Research And Implementation Of Video Surveillance Pedestrian Detection Technology

With the rapid development of electronic technology,the research of computer system to meet the needs of vision has long been a hot research direction.The most widely used is video surveillance technology,and the rapid development of video surveillance technology is derived from its powerful application.It can be seen in all fields and closely related to our life.Video surveillance technology has made outstanding contributions in maintaining social order and reducing labor costs.However,with the continuous development of this technology,people are increasingly eager to emancipate their eyes from huge video data,and improve the automation and intelligence of video surveillance.Therefore,pedestrian detection technology emerges as the times require.Pedestrian detection technology is using computer vision technology to identify pedestrian targets in videos and images.The pedestrian detection project has been listed as the key research project in many countries.The project has been given so much attention because of its advanced technology and extensive involvement: in the field of defense and military,the technology can be used in battlefield detection,target tracking and precision guidance.In the field of urban transportation,the technology can be used for intelligent transportation and violation inspection.In the field of social security,the technology can be used for human flow monitoring.Such a great prospect of development has attracted many scholars and made great progress in recent years,so it is of great significance to further study the technology.However,the technology is affected by many factors in the actual application process.First,the detection environment is complex and changeable,and the weather and light interference will increase the difficulty of detection.Secondly,the detection target is in motion,and the body,posture and clothes of pedestrians will also interfere with the detection effect.However,the biggest problem is to detect the target.Occlusion problem,the occlusion between pedestrians and the occlusion of objects to pedestrians increase the difficulty of detection.Therefore,how to improve the system performance and achieve high quality test results is a problem that needs to be solved in this field.In order to avoid the above problems,improve the system effectiveness and ensure the accuracy and timeliness of the system,the research contents can be summarized as follows:(1)In-depth investigation and Research on the development status of pedestrian detection technology at home and abroad,the significance of this study,the actual problems to be determined and the problems that may be faced in the study process in advance,as well as the heavy and difficult points in this study,and the overall planning of this study.(2)This paper outlines the widely used feature extraction algorithms and classification algorithms in the field of line detection,such as Haar,HOG,SIFT,Edgelet,SVM and AbsBoost,and summarizes the advantages and defects of the above extraction algorithms and classification algorithms,and finds out the reasons for the quality of the system detection,which lays the foundation for the next step of research.(3)The method of pedestrian detection is optimized,which is different from the original method of single feature extraction.This paper studies the method of merging a variety of features and extracting the feature of the target area.Through two different methods to detect the target area,the advanced line is roughly classified,the suspicious target is locked,and then the other classifier is selected carefully.Finally,the final result is obtained by using Bayesian decision making,and the validity and feasibility of the new method are detected by comparison experiment.(4)A new classification algorithm is proposed,which uses the cost sensitive support vector machine to solve the problem of the unbalance of positive and negative samples in the classification,and reduces the sample non equilibrium interference to the minimum by using its strong sensitivity.At the same time,the Newton step method is used to train the weak classifier,and the classifier constructed by this algorithm can avoid the over fitting problem,and ensure the difference and enhance the performance of the system classification.