A Pedestrian Image Recognition Method Based on Deep Network Model

Abstract

The invention provides a pedestrian image recognition method based on a deep network model, including: performing data preprocessing on the pedestrian image; performing an adaptive sampling algorithm on the preprocessed data to obtain batches with more difficult samples; The model extracts multi-layer features, uses sub-modules to enhance low-level features, then downscales and splices with high-level features to obtain multi-layer features, splits multi-layer features at different granularities to form a multi-branch structure, and extracts component features and global features of each branch. And splicing all the extracted features to obtain the depth representation of the pedestrian image; train the constructed network model; extract the depth representation of the query image through the trained network model, according to the cosine distance similarity between each query image and the query set , returns the recognition results for each query image. Through the multi-level and multi-granularity pedestrian re-identification depth model described above, the present invention achieves the best pedestrian re-identification performance at the present stage.

Description

technical field [0001] The invention relates to the fields of machine learning and computer vision, in particular to a pedestrian image recognition method based on a deep network model. Background technique [0002] With the development of modern society, public safety has gradually attracted people's attention. Shopping malls, apartments, schools, hospitals, office buildings, large plazas and other places with dense crowds and prone to public safety incidents have installed a large number of surveillance camera systems. The research on surveillance video is mainly reflected in the identification of visible objects, especially is pedestrian recognition. This is because pedestrians are generally the target of surveillance systems. More precisely, the task of the surveillance system is to search for a specific pedestrian in the surveillance video data, that is, the task of pedestrian re-identification. [0003] However, on the one hand, the data volume of surveillance video...

AI Technical Summary

Problems solved by technology

[0004] Among the existing pedestrian re-identification methods, the pedestrian re-identification method based on the component-based deep model has the most advanced performance. However, at the current stage, the component-based deep model often obtains component features by segmenting high-level features in the backbone network, and On the one hand, the high-level features of the deep model are highly coupled, and simply splitting the high-level features will lead to the loss of its semantic information, which will limit the performance of the model

Images