Search

Deep learning is proliferating within the field of computer vision. Road damage detection technology, as its offshoot, already plays vital role in road maintenance and traffic safety. With road damage such as potholes and cracks, accurate and efficient detection results are essential for timely road safety repair and maintenance. Therefore, road damage detection algorithms based on deep learning have attracted wide attention. YOLOv5 is an advanced target detection algorithm known for its efficient detection speed and good accuracy. However, there is still room for further improvement in its performance for road damage detection. The ability of multi-scale damage detection and spatial structure capture is not perfect. Therefore, this paper proposes three improvement points to improve the accuracy of road damage detection based on YOLOv5. The first introduced module is the Non-linear Spatial Pyramidal Pooling-Fast (NSPPF) module. This module allows for better capture of detailed features of road damage areas. Non-linear transformation and fast pyramid operation improve the sensing ability and multi-scale damage detection ability. Secondly, a combination of the CoordConv and SK attention modules is constructed. The CoordConv module fuses coordinate information with features to provide a more spatially informed representation. The SK attention module also learns correlations between global and local features, enhancing the model's ability to detect damages at different scales. This paper can better capture road injuries' spatial structure and context information by combining these two modules. Finally, experimental results on the RDD2020 dataset demonstrate the effectiveness of our model. Compared to the baseline model, the proposed improvement algorithm increases the accuracy by 2.2%, resulting in a mAP of 58.2%. This demonstrates its effectiveness and feasibility. [ABSTRACT FROM AUTHOR]

Object detection is one of the core technologies for autonomous driving. Current road object detection mainly relies on visible light, which is prone to missed detections and false alarms in rainy, night-time, and foggy scenes. Multispectral object detection based on the fusion of RGB and infrared images can effectively address the challenges of complex and changing road scenes, improving the detection performance of current algorithms in complex scenarios. However, previous multispectral detection algorithms suffer from issues such as poor fusion of dual-mode information, poor detection performance for multi-scale objects, and inadequate utilization of semantic information. To address these challenges and enhance the detection performance in complex road scenes, this paper proposes a novel multispectral object detection algorithm called MRD-YOLO. In MRD-YOLO, we utilize interaction-based feature extraction to effectively fuse information and introduce the BIC-Fusion module with attention guidance to fuse different modal information. We also incorporate the SAConv module to improve the model's detection performance for multi-scale objects and utilize the AIFI structure to enhance the utilization of semantic information. Finally, we conduct experiments on two major public datasets, FLIR_Aligned and M3FD. The experimental results demonstrate that compared to other algorithms, the proposed algorithm achieves superior detection performance in complex road scenes. [ABSTRACT FROM AUTHOR]

In response to the low matching accuracy of stereo matching algorithms in image regions with specular reflection, this paper proposes a multidimensional fusion stereo matching algorithm named MFANet. The algorithm embeds a multispectral attention module into the residual feature extraction network, utilizing two‐dimensional discrete cosine transforms to extract frequency features. In the pyramid pooling module, a coordinated attention mechanism is introduced to capture relevant positional information. In the cost aggregation part, the MFANet algorithm incorporates a three‐dimensional attention mechanism, focusing on the more important semantic information in high‐level features. By combining detailed information from low‐level features, semantic information from high‐level features, and contextual information, the algorithm generates features that are more conducive to disparity prediction. The MFANet algorithm is evaluated on three standard datasets (SceneFlow, KITTI2015, and KITTI2012). Experimental results demonstrate its robustness against specular reflection interference, accurate prediction of disparities in specular reflection pathological regions, and promising application prospects. [ABSTRACT FROM AUTHOR]

Due to the robustness of skeleton data to human scale, illumination changes, dynamic camera views, and complex backgrounds, great progress has been made in skeleton‐based video anomaly detection in recent years. The spatio‐temporal graph convolutional network has been proven to be effective in modelling the spatio‐temporal dependencies of non‐Euclidean data such as human skeleton graphs, and the autoencoder based on this basic unit is widely used to model sequence features. However, due to the limitations of the convolution kernel, the model cannot capture the correlation between non‐adjacent joints, and it is difficult to deal with long‐term sequences, resulting in an insufficient understanding of behaviour. To address this issue, this paper applies the Transformer to the human skeleton and proposes the Spatio‐Temporal Enhanced Graph‐Transformer AutoEncoder (STEGT‐AE) to improve the capability of modelling. In addition, the multi‐memory model with skip connections is employed to provide different levels of coding features, thereby enhancing the ability of the model to distinguish similar heterogeneous behaviours. Furthermore, the STEGT‐AE has a single encoder‐double decoder architecture, which can improve the detection performance by the combining reconstruction and prediction error. The experimental results show that performances of STEGT‐AE is significantly better than other advanced algorithms on four baseline datasets. [ABSTRACT FROM AUTHOR]

