dataset, SiamDCFF achieves the highest success rate, outperforming other methods detailed in this paper. • The simplified 3D-ResNet [41] is leveraged to assess the state of the Secchi disk, transforming the critical position determination problem into an optimization challenge. The objective function involves maximizing the number of visible and invisible Secchi disk images on both sides of the critical position. This method facilitates adaptive critical position determination. • A novel Gated Recurrent Unit based on auto-regression (ARGRU) is proposed, taking into account the motion characteristics of the Secchi disk. ARGRU achieves a higher success detection rate for water transparency compared to the water gauge recognition method. Notably, ARGRU outperforms RNN [42], GRU, and LSTM [43] on the dataset presented in this paper. Water transparency is traditionally assessed through visual inspection using a lowered Secchi disk (SD) into the water, with the disappearance depth of the SD recorded as the measure of transparency. However, manual and visually-dependent frequency measurements of the SD render this process labor-intensive and time-consuming. This paper presents a comprehensive machine-vision-based algorithm designed for the automatic computation of water transparency using Secchi disk videos. The algorithm leverages multiple deep neural networks, including an enhanced Siamese tracking model, a 3D-ResNet, and an improved GRU network. Trained models directly identify the SD in the video, calculating water transparency by processing pixel information. Experimental results demonstrate the algorithm's efficacy in estimating water transparency across diverse natural environments, achieving commendable accuracy (MAE = 3.6 cm MSE = 21.5 cm RMSE = 4.6 cm), rapid processing speed (average 6.87 s), and robust stability. In comparison with the former water-gauge-based algorithm, our proposed algorithm exhibits heightened efficiency and a superior detection success rate. [ABSTRACT FROM AUTHOR]

Multi-exposure image fusion (MEF) is an important area in computer vision and has attracted increasing interests in recent years. Apart from conventional algorithms, deep learning techniques have also been applied to MEF. However, although many efforts have been made on developing MEF algorithms, the lack of benchmarking studies makes it difficult to perform fair and comprehensive performance comparison among MEF algorithms, thus hindering the development of this field significantly. In this paper, we fill this gap by proposing a benchmark of multi-exposure image fusion (MEFB), which consists of a test set of 100 image pairs, a code library of 21 algorithms, 20 evaluation metrics, 2100 fused images, and a software toolkit. To the best of our knowledge, this is the first benchmarking study in the field of MEF. This paper also gives a literature review on MEF methods with a focus on deep learning-based algorithms. Extensive experiments have been conducted using MEFB for comprehensive performance evaluation and for identifying effective algorithms. We expect that MEFB will serve as an effective platform for researchers to compare the performance of MEF algorithms. • A literature review on multi-exposure image fusion is presented. • A multi-exposure image fusion benchmark is proposed. • The benchmark integrates a test set, algorithms, evaluation metrics, and a toolkit. • Comprehensive performance comparison is conducted and presented. [ABSTRACT FROM AUTHOR]

This paper proposes the planar image-space trajectory (PIST) planning algorithm for contour following using computer vision and a force control scheme. The PIST algorithm generates machining trajectories offset to a planar workpiece with an applied force direction so that a force controller can be utilized. The PIST algorithm starts with an image of the workpiece, corrects for the perspective of the camera, defines a contour following trajectory in image space, and then transfers the trajectory into the robot's workspace for playback. A case study is presented where the PIST algorithm is applied to deburring sheet metal parts for validation through experimentation. A comparison is performed between the PIST algorithm and CAD/CAM software using CAD data for path design and a laser scanner for localization. The PIST planner achieved a positional accuracy greater than 1.7 mm compared to the CAM/CAM accuracy of 1.2 mm. An admittance controller successfully accounted for the positional error to achieve satisfactory and equivalent surface finish results for both trajectory planning approaches. The PIST planner is a beneficial alternative for trajectory planning as it significantly reduced the setup and process time while adapting to deviations from the design dimensions and manufacturing imperfections. [ABSTRACT FROM AUTHOR]

• The two change detection algorithms proposed in this paper use KFCM gradient clustering algorithm. • KFCM algorithm is a classic clustering algorithm in unsupervised clustering, which can effectively cluster most of the data. • We will study how to use the special features of the image to complete the clustering of the difference maps. In recent years, computer vision, especially deep learning, has been widely used in various fields. Through the deep learning aerial image detection gradient clustering algorithm automatic recognition, it can solve the limitations of manual shooting by humans, can shoot from a high altitude to a panoramic view of a specific area, and provide a more comprehensive solution. The traditional forest resource management and management work is mainly carried out by forestry personnel to carry out a large number of investigations and investigations on the forest. This method not only consumes a lot of manpower and material resources, but also does not have real-time nature. It is difficult to deal with all kinds of forest management. Problems, causing unnecessary losses. In this regard, this paper proposes an aerial image change detection algorithm based on H-KFCM, and designs related experiments to verify and demonstrate the performance of the algorithm. In this paper, we conduct a parallel study based on deep learning on the gradient clustering algorithm of deep learning in aerial image processing. By using CUDA (Compute Unified Device Architecture) to perform large-scale parallel processing of aerial data. Can greatly shorten the time to obtain results, improve the efficiency of relevant personnel. Experiment analysis. It can be seen from the results that the deep learning parallelization program implemented in this paper has a faster calculation speed and uses less time in high-resolution images, and has a good acceleration ratio compared to the CPU. [ABSTRACT FROM AUTHOR]

