Video-based construction vehicles detection and its application in intelligent monitoring system

While vehicle detection on highways has been reported before, to the best of our knowledge, intelligent monitoring system that aims at detecting hydraulic excavators and dump trucks on state-owned land has not been explored thoroughly yet. In this paper, we present an automatic, video-based algorithm for detecting hydraulic excavators and dump trucks. Derived from lessons learned from video processing, we proposed methods for foreground detection based on an improved frame difference algorithm, and then detected hydraulic excavators and dump trucks in the respective region of interest. From our analysis, we proposed methods based on inverse valley feature of mechanical arm and spatial-temporal reasoning for hydraulic excavator detection. In addition, we explored dump truck detection strategies that combine structured component projection with the spatial relationship. Experiments on real-monitoring sites demonstrated the promising performance of our system.

A Real-Time Small Target Vehicle Detection Algorithm with an Improved YOLOv5m Network Model

To address the challenges of high complexity,poor real-time performance,and low detection rates for small target vehicles in existing vehicle object detection algorithms,this paper proposes a real-time lightweight architecture based on You Only Look Once(YOLO)v5m.Firstly,a lightweight upsampling operator called Content-Aware Reassembly of Features(CARAFE)is introduced in the feature fusion layer of the network to maximize the extraction of deep-level features for small target vehicles,reducing the missed detection rate and false detection rate.Secondly,a new prediction layer for tiny targets is added,and the feature fusion network is redesigned to enhance the detection capability for small targets.Finally,this paper applies L1 regularization to train the improved network,followed by pruning and fine-tuning operations to remove redundant channels,reducing computational and parameter complexity and enhancing the detection efficiency of the network.Training is conducted on the VisDrone2019-DET dataset.The experimental results show that the proposed algorithmreduces parameters and computation by 63.8% and 65.8%,respectively.The average detection accuracy improves by 5.15%,and the detection speed reaches 47 images per second,satisfying real-time requirements.Compared with existing approaches,including YOLOv5m and classical vehicle detection algorithms,our method achieves higher accuracy and faster speed for real-time detection of small target vehicles in edge computing.

A New Vehicle Detection Framework Based on Feature-Guided in the Road Scene

Vehicle detection plays a crucial role in the field of autonomous driving technology.However,directly applying deep learning-based object detection algorithms to complex road scene images often leads to subpar performance and slow inference speeds in vehicle detection.Achieving a balance between accuracy and detection speed is crucial for real-time object detection in real-world road scenes.This paper proposes a high-precision and fast vehicle detector called the feature-guided bidirectional pyramid network(FBPN).Firstly,to tackle challenges like vehicle occlusion and significant background interference,the efficient feature filtering module(EFFM)is introduced into the deep network,which amplifies the disparities between the features of the vehicle and the background.Secondly,the proposed global attention localization module(GALM)in the model neck effectively perceives the detailed position information of the target,improving both the accuracy and inference speed of themodel.Finally,the detection accuracy of small-scale vehicles is further enhanced through the utilization of a four-layer feature pyramid structure.Experimental results show that FBPN achieves an average precision of 60.8% and 97.8% on the BDD100K and KITTI datasets,respectively,with inference speeds reaching 344.83 frames/s and 357.14 frames/s.FBPN demonstrates its effectiveness and superiority by striking a balance between detection accuracy and inference speed,outperforming several state-of-the-art methods.

Improved YOLOv8s-Based Night Vehicle Detection

With the gradual development of automatic driving technology,people’s attention is no longer limited to daily automatic driving target detection.In response to the problem that it is difficult to achieve fast and accurate detection of visual targets in complex scenes of automatic driving at night,a detection algorithm based on improved YOLOv8s was proposed.Firsly,By adding Triplet Attention module into the lower sampling layer of the original model,the model can effectively retain and enhance feature information related to target detection on the lower-resolution feature map.This enhancement improved the robustness of the target detection network and reduced instances of missed detections.Secondly,the Soft-NMS algorithm was introduced to address the challenges of dealing with dense targets,overlapping objects,and complex scenes.This algorithm effectively reduced false and missed positives,thereby improved overall detection performance when faced with highly overlapping detection results.Finally,the experimental results on the MPDIoU loss function dataset showed that compared with the original model,the improved method,in which mAP and accuracy are increased by 2.9%and 2.8%respectively,can achieve better detection accuracy and speed in night vehicle detection.It can effectively improve the problem of target detection in night scenes.

