Pedestrian searching method based on partial shared network and cosine interval loss function

Abstract

The invention relates to a pedestrian searching method based on a partial shared network and a cosine interval loss function. The method comprises the following steps: firstly, designing a new neuralnetwork structure, enabling pedestrian detection and pedestrian re-identification parts to share shallower features, enabling the pedestrian detection and pedestrian re-identification parts to focus on respective tasks, and reducing mutual interference between the pedestrian detection and pedestrian re-identification parts to a certain extent from the perspective of improving a model structure; and secondly, deeply researching the influence of the weight of the pedestrian re-identification loss function on model optimization in multi-loss function joint optimization, and relieving the mutual interference between pedestrian detection and re-identification from the perspective of optimization by setting reasonable loss function parameters; finally, providing a more robust lookup table updating strategy, adding cosine intervals into an OIM loss function to reduce the distance between similar samples, and finally, enabling pedestrian features of network learning to be more differentiated.According to the invention, mutual interference between pedestrian detection and pedestrian re-identification can be reduced.

Description

technical field [0001] The invention relates to the technical field of computer vision applications, in particular to a pedestrian search method based on a partially shared network and a cosine interval loss function. Background technique [0002] Pedestrian re-identification aims to match target pedestrians from non-overlapping multi-camera surveillance systems. It is a very important and fast-growing research field in the field of computer vision. At present, pedestrian re-identification has many applications in the field of video surveillance, such as searching for criminal suspects from the crowd, cross-camera pedestrian tracking, pedestrian activity analysis, etc., which are of great significance to the protection of public life and property safety. Therefore, in recent years, Pedestrian re-identification technology has attracted extensive research in academia and industry. Although many person re-identification data sets and related algorithms have been proposed, ther...

AI Technical Summary

Problems solved by technology

The last one is the pedestrian re-identification loss function widely used in the end-to-end pedestrian search network OIM loss function lacks the ability to distinguish pedestrians, because it only considers correctly distinguishing different types of samples, but ignores the optimization of similar samples. similarity

Images