The method based on machine vision image processing is used to detect the surface defects of Si 3 N 4 bearing roller. Owing to the variety of defects, small area and low contrast, it is easy to miss or error detection. In this paper, an adaptive update template defect enhancement algorithm based on Gaussian model is proposed. First, a large number of surface images of Si 3 N 4 bearing roller are collected to obtain the non-defect background statistical feature, and the background characteristic curve is fitted by Gaussian model. Further, the initial background template is gained according to the Gaussian curve. Then, combined with the gray distribute of defect images and initial background template, unique adaptive update template can be established. Finally, subtraction operation and nonlinear enhancement are used to improve the comparison of defect information and background. Through inverse sorting, adaptive threshold segmentation and Canny operation, the precise positioning of defects is realized. The enhancement algorithm can effectively enhance the contrast and eliminate the influence of noise. The average detection time is 0.84s, and the detection accuracy is 96.2%. [ABSTRACT FROM AUTHOR]

A large amount of data, provided in the form of video data, is acquired during manned inspections flights of electric power lines. This data is analyzed by expert human inspectors to detect faults in the power lines infrastructure and prepare the inspection reports. This process is extremely time consuming, very expensive and prone to human error. In this paper, we present PoLIS: the Power Line Inspection Software, which has been developed with the objective of assisting the analysis of the data acquired during inspection flights. PoLIS is based on the cooperation between computer vision and machine learning techniques to automatically process video sequences acquired during inspection flights, resulting in a set of representative images per electric tower which we call Key Frames. These representative images can then be used for inspection purposes, leading to a drastic reduction of the human operators’ workload. At the core of the strategy lies an electric tower detector, which is in charge of estimating the location of the towers within the images based on the combination of a sliding window search technique and a supervised classifier. The location of the tower is then tracked using a tracking-by-registration algorithm based on direct methods, estimating the position of the tower in different images. Finally, different criteria are applied for defining whether the image corresponds to a Key Frame image or not. Extensive evaluation of the proposed strategy is conducted using videos acquired during manned helicopter inspections. The videos constituting this database contain several thousand frames representing both medium and high voltage power transmission lines in the infra-red (IR) and visible spectra. The obtained results show that the proposed strategy can reduce the large amount of data present in the inspection videos to a few Key Frames for each tower. It is also demonstrated that the learning-based approach proposed in PoLIS is appropriate for detecting electric towers, a process which is made faster and more robust by coupling it with a tower tracking algorithm. A Graphical User Interface allowing the application of PoLIS to user-provided videos is also presented in this paper, illustrating the whole process and the automated generation of an inspection report. [ABSTRACT FROM AUTHOR]

Dictionary learning has played an important role in the success of sparse representation. Although several dictionary learning approaches have been developed for image classification, discriminative dictionary pair learning, i.e., jointly learning a synthesis dictionary and an analysis dictionary, is still in its infant stage. In this paper, we proposed a novel model of Fisher discrimination dictionary pair learning (FDDPL), in which Fisher discrimination information is embedded into analysis representation, analysis dictionary, and synthesis dictionary representation. With the proposed Fisher-like discrimination term, discrimination of both synthesis dictionary representation and analysis dictionary representation is introduced into the dictionary pair learning model. An iterative algorithm to efficiently solve the proposed FDDPL and a FDDPL based classifier are also presented in this paper. The experiments on face recognition, scene categorization, gender classification, and action recognition clearly show the advantage of the proposed FDDPL. [ABSTRACT FROM AUTHOR]

Image segmentation is the process of dividing image into homogenous regions by some charasteristics and is widely used in medical diagnostics. Segmentation algorithms are used for anatomical features extraction from medical images. The Hybrid Ant Colony Optimization (ACO) – k-means and Grub Cut image segmentation algorithms for MRI images segmentation are considered in this paper. The proposed algorithms and sub-system for the medical image segmentation have been implemented. As there is no universal algorithm for medical image segmentation, image segmentation is still a challenging problem in image processing and computer vision in many real time applications and hence more research work is required. The experimental results show that the proposed algorithm has good accuracy in comparison to Grub cut. [ABSTRACT FROM AUTHOR]