Design of a road vehicle detection system based on monocular vision

In order to decrease vehicle crashes, a new rear view vehicle detection system based on monocular vision is designed. First, a small and flexible hardware platform based on a DM642 digtal signal processor （DSP） micro-controller is built. Then, a two-step vehicle detection algorithm is proposed. In the first step, a fast vehicle edge and symmetry fusion algorithm is used and a low threshold is set so that all the possible vehicles have a nearly 100% detection rate （TP） and the non-vehicles have a high false detection rate （FP）, i. e., all the possible vehicles can be obtained. In the second step, a classifier using a probabilistic neural network （PNN） which is based on multiple scales and an orientation Gabor feature is trained to classify the possible vehicles and eliminate the false detected vehicles from the candidate vehicles generated in the first step. Experimental results demonstrate that the proposed system maintains a high detection rate and a low false detection rate under different road, weather and lighting conditions.

Vehicle detection method for expressway by MPEG compressed domain

A method which extracts traffic information from an MPEG-2 compressed video is proposed. According to the features of vehicle motion, the motion vector of a macro-block is used to detect moving vehicles in daytime, and a filter algorithm for removing noises of motion vectors is given. As the brightness of the headlights is higher than that of the background in night images, discrete cosine transform （DCT）coefficient of image block is used to detect headlights of vehicles at night, and an algorithm for calculating the DCT coefficients of P-frames is introduced. In order to prevent moving objects outside the expressway and video shot changes from disturbing the detection, a driveway location method and a video-shot-change detection algorithm are suggested. The detection rate is 97.4% in daytime and 95.4% in nighttime by this method. The results prove that this vehicle detection method is effective.

Vehicle detection based on information fusion of vehicle symmetrical contour and license plate position

An efficient vehicle detection approach is proposed for traffic surveillance images, which is based on information fusion of vehicle symmetrical contour and license plate position. The vertical symmetry axis of the vehicle contour in an image is. first detected, and then the vertical and the horizontal symmetry axes of the license plate are detected using the symmetry axis of the vehicle contour as a reference. The vehicle location in an image is determined using license plate symmetry axes and the vertical and the horizontal projection maps of the vehicle edge image. A dataset consisting of 450 images （15 classes of vehicles） is used to test the proposed method. The experimental results indicate that compared with the vehicle contour-based, the license plate location-based, the vehicle texture-based and the Gabor feature-based methods, the proposed method is the best with a detection accuracy of 90.7% and an elapsed time of 125 ms.

Patch-based vehicle logo detection with patch intensity and weight matrix

A patch-based method for detecting vehicle logos using prior knowledge is proposed.By representing the coarse region of the logo with the weight matrix of patch intensity and position,the proposed method is robust to bad and complex environmental conditions.The bounding-box of the logo is extracted by a thershloding approach.Experimental results show that 93.58% location accuracy is achieved with 1100 images under various environmental conditions,indicating that the proposed method is effective and suitable for the location of vehicle logo in practical applications.

Vehicle detection algorithm based on codebook and local binary patterns algorithms

Detecting the moving vehicles in jittering traffic scenes is a very difficult problem because of the complex environment.Only by the color features of the pixel or only by the texture features of image cannot establish a suitable background model for the moving vehicles. In order to solve this problem, the Gaussian pyramid layered algorithm is proposed, combining with the advantages of the Codebook algorithm and the Local binary patterns(LBP) algorithm. Firstly, the image pyramid is established to eliminate the noises generated by the camera shake. Then, codebook model and LBP model are constructed on the low-resolution level and the high-resolution level of Gaussian pyramid, respectively. At last, the final test results are obtained through a set of operations according to the spatial relations of pixels. The experimental results show that this algorithm can not only eliminate the noises effectively, but also save the calculating time with high detection sensitivity and high detection accuracy.