Hand-drawn is one of the few visual descriptors that can directly represent visual content, and has significant research in the area of computer vision. Aiming at the problem of sparse features in the realm of hand-drawn image retrieval, hand-drawn images, and the easy deformation of hand-drawn images, this paper proposes a feature extraction method of grid resource sharing collaborative algorithm, which can be obtained utilizing precisely extracted semantic characteristics from hand-drawn images through computer multimedia-aided design Efficient and accurate retrieval results. First, the fundamental framework for obtaining semantic features is algorithm; then the attention model mechanism is the grid resource sharing collaborative introduced in the process of supervised training, and the attention structure block is introduced after the convolutional neural network's bottom layer. To locate effective semantic features, In order to accomplish high-precision retrieval, the attention structure block combines channel attention structure and spatial attention structure to build the attention structure block. The last feature descriptor is then created by combining various semantic feature levels. The proposed strategy is practical and efficient, as demonstrated by the experimental findings on the comparison database Flickr15k. In addition, in the task of hand-drawn image classification, the proposed attention mechanism greatly improves the classification accuracy. [ABSTRACT FROM AUTHOR]

During the replanting operation of a seedling tray, the end-effector needs to repeatedly grab the qualified plug seedlings in the supply tray and release them to the target tray for replanting, and in the process of grasping, the end-effector may cause some mechanical damage to the plug seedlings, thus affecting their quality. Therefore, in order to be able to adjust the position of the hand claw grasping point according to the morphological characteristics of the plug seedlings and select the optimal grasping point, this paper proposes research on the optimal grasping angle algorithm for plug seedlings based on machine vision. Firstly, a rotatable three-jaw end-effector is designed, which uses a three-jaw structure for grasping the burrowing seedlings. The three claws are driven with a telescopic cylinder to carry out clamping and relaxing actions. The rotation of the three claws is controlled with the stepper motor to adjust the optimal grasping position. Secondly, based on the pre-processing of an image of the hole tray seedling, the extraction of feature points in the region of interest, and the calculation of localization, the angle between the angular bisector of the cotyledon leaf blade of the hole tray seedling and the horizontal positive direction is solved. In this paper, two methods are designed to calculate the coordinates of feature points: one is the geometric method and the other is the center-of-mass method. Finally, the optimal grasping angle is calculated by analyzing the angle between the angular bisector of the cotyledon leaf blade and the horizontal positive direction of the cavity seedlings. According to the test, the average calculation error of the proposed algorithm is 3.12 degrees, and the average calculation time is 0.512 sec/sheet, which meet the requirements of the replanting operation. [ABSTRACT FROM AUTHOR]

Intelligent agriculture has become the development trend of agriculture in the future, and it has a wide range of research and application scenarios. Using machine learning to complete basic tasks for people has become a reality, and this ability is also used in machine vision. In order to save the time in the fruit picking process and reduce the cost of labor, the robot is used to achieve the automatic picking in the orchard environment. Cherry target detection algorithms based on deep learning are proposed to identify and pick cherries. However, most of the existing methods are aimed at relatively sparse fruits and cannot solve the detection problem of small and dense fruits. In this paper, we propose a cherry detection model based on YOLOv5s. First, the shallow feature information is enhanced by convolving the feature maps sampled by two times down in BackBone layer of the original network model to the input end of the second and third CSP modules. In addition, the depth of CSP module is adjusted and RFB module is added in feature extraction stage to enhance feature extraction capability. Finally, Soft- Non-Maximum Suppression (Soft-NMS) is used to minimize the target loss caused by occlusion. We test the performance of the model, and the results show that the improved YOLOv5s-cherry model has the best detection performance for small and dense cherry detection, which is conducive to intelligent picking. [ABSTRACT FROM AUTHOR]

The counting of surgical instruments is an important task to ensure surgical safety and patient health. However, due to the uncertainty of manual operations, there is a risk of missing or miscounting instruments. Applying computer vision technology to the instrument counting process can not only improve efficiency, but also reduce medical disputes and promote the development of medical informatization. However, during the counting process, surgical instruments may be densely arranged or obstruct each other, and they may be affected by different lighting environments, all of which can affect the accuracy of instrument recognition. In addition, similar instruments may have only minor differences in appearance and shape, which increases the difficulty of identification. To address these issues, this paper improves the YOLOv7x object detection algorithm and applies it to the surgical instrument detection task. First, the RepLK Block module is introduced into the YOLOv7x backbone network, which can increase the effective receptive field and guide the network to learn more shape features. Second, the ODConv structure is introduced into the neck module of the network, which can significantly enhance the feature extraction ability of the basic convolution operation of the CNN and capture more rich contextual information. At the same time, we created the OSI26 data set, which contains 452 images and 26 surgical instruments, for model training and evaluation. The experimental results show that our improved algorithm exhibits higher accuracy and robustness in surgical instrument detection tasks, with F1, AP, AP50, and AP75 reaching 94.7%, 91.5%, 99.1%, and 98.2%, respectively, which are 4.6%, 3.1%, 3.6%, and 3.9% higher than the baseline. Compared to other mainstream object detection algorithms, our method has significant advantages. These results demonstrate that our method can more accurately identify surgical instruments, thereby improving surgical safety and patient health. [ABSTRACT FROM AUTHOR]

