Search

Clues and traces of the universe's origin and its developmental process are deeply buried in galaxy shapes and formations. Automated galaxies classification from their images is complicated due to the faintness of the galaxy images, conflicting bright background stars, and image noise. For this purpose, the current work proposes a novel logically structured modular algorithm that analyses galaxy morphological raw brightness data to automatically detect galaxy visual center, region, and classification. First, a novel selective brightness threshold is employed to eliminate the effect of bright background stars on detecting galaxy visual centers. Second, a galaxy region detection technique is developed. Finally, a novel technique based on galaxy brightness variation patterns is employed for galaxies classification. The current work has been tested with a run on a collection of 1000 galaxies from the EFIGI catalog. Results demonstrated a success rate of 97.2% in galaxies classification with an average processing time of 0.37 s per galaxy. The high success rates and the low processing time proved the efficiency of the proposed work.

Analysing the human voice has always been a challenge to the engineering society for various purposes such as product review, emotional state detection, developing AI, and much more. Two basic grounds of voice or speech analysis are to detect human gender and the geographical region based on accent. This study presents a three-layer feature extraction method from the raw human voice to detect the gender as male or female, as well as the region from where that voice belongs. Fundamental frequency, spectral entropy, spectral flatness, and mode frequency have been calculated in the first layer of feature extraction. On the other hand, Mel Frequency Cepstral Coefficient has been used to extract the features in the second layer and linear predictive coding in the third layer. Regular voice contains some noises which have been removed with multiple audio data filtering processes to get noise-free smooth data. Multi-Output-based 1D Convolutional Neural Network has been used to recognize gender and region from a combined dataset which consists of TIMIT, RAVDESS, and BGC datasets. The model has successfully predicted the gender with 93.01% and region with 97.07% accuracy. This method works better than usual state-of-the-art methods in separate datasets along with the combined dataset on both gender and region classification.

Detection of lane region under the road boundary is an imperative module for intelligent vehicle system. Lane markings provide separate regions on the road for the vehicles to avoid the possibility of accidents. Existing methods in lane detection have limited performance using various sensor-based approaches such as Radar and LiDAR and have high operational costs. To achieve a steady and optimal lane detection, the vision-based lane region detection scheme VLDNet is proposed which utilizes a encoder-decoder network using semantic segmentation architecture. In this direction, a hybrid model using UNet and ResNet has been adopted, where UNet is used as a segmentation model and ResNet-50 is used for down-sampling the image and identifying the required features. These identified features have been then applied into UNet for up-sampling and decoding the segments of the images. The publicly available KITTI dataset have been accessed for experiments and validation of the proposed network. The method outperforms the existing state-of-the-art methods in lane region detection. The network achieves better performance using standard evaluation measures such as accuracy of 98.87%, the precision of 98.24%, recall of 96.55%, frequency weighted IoU of 97.78%, and MaxF score of 97.77%.

Stair region detection and recognition from a stair candidate image is a challenging work in the computer vision research area. In the last few decades, researchers use many recognition systems to recognize and verify the stair region from other analogous objects. However, all the verification systems such as vanishing point (VP) do not achieve the desired result for various reasons. In this regard, a method is proposed in this paper to investigate the vanishing point’s problem arising in the case of stair region verification based on the three basic criteria, i.e. focal angle of the camera, height of the camera from the ground, and distance of the camera from the stair image. For that, primarily, the stair region is extracted by utilizing the geometrical features of a stair. The detected stair candidate region is verified through the y coordinate value of the vertical VP, i.e. y y coordinate value of VP does not verify the stair region from all the scenarios. This paper investigates and justifies this problem utilizing the experimental analysis and introduces a mathematical model to estimate the location of the VP of the stair region. Finally, support vector machine (SVM) classifier is utilized instead of VP to recognize the stair candidate region and the performance of SVM is compared with respect to the VP. For that, rotational invariant uniform local binary pattern (LBP) is used for feature extraction. Stair images captured under different orientation and illumination conditions have been used to test the proposed method to evaluate the resultant accuracy.