Vehicle Detection Based on Visual Saliency and Deep Sparse Convolution Hierarchical Model

Traditional vehicle detection algorithms use traverse search based vehicle candidate generation and hand crafted based classifier training for vehicle candidate verification.These types of methods generally have high processing times and low vehicle detection performance.To address this issue,a visual saliency and deep sparse convolution hierarchical model based vehicle detection algorithm is proposed.A visual saliency calculation is firstly used to generate a small vehicle candidate area.The vehicle candidate sub images are then loaded into a sparse deep convolution hierarchical model with an SVM-based classifier to perform the final detection.The experimental results demonstrate that the proposed method is with 94.81% correct rate and 0.78% false detection rate on the existing datasets and the real road pictures captured by our group,which outperforms the existing state-of-the-art algorithms.More importantly,high discriminative multi-scale features are generated by deep sparse convolution network which has broad application prospects in target recognition in the field of intelligent vehicle.

Pedestrian and Vehicle Detection Based on Pruning YOLOv4 with Cloud-Edge Collaboration

Nowadays,the rapid development of edge computing has driven an increasing number of deep learning applications deployed at the edge of the network,such as pedestrian and vehicle detection,to provide efficient intelligent services to mobile users.However,as the accuracy requirements continue to increase,the components of deep learning models for pedestrian and vehicle detection,such as YOLOv4,become more sophisticated and the computing resources required for model training are increasing dramatically,which in turn leads to significant challenges in achieving effective deployment on resource-constrained edge devices while ensuring the high accuracy performance.For addressing this challenge,a cloud-edge collaboration-based pedestrian and vehicle detection framework is proposed in this paper,which enables sufficient training of models by utilizing the abundant computing resources in the cloud,and then deploying the well-trained models on edge devices,thus reducing the computing resource requirements for model training on edge devices.Furthermore,to reduce the size of the model deployed on edge devices,an automatic pruning method combines the convolution layer and BN layer is proposed to compress the pedestrian and vehicle detection model size.Experimental results show that the framework proposed in this paper is able to deploy the pruned model on a real edge device,Jetson TX2,with 6.72 times higher FPS.Meanwhile,the channel pruning reduces the volume and the number of parameters to 96.77%for the model,and the computing amount is reduced to 81.37%.

Image and Feature Space Based Domain Adaptation for Vehicle Detection

The application of deep learning in the field of object detection has experienced much progress.However,due to the domain shift problem,applying an off-the-shelf detector to another domain leads to a significant performance drop.A large number of ground truth labels are required when using another domain to train models,demanding a large amount of human and financial resources.In order to avoid excessive resource requirements and performance drop caused by domain shift,this paper proposes a new domain adaptive approach to cross-domain vehicle detection.Our approach improves the cross-domain vehicle detection model from image space and feature space.We employ objectives of the generative adversarial network and cycle consistency loss for image style transfer in image space.For feature space,we align feature distributions between the source domain and the target domain to improve the detection accuracy.Experiments are carried out using the method with two different datasets,proving that this technique effectively improves the accuracy of vehicle detection in the target domain.

Vehicle Detection in Still Images by Using Boosted Local Feature Detector

Vehicle detectition in still images is a comparatively difficult task. This paper presents a method for this task by using boosted local pattern detector constructed from two local features including Haar-like and oriented gradient features. The whole process is composed of three stages. In the first stage, local appearance features of vehicles and non-vehicle objects are extracted. Haar-tike and oriented gradient features are extracted separately in this stage as local features. In the second stage, Adabeost algorithm is used to select the most discriminative features as weak detectors from the two local feature sets, and a strong local pattern detector is built by the weighted combination of these selected weak detectors. Finally, vehicle detection can be performed in still images by using the boosted strong local feature detector. Experiment results show that the local pattern detector constructed in this way combines the advantages of Haar-like and oriented gradient features, and can achieve better detection results than the detector by using single Haar-like features.

