| The detection of medium resolution and complex environment is not affected by complex environment.Complex detection environment has brought great interference to target detection.Researchers at home and abroad have proposed a variety of methods to deal with the interference from complex environment.However,due to different application scenarios,the methods to solve the problem of complex environment interference are also different.Therefore,in order to make the target detection model more practical,this paper uses spatial attention mechanism to improve SSD target detection algorithm,In order to solve the problem of target detection in complex environment,a variety of target detection algorithms in complex environment are studied.This paper uses target detection technology to detect people in complex environment.In the process of personnel detection,there are some problems,such as blurred image,low detection efficiency and high real-time requirements.In this paper,the improved dark channel priority processing strategy is used for image enhancement,and the SSD(single shot multibox detector)network combined with spatial attention mechanism is used to improve the detection rate,at the same time,the real-time requirements are guaranteed,and verified on the test set and VOC data set.Using spatial attention mechanism to improve SSD target detection network,SSDSN(SSD spatial attention mechanism network)is proposed,after the original image input layer of SSD network,cov2＿1 is combined with spatial attention mechanism module＿After that,the spatial attention mechanism is introduced to improve the ability of low-level feature extraction and detail extraction of SSD target detection network,and improve the accuracy and recall rate of target detection.The detection ability of SSDSN algorithm is verified in the algorithm comparison.In view of the interference of complex environment on target detection,the complex problem of detection area is solved by region division,and the influence of image resolution and illumination on target detection is solved by improving the dark channel priority image processing strategy.Combined with ssdsn,a personnel detection method in complex environment is proposed,by comparing the detection results of Fast R-CNN,SSD,YOLOV2,YOLOV3 and SSDSN,the detection ability of SSDSN in complex environment is proved.In this paper,we use open source Django framework and canvas plug-in to develop a personnel detection system in complex environment.The system provides a simple operation interface.Through simple and convenient operation,we can quickly deploy the target detection environment in complex environment.Through detection,we can find the problems in the personnel detection method in complex environment,which is conducive to the target detection method in complex environment Optimization. |