In this paper, a hybrid visual attention model to effectively detect a distant target is proposed. The model employs the human visual attention mechanism and consists of two models, the training model, and the detection model. In the training model, some of the features are selected to train in the process of extracting and combining the early visual features from the training image of the target by bottom-up manner, and these features are trained and accumulated as trained data. When the image containing the target is input into the detection model, a task of selectively promoting only features of the target using pre-trained data is performed. As a result, the desired target is detected through the saliency map created as a result of the feature combination. The model has been tested on various images, and the experimental results demonstrate that the proposed model detected the target more accurately and faster than other previous models.

As PCNN is modulated by using the pulse-coupled synaptic mechanisms, it has a great potential for image processing in a complex real-world environment, especially in images. A new simplified pulse coupled neural network (SPCNN) is proposed. This new model uses the pixel intensity with the actual physical meanings as the input parameters instead of the abstract network parameters in the original SPCNN. In order to achieve this goal, we try to derive the general formulae of dynamic threshold and internal activity of the SPCNN according to the dynamic properties of neurons and then deduce the relationship between the pixel intensity and the abstract parameters. Then, the relationship is transformed into an objective optimization problem to obtain the appropriate abstract parameters. Finally, extensive experiments are conducted on seven widely used datasets to demonstrate the effectiveness of the proposed method and shown improvement on the salient region detection.

This paper proposes a novel multi-task learning based salient region detection method by fusing spatial and temporal features. Salient region detection has been widely used in various computer vision tasks, being as a general preprocessor to identify interest objects. Despite the recent successes, existing saliency models still lag behind the performance of human when visually perceives dynamic scenes. Most of the existing models largely rely on various spatial features. However, these spatial feature based methods have several deficiencies: (i) they can hardly adapt to the situation where moving objects are included, and (ii) they cannot model the human vision process in dynamic scenes. Recently, some saliency models introduce temporal features in their detecting process, such as the optical flow and stacking frames. The potential of temporal features for saliency optimization has been demonstrated. However, since temporal features in these models are merely used as a compensation to static features, the advantages of temporal features have not yet been fully explored. Aiming to comprehensively address these issues above, our method fuses spatial and temporal features, and learns the mapping relationship from various features to salient regions using our multi-task learning framework. The final salient region is generated by our unified Bayesian framework. The experimental results demonstrated that our proposed approach outperforms previous methods.

In this paper, we will present a new salient region detection method by exploiting its surrounding and superpixel cues. Its main highlights are: 1) An input image is quantized to 256 colors by using minimum variance quantization; 2) Saliency maps is computed based on the figure-ground segregation of the quantized image; 3) Mean saliency value of each superpixel is employed to refine saliency maps further. This can highlight salient objects robustly and suppress backgrounds evenly. Experimental results show that the proposed method produces more accurate saliency maps and performs well against twenty-one saliency models concerning three evaluation metrics on two public datasets.

Accurate recognition of polyps is crucial for early colorectal cancer diagnosis and treatment. Wireless capsule endoscopy (WCE) is a noninvasive, wireless imaging tool that allows direct visualization of the entire colon without discomfort to patients and has the potential to revolutionize the screening workup for colorectal diseases. However, current manual review is laborious and time consuming, requiring the undivided concentration of the gastroenterologist. Computational methods that can assist automated polyp recognition will enhance the outcome both in terms of diagnostic accuracy and efficiency of WCE. This review introduces the computer-assisted algorithms as applied to colorectal polyp screening, focusing on the successes of deep-learning-based strategies in the WCE sequences. We survey key applications of WCE polyp recognition, covering deep-learning-based image-level classification, lesion region detection, and pixel-accurate segmentation. We conclude by discussing emerging research challenges, possible trends, and future directions.

