| With the rapid development of society,crowded places can be seen everywhere.Statistical analysis of personnel in specific scenarios and implementation of human-computer interactive statistical analysis algorithms can provide effective and feasible technical means for urban resource optimization configuration,modern security,and commercial information collection.Based on the video surveillance system,this paper uses video analysis technology to carry out research on the statistics of specific scene personnel and its feature refinement.This not only has important theoretical significance,but also has clear practical application value.Focusing on the statistical analysis of the video-based scene personnel and the refinement of its personnel characteristics,this paper mainly carried out the following work:Firstly,this paper expounds the shortcomings of current video-based personnel statistical methods,analyzes the technical difficulties of current personnel heat statistics,and proposes an efficient and fast scene personnel statistical method based on target detection.Secondly,based on the personnel detection,this paper studies the characteristics of the scene personnel,including three characteristics of body type,gender and age.In terms of body shape recognition,the Mask RCNN is first used for human body segmentation,and then the key points of the human body are extracted based on OpenPose.Then,according to different observation angles,a body shape recognition method based on geometric measurement is proposed,and the fuzzy mathematics method is used for judgment.In terms of gender identification,this paper uses the current advanced deep residual network model for feature extraction and classification on an indoor scene dataset,which breaks through the limitation of common network on network depth and achieves good results.In terms of age identification,this paper studies it as a multi-classification problem(that is,divided into different age groups).Based on the pre-processing of samples,a densely connected deep network model is used to identify the age.The purpose of accurate age identification;Finally,the paper tests the algorithms of the above functional modules.The experimental results show that the proposed method has a good recognition effect. |
