Pedestrian library cross-channel aggregation method based on face recognition

Abstract

The invention discloses a pedestrian library cross-channel aggregation method based on face recognition. The pedestrian library cross-channel aggregation method comprises the following steps that S1,the highest-quality snapshot image in all snapshot images of one person is reserved through short-time single-channel aggregation; S11, caching the information caching queue of the snapshot image; S12, when another snapshot image is cached, comparing the snapshot image with the previous snapshot image; S13, after the set caching time is up, enabling the cached snapshot images to enter a cross-channel aggregation operation; S2, storing the snapshot images into a pedestrian library through cross-channel aggregation; S21, selecting a representative graph from each personnel file in the pedestrianlibrary; S22, setting a comparison threshold value, comparing the snapshot images entering the cross-channel aggregation operation in the step S13 with representative images of all personnel files ina pedestrian library, and searching personnel files similar to the snapshot images; and S23, finding a personnel file similar to the snapshot image, and storing the snapshot image in the personnel file.

Description

Technical field [0001] The present invention relates to the field of image processing, in particular to a cross-channel aggregation method of passerby database based on face recognition. Background technique [0002] Artificial intelligence technology is more and more widely used in the security industry, and one of the cores is face recognition technology. With the continuous landing of face recognition applications in security, a large amount of data has been accumulated, one of which is the passerby database. [0003] The so-called passerby library is formed when security applications access cameras, face capture machines and other front-end devices, which recognize faces in video streams through technologies such as video decoding and face recognition, and store them in the library. [0004] The retrieval process of the passerby library consumes a lot of GPU and CPU resources, and the results are not accurate, so it is necessary to aggregate and archive the passerby library. Ho...

AI Technical Summary

Problems solved by technology

[0004] The retrieval process of the passer-by database consumes a lot of GPU and CPU resources, and the results are inaccurate, so it is necessary to aggregate and archive the passer-by database

However, the traditional face aggregation scheme has the problems of poor aggregation effect, low aggregation efficiency, and great impact on image quality.

Images