Tunnel lining internal defect detection is essential for the safe operation of tunnels. This paper presents an automatic scheme based on rotational region deformable convolutional neural network (R2DCNN) and Ground Penetrating Radar (GPR) images for the accurate detection of defects and rebars with arbitrary orientations. The R2DCNN comprises inter-related modules, specifically, deformable convolution, feature fusion, and rotated region detection modules. In this study, synthetic GPR images, including rebars and various structural defects with different permittivities, as well as real GPR images obtained by model experiments, were constructed for the R2DCNN. Improved results were obtained while testing the R2DCNN on GPR images in comparative experiments. The mean average precision of the R2DCNN was enhanced by 8.21% compared to the R2CNN on synthetic GPR images. The R2DCNN showed satisfactory results in on-site experiments, which demonstrated the applicability of the R2DCNN to practical tunnels.

Smoke detection is very important for fire detection, because smoke usually occurs before a fire breaks out. Smoke has a variety of characteristics, such as color, shape, texture, and so on. Using these characteristics, there has been a lot of research on smoke detection based on image processing. However, smoke detection is still difficult, because these characteristics can change due to environmental factors, such as wind and lighting changes. So, in this paper, a new smoke detection method based on fire-starting regions is presented to reduce sensitivity to environmental factors. The proposed method consists of three main parts: i) generating a candidate smoke region, ii) fire starting?region detection, and iii) detecting smoke based on fire-starting regions and predicting the direction of smoke movement. As a result, the proposed smoke detection method is less sensitive to the motion of non-smoke objects (humans, shadows, vehicles, etc.) and can detect smoke quickly by determining the fire-starting region where the smoke is generated. The proposed detection method can be applied to a video surveillance system and a fire- and smoke-analysis system.

Based on the attention mechanism of human vision, an image salient region detection model is proposed by fusing multi-scale and log amplitude spectrum. The model considers the characteristics of target and background at the same time, calculates the spectral residuals at different scales, and uses “center surround difference operation” for cross scale combination. According to feature integration, multi-scale feature maps are superimposed to get the final saliency map. The simulation results show that compared with the traditional detection algorithm, the improved algorithm in this paper has obvious improvement in the detection effect and accuracy of the salient region, and reduces the misjudgment rate. The salient region detection for the aerial sea detection image has certain anti-interference and applicability.

Breast cancer is one of the most common death causes amongst women all over the world. Early detection of breast cancer plays a critical role in increasing the survival rate. Various imaging modalities, such as mammography, breast MRI, ultrasound and thermography, are used to detect breast cancer. Though there is a considerable success with mammography in biomedical imaging, detecting suspicious areas remains a challenge because, due to the manual examination and variations in shape, size, other mass morphological features, mammography accuracy changes with the density of the breast. Furthermore, going through the analysis of many mammograms per day can be a tedious task for radiologists and practitioners. One of the main objectives of biomedical imaging is to provide radiologists and practitioners with tools to help them identify all suspicious regions in a given image. Computer-aided mass detection in mammograms can serve as a second opinion tool to help radiologists avoid running into oversight errors. The scientific community has made much progress in this topic, and several approaches have been proposed along the way. Following a bottom-up narrative, this paper surveys different scientific methodologies and techniques to detect suspicious regions in mammograms spanning from methods based on low-level image features to the most recent novelties in AI-based approaches. Both theoretical and practical grounds are provided across the paper sections to highlight the pros and cons of different methodologies. The paper’s main scope is to let readers embark on a journey through a fully comprehensive description of techniques, strategies and datasets on the topic.

As smart devices are used to assist in understanding presentation slides, such as machine translation and augmented reality, in seminars and conferences, separating the projection region from the foreground and background will improve the accuracy of recognition of slides. To take on this challenge, this paper proposes a novel detection model, BASNet with channel and space attention (CS-BASNet), to accurately detect projection area. Firstly, channel module and space attention module are added to the backbone network of ResNet18 to improve the detection evaluation metrics. Secondly, to reduce the influence of color on detection, we used gray images as inputs to train the model. We also propose a projection region dataset which sets projection region as foreground. The test results on this dataset show that the evaluation metrics of CS-BASNet is better than that of BASNet, and it showed that the evaluation metrics of CS-BASNet in detecting projection region has outperformed the latest five salient object detection models.