• Performance analysis of different video-based eye blinking detection. • Development of a low-cost imaging acquisition system. • Quality metrics analysis of proposed algorithms. • Performance comparison between RGB and Grayscale images. • The red channel of the RGB camera achieves in our experiments a 100 % eye blink detection rate for event detection. The eye is considered a rich source for gathering information on our daily lives. In addition, the necessity to classify the eye as open or closed and detect eye blinks is increasing in various fields. For instance, it has been proven that eye blinks can be related to different ophthalmological conditions. Nevertheless, accurate, and robust blink estimation poses significant challenges to the efficacy of real-time applications due to the variability of eye images and light conditions. In this paper eye blinking has been monitored using a chin rest and an iOS smartphone attached to a cartesian machinery. Five video-based eye blink algorithms to classify frames have been analyzed and compared in terms of performance. To evaluate the results, different widely used statistical metrics have been employed. The results show that the proposed techniques successfully detect eye blinks, whereas the frame classifier shows different scores depending on the algorithm used. [ABSTRACT FROM AUTHOR]

Image synthesis is currently one of the most addressed image processing topic in computer vision and deep learning fields of study. Researchers have tackled this problem focusing their efforts on its several challenging problems, e.g. image quality and size, domain and pose changing, architecture of the networks, and so on. Above all, producing images belonging to different domains by using a single architecture is a very relevant goal for image generation. In fact, a single multi-domain network would allow greater flexibility and robustness in the image synthesis task than other approaches. This paper proposes a novel architecture and a training algorithm, which are able to produce multi-domain outputs using a single network. A small portion of a dataset is intentionally used, and there are no hard-coded labels (or classes). This is achieved by combining a conditional Generative Adversarial Network (cGAN) for image generation and a Meta-Learning algorithm for domain switch, and we called our approach MetalGAN. The approach has proved to be appropriate for solving the multi-domain label-less problem and it is validated on facial attribute transfer, using CelebA dataset. [ABSTRACT FROM AUTHOR]

Underwater images acquired in scattering environments are generally of poor quality because of the attenuation and backscattering of light when it passes through water with scattering particles. Polarimetric de-scattering methods can be used to significantly enhance the imaging quality in such a so-called turbid water. However, due to the complexity of polarimetric de-scattering algorithms, it is hard to achieve real-time de-scattering output from a polarimetric camera. In this paper, in order to efficiently increase the computational efficiency, the polarimetric de-scattering algorithm is optimized and a multi-threading framework is developed that enables the algorithm to run in real-time on ordinary laptops for the same polarimetric camera. We demonstrate the imaging performance by using the underwater polarimetric de-scattering system we proposed. We analyze the algorithm under different scattering conditions and discuss its optimal parameters. We find that there is a significant increase in the number of SIFT (i.e., scale invariant feature transform) feature points extracted from the processed image compared to the original image, showing that our system has potential applications in pattern recognition, computer vision, etc. [ABSTRACT FROM AUTHOR]

In recent decades, methods to find invariant points in digital images, called fiducial points, have gained great attention, mainly due to the demands of several applications in computer vision and image processing, such as the geometric matching of global structures, objects or specific regions. Among the most well known approaches are algorithms like SIFT, HOG, SURF, and their variations as A-SIFT, PCA-SIFT, surrounded by many others. Though the number of researches demonstrating the efficiency of such methods is undoubtedly enormous, the vast majority compares their performances only on pairs of images with little changes in view perspectives, objects or specific regions of the scenes. Thereby, the study of this type of technique under large viewpoint changes, called here LVC, has received little attention from the researchers. On the other hand, with the aim of filtering points of interest, most techniques have used the traditional extensive statistics. However, methods for image processing based on a new type of statistics, called non-extensive statistics, have shown to be efficient in several applications. In this paper, we present a new method, called q -SIFT, based on the non-extensive Tsallis statistics, to find fiducial points in a sequence of frames of videos under large viewpoints changes. We experimentally show the efficiency of the proposed method in video databases and propose new measurement metrics for this type of algorithm. • We present a method for Fiducial Point Description based on non-extensive physical systems. • We test the proposed method under what we call severe conditions, little tested in the literature. • We compare the proposed method with the main methods found in the literature, showing that the proposed method surpasses all others. • We present a new methodology to measure the performance of the studied algorithms. • We show that the database used represents non-extensive physical systems. [ABSTRACT FROM AUTHOR]

Abstract Face detection plays an important role in many computer vision applications. In recent years, much research has focused on extending the well-established Adaboost face detector algorithm for multi-view face detection. However, detecting in-plane and out-of-plane rotated faces simultaneously is still a challenging task today. In this paper, a very robust multi-view face detection algorithm, with its core functionality based on frontal face detection, is proposed to simultaneously detect in-plane and out-of-plane rotated faces. Moreover, only the training data in the frontal face is needed and we do not require the training data from different in-plane or out-of-plane rotation angles. In the proposed algorithm, first, techniques such as the skin filter and entropy rate superpixels (ERSs) are applied to obtain face candidates. Then, angle compensation and refinement are applied to improve the accuracy of face detection in the in-plane case. Moreover, the symmetry extension technique, i.e., extending the face candidate with its flipped version to create a face similar to the frontal one, is applied to detect out-of-plane faces without the need of training data. Simulations on the FEI, Pointing'04, BaoFace, Group, Utrecht, and WWW datasets demonstrate the proposed algorithm's effectiveness and superior performance as compared to state-of-the-art face detection methods. Graphical abstract Unlabelled Image Highlights • We detect in-plane and out-of-plane rotated faces by only the frontal face detector. • Training data from different in and out-of-plane rotation angles are not required. • Skin filters and ERSs can preserve the faces from different illumination and races. • Angle compensation computes the rotation angle for each rotated face. • Symmetry extension creates pseudo faces to well detect out-of-plane rotated faces. [ABSTRACT FROM AUTHOR]

