Multi-target Recognition Method for Vehicle Infrared Images Based on Improved YOLOv5
The rural-urban fringe is an important part of urban construction.it is difficult to effectively deploy detection equip-ment,the night supervision of vehicle targets in this area always is a difficult problem for urban management.This paper provides an intelligent approach of detecting multiple moving targets for solving the problem in infrared night vision images based on UAV plat-forms,presents a multi-moving target recognition method based on improved YOLOv5 in infrared night vision conditions,and analyzes the characteristics of traffic objects and impact of vehicle parking on road infrared radiation,etc.Convolutional block attention module(CBAM)attention mechanism is introduced to extract and integrate spatial with channel information to enhance the expression ability of the network on the target.By combining the advantages of efficient IOU loss and focal loss,the EIoU-focal loss function is used to replace the CIoU loss function,solve the disadvantages of sample imbalance,low resolution of infrared image,large noise interference and low contrast between target and background,and improve the detection accuracy.By adding the DCN to dynamically adjust the shape of the convolution kernel,it can adapt to the deformation of the object in images,and reduce the recognition influence caused by irregular shape and many changes.Finally,experiments and data comparisons between improved network and typical network are im-plemented on public dataset,the results show that for multi targets recognition,the improved network based on YOLOv5 has higher recognition results,it increases the accuracy,recall rate andF1value by 3.9％,4.1％and 4.4％,respectively.
deep learningmulti-object recognitionYOLOv5deformable convolutionattention mechanisms
NETL
NSTL
万方数据