Salient regions provide important cues for scene understanding to the human vision system. However, whether the detected salient regions are helpful in image blur estimation is unknown. In this study, a salient region guided blind image sharpness assessment (BISA) framework is proposed, and the effect of the detected salient regions on the BISA performance is investigated. Specifically, three salient region detection (SRD) methods and ten BISA models are jointly explored, during which the output saliency maps from SRD methods are re-organized as the input of BISA models. Consequently, the change in BISA metric values can be quantified and then directly related to the difference in BISA model inputs. Finally, experiments are conducted on three Gaussian blurring image databases, and the BISA prediction performance is evaluated. The comparison results indicate that salient region input can help achieve a close and sometimes superior performance to a BISA model over the whole image input. When using the center region input as the baseline, the detected salient regions from the saliency optimization from robust background detection (SORBD) method lead to consistently better score prediction, regardless of the BISA model. Based on the proposed hybrid framework, this study reveals that saliency detection benefits image blur estimation, while how to properly incorporate SRD methods and BISA models to improve the score prediction will be explored in our future work.

This paper aims to develop an efficient method of remote sensing object detection with deep learning in an end-to-end manner. We proposed a method of region detection followed by target detection to detect small-scale targets from large-scale remote sensing images, with a light-weighted neural network trained on our own dataset. For a simulation of practical detection algorithm deployment, which requires both accuracy and efficiency on hardware of limited compute power, the whole algorithm was implemented and tested on NVIDIA Jetson AGX Xavier embedded platform.

We are pleased to introduce a first-of-its-kind algorithm that combines in-silico region detection and spatial gene-expression imputation. Spatial transcriptomics by 10X Visium (ST) is a new technology used to dissect gene and cell spatial organization. Analyzing this new type of data has two main challenges: automatically annotating the major tissue regions and excessive zero values of gene-expression due to high dropout rates. We developed a computational tool—MIST—that addresses both challenges by automatically identifying tissue regions and estimating missing gene-expression values for individual tissue regions. We validated MIST detected regions across multiple datasets using manual annotation on the histological staining images as references. We also demonstrated that MIST can accurately recover ST’s missing values through hold-out experiments. Furthermore, we showed that MIST could identify subtle intra-tissue heterogeneity and recover spatial gene-gene interaction signals. We therefore strongly encourage using MIST prior to downstream ST analysis because it provides unbiased region annotations and enables accurately de-noised spatial gene-expression profiles.

Traditional deep learning-based vehicle detection methods are often designed using a pyramid of filters with multiple scales and sizes; therefore, the processing time is slow due to the large number of scales used and because the classifier runs at all scales. Recently, a deep learning-based region proposal network was introduced to detect vehicles that only employ the network one time regardless of the size of the input image. In object detection, deep learning-based region proposal networks have achieved state-of-the-art performance in terms of accuracy. These systems achieve a very high accuracy under normal driving conditions; however, their performance decreases under difficult driving conditions such as in snow, rain, or fog. In addition, the current state-of-the-art system-based region proposal networks still fail to satisfy the real-time requirements of the driving assistant systems. More recently, the identification of local patterns has been shown to improve the performance of the traditional deep-learning systems; hence, this paper investigates local patterns in region proposal networks to improve their accuracy. Depth information is also investigated to improve the processing time of current region proposal networks. Our experimental results show that the proposed system obtains better performance than the state-of-the-art object region detection systems in terms of both accuracy and processing time.

In this paper, a motion region detection and pattern extraction algorithm is proposed for motion recognition based on the image as the distance changes between the radar sensor and human motion. This algorithm, which was developed by combining the characteristics of the radar motion image and image processing techniques, compensates for the discontinuous data that can occur when removing clutter for motion data detection and extracts motion patterns applicable to the recognition learning field. According to the motion data measured by a radar sensor, the algorithm performance is verified by an experiment.