• This paper proposes multiple sets of laser lines to assist belt defect detection. • The solution proposed in this paper can precisely detect the torn area of the belt. • This method supports users to customize the alarm level of tear detection. • This method is more accurate than the comparison method mentioned in the article. Conveyor is mainly used for ore collection and transportation, and it plays an important role in mining operations. Due to the harsh mining environment, the conveyor belt often has longitudinal tear. If the tears can not be found in time, a safety accident will occur and cause significant economic losses. In order to accurately and efficiently detect the longitudinal tearing of the conveyor belt, this paper proposes a computer vision detection algorithm based on multiple sets of lasers. This algorithm can accurately segment the laser stripe area through image processing and analysis, and accurately locate the torn area of the conveyor belt. And this method supports users to customize the alarm level of tear detection, which meets different industrial production scenarios. The experimental results show that the method proposed in this paper can accurately and quickly detect the longitudinal tearing defect of the conveyor belt. [ABSTRACT FROM AUTHOR]

• An algorithm for identifying and locating sugarcane stem nodes is proposed. • Use the horizontal projection function of binary image to intercept the region of interest. • Use vertical projection function and local pixels to locate sugarcane stem nodes. • The algorithm has good performance in recognition rate, accuracy and time-consuming. Aiming at the difficulty of sugarcane nodes identification and location during automatic cutting of sugarcane seeds, based on machine vision system, this paper proposed a sugarcane nodes identification algorithm based on sum of local pixel of minimum points of vertical projection function. Firstly, according to the color and texture characteristics of yellow sugarcane, the captured RGB image of sugarcane was converted into HSV color image. Then, the S-component image in the HSV image was extracted, and the S-component image was binarized by the Otsu algorithm to obtain the binary image, and then the binary image was processed by morphological closed operation, which can eliminate the noise and fill holes of binary image. Subsequently, the sugarcane area was segmented as the region of interest through the horizontal projection of the binary image, which lowered the amount of calculation while reducing interference. Finally, the vertical projection function of the binary image of the region of interest was established, and the function was continuously derived to obtain the minimum points, and then the position of sugarcane nodes was preliminarily determined by the obtained minimum points. Then, the final position of the sugarcane nodes were determined according to the number of nodes to be identified and the sum of pixels of each 5 columns on both sides of the minimum points. The pixel column positions corresponding to the minima of the sum are the accurate nodes positions determined by the proposed algorithm. The experimental results show that the algorithm proposed in this paper has a single node identification rate of 100%, with an average time consumption of 0.15 s, and a position deviation of less than 0.34 mm; a double nodes identification rate of 98.5%, with an average time consumption of 0.21 s, and a position deviation of less than 0.42 mm. Compared with other nodes identification algorithms mentioned in this paper, it has higher identification rate and accuracy. [ABSTRACT FROM AUTHOR]

We propose a novel fully convolutional network architecture for shapes, denoted by Shape Fully Convolutional Networks (SFCN) . 3D shapes are represented as graph structures in the SFCN architecture, based on novel graph convolution and pooling operations , which are similar to convolution and pooling operations used on images. Meanwhile, to build our SFCN architecture in the original image segmentation fully convolutional network (FCN) architecture, we also design and implement a generating operation with bridging function. This ensures that the convolution and pooling operation we have designed can be successfully applied in the original FCN architecture. In this paper, we also present a new shape segmentation approach based on SFCN. Furthermore, we allow more general and challenging input, such as mixed datasets of different categories of shapes which can prove the ability of our generalisation. In our approach, SFCNs are trained triangles-to-triangles by using three low-level geometric features as input. Finally, the feature voting-based multi-label graph cuts is adopted to optimise the segmentation results obtained by SFCN prediction. The experiment results show that our method can effectively learn and predict mixed shape datasets of either similar or different characteristics, and achieve excellent segmentation results. [ABSTRACT FROM AUTHOR]

