Pedestrian re-identification method and device based on spatial constraints

Abstract

The present invention proposes a pedestrian re-identification method and equipment based on space constraints, including: obtaining image information in the observation scene for analysis, and calculating a projection matrix; the intelligent front end performs single-shot tracking on the corresponding pedestrian target, and generates a pedestrian target tracking trajectory , which is transformed into a three-dimensional coordinate trajectory through coordinate mapping; the background server receives the pedestrian target tracking trajectory returned from the intelligent front end, and aggregates the pedestrian target tracking trajectory to obtain the aggregated trajectory; The following pedestrian target tracking trajectory is sampled; the background server receives the pedestrian image to be retrieved, extracts the characteristics of the pedestrian image as the retrieval feature, compares the retrieval feature with the stored multiple features, and searches for the pedestrian target trajectory that is successfully compared. And sort according to the confidence, and return the search results. The present invention improves the accuracy of pedestrian search through three-dimensional space constraints.

Description

technical field [0001] The invention relates to the technical field of intelligent video analysis, in particular to a pedestrian re-identification method and equipment based on space constraints. Background technique [0002] With the rapid development of the economy and the flow of population, especially in large cities, there are more densely populated areas, and these densely populated areas are sensitive areas where security issues frequently occur, so these sensitive areas have also become key areas in urban construction. worry. In order to prevent various problems such as theft, robbery and fights in key sensitive areas, relevant departments have installed a large number of cameras in relevant locations for real-time monitoring of these areas, but due to long distances or problems with clarity, in After an emergency occurs, the suspect cannot be accurately located and tracked, and the expected effect cannot be achieved. [0003] The traditional pedestrian re-id metho...

AI Technical Summary

Problems solved by technology

In order to prevent various problems such as theft, robbery and fights in key sensitive areas, relevant departments have installed a large number of cameras in relevant locations for real-time monitoring of these areas, but due to long distances or problems with clarity, in After an emergency, the suspect cannot be accurately located and tracked, and the expected effect cannot be achieved

[0003] The traditional pedestrian re-id method is relatively simple. It imports video images from multiple lenses and only relies on feature comparison to process them. It does not use effective three-dimensional spatial information or even temporal information. These information are used as the target It is very valuable to characterize the continuous trajectory of

Images