Research On Remaining Object Detection Algorithm Based On Improved YOLOv2 Network

With the social importance of public safety issues,intelligent video surveillance technology plays an increasingly important role in security,as an effective security technology to be applied and promoted.Remnant detection technology has become a key part of intelligent video surveillance systems,helping to recover legacy objects in time or eliminate the potential hazards of unidentified remnants in video surveillance areas.The existing detection algorithm for legacy detection has a higher false detection rate and missed detection rate in the foreground detection under crowded monitoring conditions and can not solve the problems such as background movement and interference caused by occultation of the legacy object.In order to improve the accuracy and real-time of detection of remnants in complex environment,this paper deeply studies and analyzes target detection and remnant detection based on depth learning,and applies YOLOv2 target detection to the detection of remnant.Remaining Object Detection Algorithm Based on Improved YOLOv2 Network.The main research work of this paper is as follows:(1)In view of YOLOv2 small target detection accuracy is not high,this paper presents YOLOv2-A detection network.By introducing deep residual network thought,deep features and shallow features are merged many times,and the performance of small targets detection is improved without increasing the original network computing time.At the same time,YOLOv2-A is used to replace the traditional target detection method to extract the foreground object,which can effectively eliminate the interference of the non-object objects such as resident pedestrians and animals.Finally,carry on the analysis and judgment of the remnant to mark the remnant.(2)Aiming at the shortcomings of ORB algorithm such as scale invariance,this paper proposes ORBS(ORB-SURF)feature matching algorithm.The ORBS algorithm uses the SURF algorithm to extract the features and makes the feature points have the scale invariance.The PROSAC algorithm is used to obtain the stable accurate matching points based on the Hamming distance rough matching.At the same time,the ORBS algorithm is applied to the detection of the relics proposed in the above to solve the problem of background movement and occlusion of the relics,and a detection algorithm based on the improved ORB is proposed.(3)This paper designs and implements an intelligent detection system for the remnants of the actual monitoring scenario.The whole framework of the system and the basic principle of the module are introduced in detail,and the function design and implementation of the system server and client are also introduced.