Stereo 3D reconstruction of ocean waves is gaining more and more popularity in the oceanographic community and industry. Indeed, recent advances of both computer vision algorithms and computer processing power now allow the study of the spatio-temporal wave field with unprecedented accuracy, especially at small scales. Even if simple in theory, multiple details are difficult to be mastered for a practitioner, so that the implementation of a sea-waves 3D reconstruction pipeline is in general considered a complex task. For instance, camera calibration, reliable stereo feature matching and mean sea-plane estimation are all factors for which a well designed implementation can make the difference to obtain valuable results. For this reason, we believe that the open availability of a well tested software package that automates the reconstruction process from stereo images to a 3D point cloud would be a valuable addition for future researches in this area. We present WASS ( http://www.dais.unive.it/wass ), an Open-Source stereo processing pipeline for sea waves 3D reconstruction. Our tool completely automates all the steps required to estimate dense point clouds from stereo images. Namely, it computes the extrinsic parameters of the stereo rig so that no delicate calibration has to be performed on the field. It implements a fast 3D dense stereo reconstruction procedure based on the consolidated OpenCV library and, lastly, it includes set of filtering techniques both on the disparity map and the produced point cloud to remove the vast majority of erroneous points that can naturally arise while analyzing the optically complex nature of the water surface. In this paper, we describe the architecture of WASS and the internal algorithms involved. The pipeline workflow is shown step-by-step and demonstrated on real datasets acquired at sea. [ABSTRACT FROM AUTHOR]

Nowadays, the technology of using computer vision to calibrate objects is widely used, which has a huge market demand in many fields. This paper provides a calibration method of computer vision system based on dual attention neural network. This paper uses the camera to simulate human eyes to obtain three-dimensional images. After obtaining the three-dimensional images, the images are input into the Residual Network (ResNet) model, and the weight of ResNet is repeatedly updated so as to accurately identify the images. On this basis, introduces dual attention mechanism that an algorithm is used in natural language to the visual image processing, using multistage feature extraction method to extract the three-dimensional image for each characteristic of regional. After extracting the feature area, the accuracy of the feature area is constantly updated to the minimum. Besides, the feature areas are brought into the calibration algorithm of Zhang Zhengyou system to obtain the spatial coordinates of the objects in the attention area. This method can realize the space position calibration of specific objects under various complex backgrounds and calculate the distance from the calibrated objects, which can not only calibrate the system but also identify it, and greatly improve the reliability and accuracy of the calibration process. • The dual attention model realizes object calibration in complex background. • The dual attention mechanism can be used to calibrate and identify the system. • The proposed algorithm makes the calibration error less than 0.1%. • The algorithm is used to identify and calibrate parking space. [ABSTRACT FROM AUTHOR]

Automatic detection of defective oranges by computer vision system is not easy because of the uneven lightness distribution on the surface of oranges. It means that the methods only directly using global segmentation provide unsatisfactory results when orange images present faint defect characters or inhomogeneous surface. The contrast between sound and defective regions can be used to produce more accurate segmentation results, which is more capable of detecting pixels lying around the defect boundary on orange surface based on the local segmentation method. In this paper, we study and propose a sliding comparison window local segmentation algorithm and also presents the detailed image processing procedure including removal of background pixels, image binarization using local segmentation, image subtraction, image morphological modification, removal of stem end pixels for detecting surface defect in an orange gray-level image. This method is an original contribution that allows successful segmentation of various types of surface defects (e.g., insect injury, wind scarring, thrips scarring, scale infestation, canker spot, dehiscent fruit, copper burn, phytotoxicity).The image segmentation algorithm was tested with 1191 samples of oranges. The proposed algorithm was able to correctly detect 97% of the defective orange. Future work will be focused on whole surface and fast on-line inspection. [ABSTRACT FROM AUTHOR]

In the scenario of big data, incorporating multiple attribute descriptions of the same subject (i.e., so-called multi-view data analysis) and the developing effective data visualization techniques have been one of the most important topics in the computer vision and machine learning community. In this paper, a multiple view source attributes embedding framework is proposed that can assemble various object attributes together and embed them into a low dimensional space to realize visual data structure demonstration. Classical sole attribute data embedding approaches such as MDS and SHE can be utilized under this framework. The core idea of this framework is to achieve a uniform contribution merging, which means every source attribute will contribute equally to the final embedded output. The underlying reason for this idea is that now that an attribute is input as a source and no prior weight is given beforehand, it should be treated equally. The most exciting advantage of this strategy is that it can avoid prejudice caused by major attributes (exerting due to their evident quantity advantage), which may lead the minor attributes be ignored. In addition, we propose a simple iterating algorithm to implement this framework, and plenty of experiments are conducted under diverse source attribute configurations to validate its effectiveness. [ABSTRACT FROM AUTHOR]

Learning-based partial differential equations (PDEs), which combine fundamental differential invariants into a nonlinear regressor, have been successfully applied to several computer vision and image processing problems. However, the gradient descent method (GDM) for solving the linear combination coefficients among differential invariants is time-consuming. Moreover, when the regularization or constraints on the coefficients become more complex, it is troublesome or even impossible to deduce the gradients. In this paper, we propose a new algorithm, called fast alternating time-splitting approach (FATSA), to solve the linear combination coefficients. By minimizing the difference between the expected output and the actual output of PDEs at each time step , FATSA can solve the linear combination coefficients much faster than GDM. More complex regularization or constraints can also be easily incorporated. Extensive experiments demonstrate that our proposed FATSA outperform GDM in both speed and quality. [ABSTRACT FROM AUTHOR]

