Personnel and vehicle target classification system and method

Abstract

The invention relates to a personnel and vehicle target classification system and method. The method comprises the following steps: (1) conducting physical partition on video according to setting byte amount to obtain a plurality of video blocks; (2) saving the partitioned video blocks at different child nodes in sequence; (3) conducting logic slicing on a video block according to the key frame of each video block to obtain a plurality of video slices; (4) analyzing each video slice to obtain a key value pair of the video slice; (5) conducting personnel and vehicle classification on the video slices according to the key value pairs obtained from analysis to obtain personnel target information and vehicle target information; (6) saving the personnel target information and the vehicle target information after classification in different catalogs respectively. Compared with the prior art, the method is easy to operate, efficient in processing, and effective in unziping.

Description

technical field [0001] The invention relates to the technical field of video surveillance, in particular to a system and method for classifying people and vehicles. Background technique [0002] The amount of video surveillance data is huge, and with the strengthening of high-definition and ultra-high-definition trends, the scale of video surveillance data will increase exponentially; however, due to the lack of methods for fast and effective analysis of massive video data, the information utilization rate is extremely high. Low. [0003] The development of cloud computing technology provides conditions for efficient massive video analysis. But now the cloud platform is mainly oriented to processing massive text data, and has good results in log analysis, web page processing and other fields. There are the following problems when performing parallel human-vehicle target classification on video data on this cloud platform: [0004] First of all, video data is unstructured ...

AI Technical Summary

Problems solved by technology

However, there is correlation between video frames. When classifying human and vehicle objects in the video, it is necessary to detect and track the objects in the video. Therefore, it cannot be randomly divided according to the byte size, otherwise the key frame cannot be found, The lack of header files and other issues lead to decoding failures, which eventually lead to the inability to perform intelligent video analysis

[0006] Finally, although the video data can be cut according to the specified size with a video cutter first, and then uploaded to the distributed file system for intelligent analysis, this method not only increases the workload, but also increases the workload when faced with massive video data. is difficult to operate

Images