Due to low line extraction and matching accuracy, existing point-line fusion visual SLAM（simultaneous localization and mapping） methods have low localization accuracy. This paper improved existing point-line fusion stereo visual SLAM method from multiple dimensions. Firstly, this paper improved the LSD（line segment detection） extraction quality through internal parameters adjustment and length threshold filtering. Secondly, on the basis of geometric constraints, it abstracted the line feature matching as a sparse minimization problem to find best match of line segment. Thirdly, on the basis of image texture score and number of successfully matched features, this paper improved back-end point-line feature fusion strategy by assigning point-line feature weights. Compared with LSD algorithm and LBD（line binary descriptor） matching method, the number of short line segments decreased by 67% on average and matching accuracy increased by 18% on average. Compared with PL-SLAM, localization error dropped by about 15% and 45% respectively on the EuRoC and KITTI datasets. Experiments demonstrate that the proposed method enhances localization accuracy of SLAM system effectively by improving various modules performance of point-line SLAM method in multiple dimensions. [ABSTRACT FROM AUTHOR]

For facing of the problems caused by the YOLOv4 algorithm's insensitivity to small objects and low detection precision in traffic light detection and recognition, the Improved YOLOv4 algorithm is investigated in the paper using the shallow feature enhancement mechanism and the bounding box uncertainty prediction mechanism. The shallow feature enhancement mechanism is used to extract features from the network and improve the network's ability to locate small objects and color resolution by merging two shallow features at different stages with the high-level semantic features obtained after two rounds of upsampling. Uncertainty is introduced in the bounding box prediction mechanism to improve the reliability of the prediction of the bounding box by modeling the output coordinates of the prediction bounding box and adding the Gaussian model to calculate the uncertainty of the coordinate information. The LISA traffic light data set is used to perform detection and recognition experiments separately. The Improved YOLOv4 algorithm is shown to have a high effectiveness in enhancing the detection and recognition precision of traffic lights. In the detection experiment, the area under the PR curve value of the Improved YOLOv4 algorithm is found to be 97.58%, which represents an increase of 7.09% in comparison to the 90.49% score gained in the Vision for Intelligent Vehicles and Applications Challenge Competition. In the recognition experiment, the mean average precision of the Improved YOLOv4 algorithm is 82.15%, which is 2.86% higher than that of the original YOLOv4 algorithm. The Improved YOLOv4 algorithm shows remarkable advantages as a robust and practical method for use in the real-time detection and recognition of traffic signal lights. [ABSTRACT FROM AUTHOR]

Machine Learning (ML) methodologies have become a worthy choice for X-ray screening. X-ray imaging is a method used for the identification of bone fractures. But every so often, the dimensions and locations of a fracture are detected incorrectly. Therefore, this paper aims to design a system that would accurately detect and classify fractured and non-fractured bone images. This proposed system comprises four principal phases. The first is the image acquisition phase, in which few input images are acquired from the X-ray machine, and few input images are collected from the imaging centre. The second is a pre-processing phase to see their informative areas like location, edges and shape. So, the wavelet transform method helps to compress the images to store and remove noises. The third is the feature extracting phase; the Harris corner detection technique is used to locate the broken points as corner features and intensifies the quality of the X-ray images. Before applying the Harris corner algorithm, the image sharpening method was used. Then images are equipped to be fed into a fourth phase, i.e. classification phase; two techniques are used in this phase: Support Vector Machine (SVM) and Error Backpropagation Neural Network (EBP-NN). Classification by SVM and EBP-NN both methods are tested on various fractured and non-fractured bone images. Finally, found that classification done by the SVM technique is more accurate than EBP-NN. The classification by the SVM approach has higher efficiency and a higher classification rate of 94%, which is 2% more than the EBP-NN technique. [ABSTRACT FROM AUTHOR]

Feature point extraction and matching work together and play a crucial role in many machine vision tasks, such as visual simultaneous localization and mapping (SLAM) which is the basis of robotics. However, feature matching has not been fully studied as much as feature extraction. The most widely used feature matching method is still the classical brute-force matching. In this paper, by fusing the data of inertial measurement unit (IMU), a novel feature points matching algorithm with online learning fault detection, where the global matcher and local matcher operate alternately and switch automatically, is proposed to improve matching accuracy. Taking the uncertainties of the initial states of camera and IMU into account in the initialization phase, the system starts with the global mode to accomplish the matching task and initialize the state variables which are used to align the data of two sensors. Then, once completing the initialization, the algorithm will automatically switch to our new-designed local matcher with feature compensation, which bounds the search space and improves the matching accuracy via performing on the support region according to the results of IMU pre-integration rather than the whole image. In addition, for the purpose of effectively dealing with the misestimation of support region induced by IMU failures, an online learning fault detector based on support vector machines is newly developed, which can trigger the action of mode switcher and runs simultaneously in another thread. Furthermore, a timer is attached to the fault detector in order to activate the detection periodically, which can further reduces the cost of computing resources. Finally, the performance of the present algorithm is validated on public datasets. • A dual-mode automatic switching feature points matching algorithm is designed. • The algorithm consists of a global matcher, a local matcher and a fault detector. • A local matcher with feature compensation is proposed to improve matching accuracy. • A SVM-based online learning fault detector is presented to address IMU faults. [ABSTRACT FROM AUTHOR]