Video summarization condenses a video by extracting its informative and interesting segments. In this article, a novel video summarization approach is proposed based on spatiotemporal salient region detection. The proposed approach first segments a video into a set of shots which are ranked with spatiotemporal saliency scores. The score for a shot is computed by aggregating the frame level spatiotemporal saliency scores. This approach detects spatial and temporal salient regions separately using different saliency theories related to objects present in a visual scenario. The spatial saliency of a video frame is computed using color contrast and color distribution estimations and center prior integration. The temporal saliency of a video frame is estimated as an integration of local and global temporal saliencies computed using patch level optical flow abstractions. Finally, top ranked shots with the highest saliency scores are selected for generating the video summary. The objective and subjective experimental results demonstrate the efficacy of the proposed approach.

Endoscopic ultrasonography (EUS) has shown great advantages in the diagnosis and staging of gastrointestinal malignant tumors. However, EUS based diagnosis is limited by variability in the examiner’s subjective interpretation to differentiate between normal and early esophageal carcinoma. In this paper, we propose a novel approach aiming at automatic abnormal region detection from the esophageal EUS images; the contribution is three-fold: first, we present a series of preprocessing strategies developed specifically for the enhancement of EUS images to aid the estimation in the subsequent works. Second, we provide an automatic layer segmentation method based on the multiple surface graph searching approach with incorporation of geometric constraints, which is applied to segment the EUS images into five discernible layers. Third, we introduce the novel feature extraction strategy by utilizing the features from each column in the segmented layers. The SVM classifier is then applied to fulfill the normal and early esophageal carcinoma classification. Subsequently, a clustering method is used to assemble the abnormal columns together so as to detect the abnormal region. Experimental results show that our method has demonstrated its robustness even facing noisy EUS images, and has achieved high accuracy in detecting abnormal region.

Being an efficient method of information fusion, multi-focus image fusion has attracted increasing interests in image processing and computer vision. This paper proposes a multi-focus image fusion method based on focus region detection using mean filter and guided filter. Firstly, a novel focus region detection method is presented, which uses guided filter to refine the rough focus maps obtained by mean filter and difference operator. Then, An initial decision map is got via the pixel-wise maximum rule, and optimized to generate final decision map by using guided filter again. Finally, the fused image is obtained by the pixel-wise weighted-averaging rule with the final decision map. Experimental results demonstrate that the novel focus region detection method has stronger robustness to different noises, and higher computational efficiency than other focus measures. Furthermore, the proposed fusion method implements efficiently and outperforms some state-of-the-art approaches both in visual effect and objective evaluation.

In recent years, extensive research is carried out in computer assisted interpretation carried out for cancer classification. Computer aided Interpretations are involves with pre-processing, contrast enhancement, segmentation, appropriate feature extraction and classification. Though considerable research is carried out in developing contrast enhancement and image segmentation techniques, cancer regions could not be isolated and extracted efficiently. Hence this work focuses on developing efficient image segmentation techniques for isolating the cancer region and also identifying suitable descriptors for describing the cancer region. Hence this work focuses to introduce a simple and easy approach for detection of cancerous tissues in mammals. Detection phase is followed by segmentation of the region in an image. Our approach uses simple image processing techniques such as averaging and thresholding along with a Max-Mean and Least-Variance technique for cancer detection. Experimental results demonstrate the effectiveness of our approach.

Document images are more popular in today’s world and being made available over the internet for Information retrieval. The document images becomes a difficult task compared with digital texts and edge detection is an important task in the document image retrieval, edge detection indicates to the process of finding sharp discontinuation of characters in the document images. The single edge detection methods causing the weak gradient and edge missing problems adopts the method of combining global with local edge detection to extract edge. The global edge detection obtains the whole edges and uses to improve adaptive smooth filter algorithm based on canny operator. These combinations increase the detection efficiency and reduce the computational time. In addition, the proposed algorithm has been tested through real-time document retrieval system to detect the edges in unstructured environment and generate 2D maps. These maps contain the starting and destination points in addition to current positions of the objects. This proposed work enhancing the searching ability of the document to move towards the optimal solution and to verify the capability in terms of detection efficiency.

Residual yarn detector plays an important role in the pipeline of spinning-linked winding systems. This research proposed an image-based method to improve the traditional detectors who have weaknes...