Recently a delta-density based clustering (DDC) algorithm was proposed to cluster data efficiently by fast searching density peaks. In the DDC method, the density and a new-defined criterion delta-distance are utilized. The examples with anomalously large delta-density values are treated as cluster centers, then the remaining are assigned the same cluster label as their neighbor with higher density. However there are two challenges for the DDC algorithm. First, no rules are available to judge density-delta values as “anomalously large” or not. Second, the decision graph might produce the redundant examples with “anomalous large” density-delta values, as we define as the “decision graph fraud” problem. In this paper, an improved and automatic version of the DDC algorithm, named as 3DC clustering, is proposed to overcome those difficulties. The 3DC algorithm is motivated by the divide-and-conquer strategy and the density-reachable concept in the DBSCAN framework. It can automatically find the correct number of clusters in a recursive way. Experiments on artificial and real world data show that the 3DC clustering algorithm has a comparable performance with the supervised-clustering baselines and outperforms the unsupervised DDCs, which utilize the novelty detection strategies to select the “anomalously large” density-delta examples for cluster centers. [ABSTRACT FROM AUTHOR]

Finding correspondences between pair of images of the same scene is a key problem in computer vision. When matching images undergo viewpoint change, partial occlusion, clutters and illumination change, there will be a lot of mismatches due to the limited repeatability and discriminative power of features. In this paper, we propose a robust matching algorithm that can remove false matches and propagate the correct ones to obtain more matches, thus improve the matching accuracy. First, extracts SURF (Speeded Up Robust Features) descriptors from the input two images, which can be used to build the proximity matrix. Then perform SVD (Singular Value Decomposition) on the proximity matrix to obtain the initial matches. Third, refine the initial matches by the unique property of Delaunay triangulation that can produce the maximum cliques of the two Delaunay graph. Finally, recover the lost matches with the constraint of dual graph of Voronoi. Experimental results on Oxford datasets indicate that our algorithm can improve the match performance compared to the RANSAC-based method. [ABSTRACT FROM AUTHOR]

Camera calibration is a basic and crucial problem in computer vision and photogrammetry. The traditional calibration approach based on 2D planar target fails to give reliable and accurate results due to the inaccurate localization of feature points in some calibration images. An accurate and robust estimation method for camera parameters based on RANdom SAmple Consensus (RANSAC) algorithm is proposed to exclude the unreliable images in this study. First, the distance between circular point and image of the absolute conic is defined, and the recommended threshold value is given by computer simulation. Second, RANSAC has been utilized to pick out a subset of calibrating images automatically, and linear algebraic approximation is performed to estimate the intrinsic parameters and external parameters. Finally, all the camera parameters including lens distortion parameters are refined by the non-linear searching algorithm. Numerical simulation and practical experiment in this paper demonstrate the accuracy and robustness of the proposed method. The experimental results show that the proposed method is more robust and efficient in improving the calibration accuracy than the traditional methods. [ABSTRACT FROM AUTHOR]

This paper presents an intelligent, automatically controlled camera based on visual feedback. The camera housing contains actuators that change the orientation of the camera – enabling a full rotation around the vertical axis (pan) and 90° around the horizontal axis (tilt). A system for acquisition, processing image analysis and a camera driver are implemented in the FPGA Xilinx Spartan-6 device. An original, innovative reconfigurable system architecture has been developed. The FPGA device is connected directly to the eight independently operated SRAM memory banks. A prototype device has been constructed with a real-time tracking algorithm, enabling an automatically control of the position of the camera. The device has been tested indoors and outdoors. The camera is able to keep a tracked object close to the center of its field of view. The power consumption of the control system is 2 W. A reconfigurable part reaches the computing performance of 3200 MOPS. [ABSTRACT FROM AUTHOR]

This paper proposes two exposure based recursive histogram equalization methods for image enhancement. The proposed methods are very effective for images acquired in low light condition like underwater sequences or night vision images. The first method is recursive exposure based sub-image histogram equalization (R-ESIHE) that recursively performs ESIHE [20] method till the exposure residue among successive iteration is less than a predefined threshold. The second method is named as recursively separated exposure based sub image histogram equalization (RS-ESIHE) that performs the separation of image histogram recursively; separate each new histogram further based on their respective exposure thresholds and equalize each sub histogram individually. The experimental results show that low exposure image enhancement problem was not addressed by earlier HE based methods, has been efficiently handled by these new methods. The performance evaluation of new methods is done in terms of image information content as well as visual quality inspection. The proposed methods outperforms earlier HE based contrast enhancement algorithms specifically for low light images. [ABSTRACT FROM AUTHOR]

