Specific pedestrian tracing method and system based on adaptive component feature combination query

Abstract

The invention discloses a specific pedestrian tracing method and system based on adaptive component feature combination query. The method comprises the steps that pedestrian target representative frames are extracted from an input video; structured semantic description of a pedestrian image is obtained according to the pedestrian target representative frames, and overall features of the pedestrian image are extracted; semantic component segmentation is performed on the pedestrian target representative frames to generate pedestrian components, and component features of the pedestrian components are extracted; and combined measurement is performed on the structured semantic description of the pedestrian image, the overall features and the component features as well as characteristics of a query image, a target similarity between each pedestrian and a query person is obtained, and the pedestrian corresponding to the maximum similarity is a traced person. Through the specific pedestrian tracing method and system, pixel-level segmentation is performed on the pedestrian image to extract the component features, similarity measurement between components is more pertinent than global feature measurement, a visual angle problem can be better solved, and the efficiency and accuracy of tracing of a specific pedestrian are improved.

Description

technical field [0001] The invention relates to tracking videos of specific pedestrians, in particular to a method and system for tracking specific pedestrians based on the combination of features of query adaptive components, and belongs to the field of video investigation business. Background technique [0002] With the extensive construction of safe cities and the popularity of surveillance in various places, the amount of video surveillance data has become larger and larger, which has brought great challenges to criminal investigation and detection. How to quickly and accurately extract target suspects from these massive databases become the key to solve the case. The traditional pedestrian tracking method can effectively solve the problems of missed detection and false detection that may be caused by long-term manual retrieval, but the matching efficiency is relatively low. The main problems are: (1) The pedestrian representative image is not extracted accurately, resul...

AI Technical Summary

Problems solved by technology

The traditional pedestrian tracking method can effectively solve the problems of missed detection and false detection that may be caused by long-term manual retrieval, but the matching efficiency is relatively low. The main problems are: (1) The pedestrian representative image is not extracted accurately, resulting in inaccurate detection. Locating the pedestrian image, the image is associated with many people or incomplete, which makes the subsequent feature extraction invalid; (2) only extracts the global features of the pedestrian, and the pedestrian expression is not fine, which makes the pedestrian expression biased or the discrimination is insufficient; (3) All pedestrians use a unified The matching method does not consider individual differences, resulting in limited use scenarios and individuals

Images