Anomaly region detection aims at finding spatial outliers or spatial anomalous clusters. Generally, detection approaches cover spatial neighbor's discovery with spatial attributes and anomaly measurement of spatial regions according to non-spatial attributes. In this article, an anomaly region detection method using Delaunay minimal spanning tree (DMST for short) is proposed. First, a Delaunay minimal spanning tree is constructed. Then, the current longest edge of the tree is iteratively cut and anomaly regions are concurrently detected. Finally, the shortest edge of the related bipartite graph is taken as the anomaly measurement. The proposed method could avoid the disturbance of bad reference neighbors and generate anomaly regions keeping atomicity.

To deal with the serious visual artifacts caused by the consistent restoration algorithm on local motion blurred images, a novel two stage blur region detection method including coarse location and region refinement for local blurred images is proposed in this paper. First, blur feature discriminant criteria in frequency domain and spatial domain is defined to generate blur feature image; then, binarization is used on the blur feature image to obtain the coarse blur region detection map; subsequently, morphological methods are used to process the clear and blur areas and the trimap is obtained. Finally, the refinement detection of blur region is achieved by combination of the trimap and automatic KNN matting algorithm. Experimental results show that the proposed algorithm can detect the local blurred region quickly and effectively, and it has outperformed other traditional methods in the application of blur region detection and image restoration as well.

The moving object detection from a stationary video sequence is a primary task in various computer vision applications. In this proposed system; three processing levels are suppose to perform: detects moving objects region from the background image; reduce noise from the pixels of detected region and extract meaningful objects and their features (area of object, center point of area etc.). In this paper; background subtraction techniques is used for segments moving objects from the background image, which is capable for pixel level processing. Morphology operation (Erosion and dilation) are used to remove pixel to pixel noise. In last level, CCL algorithm is used for sorts out foregrounds pixels are grouped into meaningful connected regions and their features.

In recent times, there has been a tremendous progress in image dehazing for computer vision applications, while the sky region processed by these algorithms tends to degrade by noise and color distortion. In this paper, an improved dark channel prior algorithm is proposed which detects the sky region first and divides the image into sky region and non-sky region and then estimates the transmission of two parts separately, followed by combining with refining step. The proposed algorithm also accurately corrects the transmission of the sky region to avoid noise and color distortion. Experimental results show a greater quality improvement in the output images than the existing strategies.

International audience; Precisely detecting solar Active Regions (AR) from multi-spectral images is a challenging task yet important in understanding solar activity and its influence on space weather. A main challenge comes from each modality capturing a different location of these 3D objects, as opposed to more traditional multi-spectral imaging scenarios where all image bands observe the same scene. We present a multi-task deep learning framework that exploits the dependencies between image bands to produce 3D AR detection where different image bands (and physical locations) each have their own set of results. We compare our detection method against baseline approaches for solar image analysis (multi-channel coronal hole detection, SPOCA for ARs (Verbeeck et al., 2013)) and a state-of-the-art deep learning method (Faster RCNN) and show enhanced performances in detecting ARs jointly from multiple bands.

The automatic localization and classification of defects on surfaces helps to ensure the quality of industrially manufactured products. For the development of such automatic systems, a measure is needed that allows a profound optimization and comparison. However, there is currently no measure dedicated to surface defect localization in particular and measures from related fields are unsuitable. Thus, we present the Region Detection Rate (RDR) which is specialized on defect localization since it is evaluated in a defect-wise manner. It entails a set of rules that define the circumstances under which a defect is considered detected and a prediction is considered a false positive. The usability of the RDR is qualitatively demonstrated on examples from three different datasets, one of which has been annotated as part of this work. We hope that the new measure supports the development of future automatic surface defect localization systems and to raise a discussion about the suitability of measures with regard to this task.