This paper proposes a novel method to address the registration of images with affine transformation. Firstly, the Maximally Stable Extremal Region (MSER) detection method is performed on the reference image and the image to be registered, respectively. And the coarse affine transformation matrix between the two images is estimated by the matched MSER pairs. Two circular regions containing roughly the same image content are also obtained by fitting and normalizing the centroids of the matched MSERs from the two images. Secondly, a scale invariant and approximate affine transformation invariant feature point detection algorithm based on the Gabor filter decomposition and phase congruency is performed on the two coarsely aligned regions, and two feature point sets are achieved, respectively. Finally, the affine transformation matrix between the two feature point sets is obtained by using a probabilistic point set registration algorithm, and the final affine transformation matrix between the reference image and the image to be registered is achieved according to the coarse affine transformation matrix and the affine transformation matrix between the two feature point sets. Several sets of experiments demonstrate that our proposed method performs competitively with the classical scale-invariant feature transform (SIFT) method for images with scale changes, and performs better than the traditional MSER and Affine-SIFT (ASIFT) methods for images with affine distortions. Moreover, the proposed method shows higher computation efficiency and robustness to illumination change than some existing area-based or feature-based methods do. [ABSTRACT FROM AUTHOR]

Establishing correct correspondences between two faces with different viewpoints has played an important role in 3D face reconstruction and other computer-vision applications. Usually, face images are considered to lack sufficient distinctive features to establish a large number of correspondences on uncalibrated images. In this paper, we investigate pore-scale facial features, which are formed from pores, fine wrinkles, and hair. These features have many characteristics that make them suitable for matching facial images under different variations. Using both biological observation and computer-vision consideration, a new framework is devised for pore-scale facial-feature extraction and matching. The matching difficulty under various skin appearances of different subjects and imaging distortion is also analyzed. For further improving the matching performance and tackling distortions such as varying illuminations and unfocused blurring, a pore-to-pore correspondences dataset is established for training a more distinctive and compact descriptor. Experiments are conducted on a face database containing 105 subjects, and the results prove that the pore-scale features are highly distinctive; face images with a minimum resolution of 600×700 (0.4 mega) pixels contain sufficient details to perform a reliable matching in different poses. Generally, our algorithm can establish between 500 and 2000 correct correspondences on a pair of uncalibrated face images of the same person. Furthermore, the proposed methods can be applied to face recognition, 3D reconstruction, etc. [ABSTRACT FROM AUTHOR]

The image parsing process gives labels to image regions, as well as information including shape, semantics and context. Although it is one of the most important features in the human visual system, automatic image parsing using computer vision techniques remains difficult due to computational issue. In this paper we introduce a novel method to address this limitation. Our system models an image as a set of regions and uses a novel hypotheses generation algorithm to get possible image parsing solutions for final re-scoring. Our proposed hypotheses generation algorithm, called Loopy Dynamic Programming (LDP), handles large search space efficiently and gives good parsing hypotheses for testing. With such capacity, we are able to apply more precise and complex image models to achieve better performance. In addition, our system can perform image segmentation, detection and classification simultaneously. Experimental results using Pascal VOC 2007 dataset show that the proposed technique achieves very promising performance in all the three tasks. [ABSTRACT FROM AUTHOR]

The tilt angle of a machine is important for its stability and technological operations. Most sensors are gravity or inertial based, and their outputs can be affected by vibration induced errors, random errors or drift. A sensor that works on a different physical principle in collaboration with these sensors could decrease the errors and thus increase the trustworthiness of the measurement. This paper deals with the optical determination of the machine’s tilt angle, using camera images of the trees. This method assumes the surrounding forest, in the machine’s location, as a base for the analysis and it is focused on the feasibility of the idea. The Sobel algorithm for edge detection is used to process the inputs from the camera. A simple and fast algorithm inspired by the fringe pattern has been designed. The results show that the initial assumption of the upright position of the trees in the direction of gravity is sufficient and can be used to determine the machine’s position on a slope, with an average error in the range of 0.63–1.63°. This method can also be utilized to determine a vehicle’s or robot’s inclinations in an urban environment, which is possible without the modification of the method. [ABSTRACT FROM AUTHOR]

Image enhancement can accentuate image feature and is necessary process in image processing. This work focuses on fusing multi-exposure image sequences low-light image enhancement. Inspired by the classical non-local means in computer vision, we proposed an improved deep neural network framework with attentions for image enhancement. Firstly, the original image was preprocessed in different dimensions. we get the edge images using an edge extracted algorithm and fusion multi exposed images to get an better initial images based on fully convolutional neural network with position and channel attention mechanism. Secondly, the head network is constructed by fully convolutional neural network. For capturing long-range dependencies between features maps, we designed a non-local attention module for head network to get better enhancement image. Finally, emerging the original images, edge image and fusion image as the input of the head network, it can enhance the images to get high-quality images. Experiments show that our framework proposed in this paper is effective and the attention mechanism play a significant hole in the network. • Designed channel and space attention mechanism to fusion multi exposure images. • Designed non-local attention mechanism in feature maps domain for enhancement network. • Proposed an emerging deep neural network framework with multiple attention mechanism. [ABSTRACT FROM AUTHOR]