Real-Time Front Vehicle Detection Algorithm Based on Local Feature Tracking Method

This paper proposes an algorithm that extracts features of back side of the vehicle and detects the front vehicle in real-time by local feature tracking of vehicle in the continuous images.The features in back side of the vehicle are vertical and horizontal edges,shadow and symmetry.By comparing local features using the fixed window size,the features in the continuous images are tracked.A robust and fast Haarlike mask is used for detecting vertical and horizontal edges,and shadow is extracted by histogram equalization,and the sliding window method is used to compare both side templates of the detected candidates for extracting symmetry.The features for tracking are vertical edges,and histogram is used to compare location of the peak and magnitude of the edges.The method using local feature tracking in the continuous images is more robust for detecting vehicle than the method using single image,and the proposed algorithm is evaluated by continuous images obtained on the expressway and downtown.And it can be performed on real-time through applying it to the embedded system.

Night Vehicle Detection Using Variable Haar-Like Feature

This paper proposes a night-time vehicle detection method using variable Haar-like feature.The specific features of front vehicle cannot be obtained in road image at night-time because of light reflection and ambient light,and it is also difficult to define optimal brightness and color of rear lamp according to road conditions.In comparison,the difference of vehicle region and road surface is more robust for road illumination environment.Thus,we select the candidates of vehicles by analysing the difference,and verify the candidates using those brightness and complexity to detect vehicle correctly.The feature of brightness difference is detected using variable horizontal Haar-like mask according to vehicle size in the location of image.And the region occurring rapid change is selected as the candidate.The proposed method is evaluated by testing on the various real road conditions.

Design of motion control system of pipeline detection AUV

A great number of pipelines in China are in unsatisfactory condition and faced with problems of corrosion and cracking,but there are very few approaches for underwater pipeline detection.Pipeline detection autonomous underwater vehicle(PDAUV) is hereby designed to solve these problems when working with advanced optical,acoustical and electrical sensors for underwater pipeline detection.PDAUV is a test bed that not only examines the logical rationality of the program,effectiveness of the hardware architecture,accuracy of the software interface protocol as well as the reliability and stability of the control system but also verifies the effectiveness of the control system in tank experiments and sea trials.The motion control system of PDAUV,including both the hardware and software architectures,is introduced in this work.The software module and information flow of the motion control system of PDAUV and a novel neural network-based control(NNC) are also covered.Besides,a real-time identification method based on neural network is used to realize system identification.The tank experiments and sea trials are carried out to verify the feasibility and capability of PDAUV control system to complete underwater pipeline detection task.

Road boundary estimation to improve vehicle detection and tracking in UAV video

Video processing is one challenge in collecting vehicle trajectories from unmanned aerial vehicle(UAV) and road boundary estimation is one way to improve the video processing algorithms. However, current methods do not work well for low volume road, which is not well-marked and with noises such as vehicle tracks. A fusion-based method termed Dempster-Shafer-based road detection(DSRD) is proposed to address this issue. This method detects road boundary by combining multiple information sources using Dempster-Shafer theory(DST). In order to test the performance of the proposed method, two field experiments were conducted, one of which was on a highway partially covered by snow and another was on a dense traffic highway. The results show that DSRD is robust and accurate, whose detection rates are 100% and 99.8% compared with manual detection results. Then, DSRD is adopted to improve UAV video processing algorithm, and the vehicle detection and tracking rate are improved by 2.7% and 5.5%,respectively. Also, the computation time has decreased by 5% and 8.3% for two experiments, respectively.