y publicar datos. Por ejemplo, en el Internet se halla un enorme número de tablas publicadas en HTML integradas en documentos PDF, o que pueden ser simplemente descargadas de páginas web. Sin embargo, las tablas no siempre son fáciles de interpretar pues poseen una gran variedad de características y son organizadas en diferentes formatos. De hecho, se han desarrollado muchos métodos y herramientas para la interpretación de tablas. Este trabajo presenta la implementación de un algoritmo, basado en campos aleatorios condicionales (CRF, Conditional Random Fields), para clasificar las filas de una tabla como fila de encabezado, fila de datos y fila metadatos. La implementación se complementa con dos algoritmos para reconocer tablas en hojas de cálculo, específicamente, basados en reglas y detección de regiones. Finalmente, el trabajo describe los resultados y beneficios obtenidos por la aplicación del algoritmo para tablas HTML, obtenidas desde la web y las tablas en forma de hojas de cálculo, descargadas desde el sitio de la Agencia Nacional de Petróleo de Brasil.//Tables are widely adopted to organize and publish data. For example, the Web has an enormous number of tables, published in HTML, embedded in PDF documents, or that can be simply downloaded from Web pages. However, tables are not always easy to interpret due to the variety of features and formats used. Indeed, a large number of methods and tools have been developed to interpreted tables. This work presents the implementation of an algorithm, based on Conditional Random Fields (CRFs), to classify the rows of a table as header rows, data rows or metadata rows. The implementation is complemented by two algorithms for table recognition in a spreadsheet document, respectively based on rules and on region detection. Finally, the work describes the results and the benefits obtained by applying the implemented algorithm to HTML tables, obtained from theWeb, and to spreadsheet tables, downloaded from the Brazilian National Petroleum Agency.

This work presents a vision based security robot perception and control strategy for semi-structured and unstructured roads navigation. The main contributions contain deep learning technique for road recognition and a hybrid navigation scheme. A deep convolutional neural network is employed to perform pixel-wise segmentation and thus to find road regions. Secondly, based on the segmented regions, an edge extraction algorithm is designed to extract and fit the road boundaries. To ensure the robustness of navigation, the region detection algorithm is proposed to ensure the robot to movement on the traversable area. Experimental results verify the effectiveness of proposed visual navigation approaches.

© Springer International Publishing Switzerland 2016. Image analysis is a key area in the computer vision domain that has many applications. Genetic Programming (GP) has been successfully applied to this area extensively, with promising results. Highlevel features extracted from methods such as Speeded Up Robust Features (SURF) and Histogram of Oriented Gradients (HoG) are commonly used for object detection with machine learning techniques. However, GP techniques are not often used with these methods, despite being applied extensively to image analysis problems. Combining the training process of GP with the powerful features extracted by SURF or HoG has the potential to improve the performance by generating high-level, domain-tailored features. This paper proposes a new GP method that automatically detects different regions of an image, extracts HoG features from those regions, and simultaneously evolves a classifier for image classification. By extending an existing GP region selection approach to incorporate the HoG algorithm, we present a novel way of using high-level features with GP for image classification. The ability of GP to explore a large search space in an efficient manner allows all stages of the new method to be optimised simultaneously, unlike in existing approaches. The new approach is applied across a range of datasets, with promising results when compared to a variety of well-known machine learning techniques. Some high-performing GP individuals are analysed to give insight into how GP can effectively be used with high-level features for image classification.

Detecting ellipses accurately and efficiently for real-world images is crucial for various visual-based applications. Most existing methods employ detection strategies throughout images, while most time is spent on the non-ellipse region. Meanwhile, small ellipses are often miss-detected due to the low resolution and fixed parameters of detectors. In this paper, we proposed an effective ellipse detector benefiting from the region detection method, which provides a basic estimation on the region and size of ellipses. Then, a two-level arc pruning strategy is proposed to detect ellipses efficiently while limiting false-positive and false-negative results. Furthermore, for the pre-estimated region without detected ellipses, interpolation method is employed to enlarge the target region, which makes small and blur ellipses to be detected. Experimental results demonstrate that the proposed method achieves competitive accuracy compared with the state-of-the-art methods.

High quality document layout analysis is fundamental to the accurate processing of handwritten textual material, on both the level of individual lines and higher order zones demarking textual and non-textual content. We present an artificial neural network based approach to prediction of either that is implemented as part of a libre optical character recognition package and highly reconfigurable for a variety of tasks. Experiments on different openly available datasets show competitive results to state-of-the-art methods.

