Long-term cross-camera target association method and system based on color and texture suppression

Abstract

The invention relates to a long-term cross-camera target association method and system based on color and texture suppression. The method comprises the following steps: inputting a training image in a long-term cross-camera target association data set; performing texture suppression on the training image; performing color suppression on the image subjected to texture suppression; performing identity feature extraction on the image subjected to color and texture suppression by using an identity representation model; supervised learning is carried out on the extracted identity features, and parameter optimization of the identity representation model is realized; and extracting identity features of the to-be-queried pedestrian image and the candidate pedestrian image by using the learned identity representation model, calculating the distance between the features as a similarity score, and selecting the candidate pedestrian with the highest similarity score as an associated target in the cross-camera scene. According to the method, the problem that target pedestrians are difficult to match in a long-term cross-camera scene can be solved, the method can be introduced into an intelligent monitoring system, a target tracking system and a target association system, and a more complete intelligent monitoring technology is realized.

Description

technical field [0001] The invention belongs to the field of robot vision technology and intelligent monitoring, and in particular relates to a long-term cross-camera target association method and system based on color and texture suppression; Influence, solve the problem of uncertain appearance characteristics of pedestrians in long-term scenes, so as to achieve long-term cross-camera target association and matching. Background technique [0002] The cross-camera target association task aims to solve how to match pedestrian target identities in cross-camera scenarios to achieve continuous tracking and trajectory association. In particular, the existing cross-camera target association tasks are often aimed at short-term scenarios, that is, it is assumed that the target pedestrian will not change clothes in the cross-camera scenario. In order to break through this limitation and no longer restrict whether the target pedestrian's clothing changes, the need to solve long-term ...

AI Technical Summary

Problems solved by technology

Under the premise of ensuring that the identity information is not lost, the color and texture information in the pedestrian image is weakened, and the influence of the color and texture information is weakened, and the identity information of the pedestrian is focused on, so as to solve the problem of pedestrian clothing change in the long-term target matching task.

Images