Research On Pedestrian Vehicle Target Detection Algorithm Based On Deep Learning

With the rapid development of urbanization and the gathering of population,the road congestion becomes more and more serious,which also brings the problem of frequent traffic accidents.Under this background,intelligent transportation system has received more and more attention.Aiming at the problem that the traditional target detection algorithm cannot meet the real-time and accuracy in complex environment,the pedestrian vehicle target detection algorithm based on deep learning can effectively reduce the pressure of urban road management and improve the frequent accidents caused by traffic congestion,which has high research and application value.Compared with other detection methods,the target detection method based on YOLO(You only look once)has a relatively fast target detection speed due to its fewer parameters,which is more suitable for real-time detection in traffic scenes.Therefore,this paper proposes an intelligent traffic target detection system based on YOLO.The main work of this paper is as follows:(1)Aiming at the problem that YOLOV4 is difficult to meet the high precision requirements of pedestrian vehicle detection,this paper proposes an AM-YOLO pedestrian vehicle detection algorithm based on YOLOV4.Firstly,a Convolutional Block Attention Module(CBAM)is added to the YOLOV4 network,which enables the network to focus more on important features and ignore some irrelevant information.The addition of CBAM module makes the network focus on important features in both channel and space dimensions.Secondly,the path aggregation network is improved.Through a series of up-sampling and multiple splicing operations between the three feature maps,the feature maps are efficiently fused between the upper level and the bottom level,and the detection effect is significantly improved.Finally,the K-means++ algorithm is used to re-cluster the prior boxes suitable for the detection of the current pedestrian vehicle data set,helping the model to more accurately locate and identify the pedestrian vehicle target.The experimental results show that the detection accuracy of AM-YOLO algorithm is improved by 1.96%compared with YOLOV4 algorithm,and the detection ability of AM-YOLO algorithm is significantly enhanced.(2)Aiming at the problem that YOLOV5s is difficult to meet the high real-time requirements of pedestrian vehicle detection,this paper proposes an improved pedestrian vehicle detection algorithm based on YOLOV5s.Firstly,Ghost module is used to replace some modules of YOLOV5s backbone network.GhostNet is a lightweight network,which uses linear operation instead of partial convolution.The number of parameters in the network is greatly reduced,which meets the real-time requirements of pedestrians and vehicles.Secondly,the integration of Efficient Channel Attention(ECA)with Ghost modules and the original modules of the network can not only greatly reduce the number of parameters,but also improve the learning ability of important features of the network and maintain the feature processing ability of the network.Finally,based on the Complete Intersection Over Union(CIOU)loss function of the original network,the ratio of width to height to distance loss of the boundary frame is changed to consider the ratio of width to height of the predicted frame and the real frame respectively,so that the loss function can converge effectively in the training process of the network.The experimental results show that the improved algorithm reduces the number of YOLOV5s parameters by 28%and model size by 27%while maintaining high accuracy,which saves hardware cost and can meet the requirements of practical applications.(3)Design of pedestrian vehicle target detection system with PyQt5.Combined with the pedestrian vehicle target detection algorithm proposed in this paper,a pedestrian vehicle image and video detection system is designed.First of all,data enhancement is carried out on the pictures in the traffic scene.Then,object detection can be performed on pedestrian and vehicle pictures or videos.The experimental results show that the system can accurately detect the pictures or videos of pedestrians and vehicles,and display the detection results in real time on the system interface.