The algorithms based on decomposition are regarded as a promising optimizer for the multi-objective optimization problems (MOPs). However, it is difficult for the algorithms based on decomposition to handle MOPs with the complicated feature, because they adopt the fixed weight vectors. In this paper, we propose an adaptive method with the region detection strategy for the decomposition-based evolutionary algorithm (aMOEA/D-RD) to adjust the weight vectors. In the proposed algorithm, some useless weight vectors, which are not associated to any solution in the successive generations, are found through the region detection method. Then these weight vectors are adaptively adjusted by the solutions in the most crowed subregion. After the adjustment of weight vectors, the distribution of the weight vectors can be more suit to approximate the true the Pareto optimal front of MOPs. Comparative experiments on benchmark with various geometric features have been performed and the simulation results show the effectiveness and the competitiveness of our proposal algorithm.

Whole-genome bisulfite sequencing (WGBS) provides a precise measure of methylation across the genome, yet presents a challenge in identifying regions that are differentially methylated (DMRs) between different conditions. A number of methods have been proposed which mainly focusing on the setting of two-group comparison. We develop a DMR detecting method MethCP for WGBS data, which is applicable for a wide range of experimental designs beyond the two-group comparisons, such as time-course data. MethCP identifies DMRs based on change point detection, which naturally segments the genome and provides region-level differential analysis. For simple two-group comparison, we show that our method outperforms developed methods in accurately detecting the complete DM region on a simulated dataset and an Arabidopsis dataset. Moreover, we show that MethCP is capable of detecting wide regions with small effect sizes, which can be common in some settings but existing techniques are poor in detecting such DMRs. We also demonstrate the use of MethCP for time-course data on another dataset following methylation throughout seed germination in Arabidopsis.

Image steganography is the operation of hiding a message into a cover image. the message can be text, codes, or image. Hiding an image into another is the proposed approach in this paper. Based on LSB coding, a k-LSB-based method is proposed using k least bits to hide the image. For decoding the hidden image, a region detection operation is used to know the blocks contains the hidden image. The resolution of stego image can be affected, for that, an image quality enhancement method is used to enhance the image resolution. To demonstrate the effectiveness of the proposed approach, we compare it with some of the state-of-the-art methods.

Text detection and recognition in natural scene images is a computer vision problem that remained a challenge for computer engineers for quite a long time. The new advancements in deep learning have revolutionized the world of computer vision. This paper attempts to build a Deep Learning (DL) based Text detection and recognition model for interpreting the text in natural scene images. The proposed model consists of three stages namely candidate text region detection, text region extraction, and text recognition. The natural scene image is first fed to the candidate text region detection mechanism which extracts potential regions containing text characters. The regions containing non-text which are introduced in the first stage of processing are filtered in the second stage. The set of text regions resulted from the second stage is then recognized in the final stage. Maximally Stable Extremal Region (MSER) algorithm is used in the candidate text region detection. Two convolutional neural networks, one in the text region extraction stage and the other one in the text recognition stage, are used in the proposed model. Text detection in natural scenes is not an easy problem as it appears. The complexity of detection and recognition of text characters in natural scene images is mainly due to the diversity of the textual characters and the natural scene, presence of various disturbances, different illumination conditions, different color, size, and area of the text. ICDAR-2011, ICDAR-2013, CHARS-74K, and CIFAR-100 datasets are used for training and validating our models.

Road perception and lane detection techniques have been widely studied in the past 2 decades and applied in many advanced driver assistance systems. However, the development of a robust lane detection system that can deal with various road conditions and efficiently evaluate its detection results in real time is still of great challenge. Therefore the construction of the evaluation system is one of the primary objectives for lane detection systems. In this section, a sensor fusion-based road region detection based on camera and Lidar and a vision-based lane detection and evaluation system based on integrating two distinct lane detection algorithms are proposed. The proposed road perception and lane detection system can benefit the driver's intention inference. Specifically, the road curvatures, the lane positions, and the lane styles can be used as input to the driver intention inference module.