3D Vehicle Detection Algorithm Based onMultimodal Decision-Level Fusion

3D vehicle detection based on LiDAR-camera fusion is becoming an emerging research topic in autonomous driving.The algorithm based on the Camera-LiDAR object candidate fusion method(CLOCs)is currently considered to be a more effective decision-level fusion algorithm,but it does not fully utilize the extracted features of 3D and 2D.Therefore,we proposed a 3D vehicle detection algorithm based onmultimodal decision-level fusion.First,project the anchor point of the 3D detection bounding box into the 2D image,calculate the distance between 2D and 3D anchor points,and use this distance as a new fusion feature to enhance the feature redundancy of the network.Subsequently,add an attention module:squeeze-and-excitation networks,weight each feature channel to enhance the important features of the network,and suppress useless features.The experimental results show that the mean average precision of the algorithm in the KITTI dataset is 82.96%,which outperforms previous state-ofthe-art multimodal fusion-based methods,and the average accuracy in the Easy,Moderate and Hard evaluation indicators reaches 88.96%,82.60%,and 77.31%,respectively,which are higher compared to the original CLOCs model by 1.02%,2.29%,and 0.41%,respectively.Compared with the original CLOCs algorithm,our algorithm has higher accuracy and better performance in 3D vehicle detection.

Optimal Deep Convolutional Neural Network for Vehicle Detection in Remote Sensing Images

Object detection(OD)in remote sensing images(RSI)acts as a vital part in numerous civilian and military application areas,like urban planning,geographic information system(GIS),and search and rescue functions.Vehicle recognition from RSIs remained a challenging process because of the difficulty of background data and the redundancy of recognition regions.The latest advancements in deep learning(DL)approaches permit the design of effectual OD approaches.This study develops an Artificial Ecosystem Optimizer with Deep Convolutional Neural Network for Vehicle Detection(AEODCNN-VD)model on Remote Sensing Images.The proposed AEODCNN-VD model focuses on the identification of vehicles accurately and rapidly.To detect vehicles,the presented AEODCNN-VD model employs single shot detector(SSD)with Inception network as a baseline model.In addition,Multiway Feature Pyramid Network(MFPN)is used for handling objects of varying sizes in RSIs.The features from the Inception model are passed into theMFPNformultiway andmultiscale feature fusion.Finally,the fused features are passed into bounding box and class prediction networks.For enhancing the detection efficiency of the AEODCNN-VD approach,AEO based hyperparameter optimizer is used,which is stimulated by the energy transfer strategies such as production,consumption,and decomposition in an ecosystem.The performance validation of the presentedmethod on benchmark datasets showed promising performance over recent DL models.

Vision-Based On-Road Nighttime Vehicle Detection and Tracking Using Taillight and Headlight Features

An important and challenging aspect of developing an intelligent transportation system is the identification of nighttime vehicles. Most accidents occur at night owing to the absence of night lighting conditions. Vehicle detection has become a vital subject for research to ensure safety and avoid accidents. New vision-based on-road nighttime vehicle detection and tracking system are suggested in this survey paper using taillight and headlight features. Using computer vision and some image processing techniques, the proposed system can identify vehicles based on taillight and headlight features. For vehicle tracking, a centroid tracking algorithm has been used. Euclidean Distance method has been used for measuring the distances between two neighboring objects and tracks the nearest neighbor. In the proposed system two flexible fixed Region of Interest (ROI) have been used, one is the Headlight ROI, and another is the Taillight ROI that could adapt to different resolutions of the images and videos. The achievement of this research work is that the proposed two ROIs can work simultaneously in a frame to identify oncoming and preceding vehicles at night. The segmentation techniques and double thresholding method have been used to extract the red and white components from the scene to identify the vehicle headlights and taillights. To evaluate the capability of the proposed process, two types of datasets have been used. Experimental findings indicate that the performance of the proposed technique is reliable and effective in distinct nighttime environments for detection and tracking of vehicles. The proposed method has been able to detect and track double lights as well as single light such as motorcycle light and achieved average accuracy and average processing time of vehicle detection about 97.22% and 0.01 s per frame respectively.