Superpixel is one of the most popular image over-segmentations with broad applications in the computer vision field to reduce their computations by replacing pixels as primitives. The main concerns of one superpixel generation algorithm are its accuracy and efficiency. One of the most important things in superpixel accuracy is to fit the image boundaries tightly with a few pixels as possible (namely minimal contour density, which is measured by the percent of superpixel contour pixels in the whole image). In this paper, we propose a new fast algorithm based on the clustering method to produce superpixels accurately with low contour density. First, we adopt the linear path from a pixel to one superpixel seed to define a regular term and propose a new distance measurement between them. In addition, we introduce the gradient and Local Binary Pattern (LBP) features and propose formulas of parameters in the proposed method adaptively. In this way, we can use the new distance measurement to group pixels as initial regions adaptively and produce the final superpixels by merging those small ones. Finally, we test the new algorithm on two public datasets and compare it with the state-of-the-art. Our method can generate superpixels with lower contour density while being competitive in accuracy and computational time. [ABSTRACT FROM AUTHOR]

Harris corner detection is an algorithm frequently used in image processing and computer vision applications to detect corners in an input image. In most modern applications of image processing, there is a need for real time implementation of algorithms such as Harris corner detection in hardware systems such as field-programmable gate arrays (FPGAs). FPGAs allow faster algorithmic throughput, which is required to match real time speeds or cases where there is a requirement to process faster data rates. High level synthesis tools offer higher abstraction level to designers with continued verification during the design flow and hence are getting popular with the design community. This paper proposes a high speed and area optimized implementation of a Harris corner detection algorithm. The proposed implementation was actualized using a novel high-level synthesis (HLS) design method based on application-specific bit widths for intermediate data nodes. Register transfer level (RTL) code was generated using MATLAB HDL coder for HLS. The generated hardware description language (HDL) code was implemented on Xilinx ZedBoard using Vivado software and verified for functionality in real time with input video stream. The obtained results are superior to those of previous implementations in terms of area(smaller gate count on target FPGA) and speed for the same target board. [ABSTRACT FROM AUTHOR]

Image Contrast Enhancement (ICE) is a crucial step in several image processing and computer vision applications. Its main objective is to improve the quality of the visual information contained in the processed images. The presence of noise and small sets of pixels in images are not only irrelevant for their visualization. It also negatively affects the improvement process of ICE schemes since the inclusion of irrelevant information avoids the appropriate distribution of significant pixel intensities in the enhanced image. As a consequence of this effect, most of the proposed ICE methods present different associated problems such as the production of undesirable artifacts, noise amplification, over saturation and bad human visual perception. In this paper, an Image Contrast Enhancement (ICE) method for grayscale and color images is presented. The proposed approach has the propriety of eliminating noisy and irrelevant information in order to improve the distribution capacity of significant pixel intensities in the enhanced image. Our method eliminates multiple groups of a very small number of pixels that, according to their characteristics, do not represents any object or important detail of the image. This process is done by the Mean-shift algorithm, which is used to replace such sets of irrelevant pixels in the original histogram by significant pixel densities represented by local maxima. Then, the Moth Swarm Algorithm (MSA) is used to redistribute the pixel intensities of the reduced histogram so that the value from Kullback–Leibler entropy (KL-entropy) has been maximized. The proposed approach has been tested considering different public datasets commonly used in the literature. Its results are also compared with those produced by other well-known ICE techniques. Evaluation of the experimental results demonstrates that the proposed approach highlights the important details of the image also improving its human visual appearance. • It uses an elimination process to reduce irrelevant features. • It uses the KL-entropy for the enhancement process. • It considers the use of the MSA to find the best enhanced image. • Its results are competitive in comparison with other state-of-the art methods. [ABSTRACT FROM AUTHOR]

This paper proposes an automatic crack detection, localization, and quantification method using the integration of a faster region proposal convolutional neural network (Faster R-CNN) algorithm to detect crack regions. The regions were located using various bounding boxes and a modified tubularity flow field (TuFF) algorithm to segment the crack pixels from the detected crack regions. A modified distance transform method (DTM) was used to measure crack thickness and length in terms of pixel measurement. To validate the proposed method, 100 images were taken in different places with complex backgrounds containing different angles and distances between the camera and the objects. The results obtained from the Faster-R-CNN-based crack damage detection had a 95% average precision. The pixel-level segmentation performance of the modified TuFF algorithm exhibited an authentic outcome, with 83% intersection over union. Finally, the modified DTM algorithm provided 93% accuracy with respect to crack length and thickness with a 2.6 pixel root mean square error. • The first proposal of an automatic hybrid crack segmentation and quantification under complex backgrounds and different lightening conditions. • The hybrid method successfully removes the limitation of independent application of deep learning and image processing method. • The proposed method achieved the AP of the Faster R-CNN is 95%, and the IoU of the modified TuFF with CLAHE is 83%. • The original DTM method is improved to measure the thickness and length of the segmented cracks with an RMS error of 2.6 pixels, providing crack length accuracy of 93%. [ABSTRACT FROM AUTHOR]