Crowd counting method based on deep learning and apparatus thereof

Abstract

Embodiments of the invention disclose a crowd counting method based on deep learning and an apparatus thereof. The method comprises the following steps of dividing a crowd picture to be detected into a plurality of picture blocks; based on an RPN candidate frame acquired from training in advance, generating a model, and determining head candidate frame areas in the plurality of picture blocks and degrees of confidence of the head candidate frame areas; according to the degrees of confidence, screening the determined head candidate frame areas and acquiring an area to be detected; and based on a Fast-RCNN correction model acquired from the training in advance, carrying out classification prediction on the area to be detected, and according to a classification prediction result, determining a crowd number included in the crowd picture to be detected. In the embodiments of the invention, the crowd counting method under high resolution is provided, and accuracy and robustness of crowd counting are increased.

Description

technical field [0001] Embodiments of the present invention relate to the technical fields of computer vision and machine learning, and in particular to a deep learning-based crowd counting method and device. Background technique [0002] Video-based crowd counting has a wide range of applications. Whether it is the analysis of the advertising effect of advertising machines, the early warning of the number of people in security monitoring, or the analysis of the flow of people in tourist attractions, the crowd counting technology based on video is a method worth promoting. Traditional crowd counting methods based on face recognition mostly rely on manual feature extraction and professional domain knowledge, which can achieve certain results. However, when counting portraits with different postures and different sides, feature extraction becomes more difficult, so that these methods cannot make more accurate judgments. [0003] At present, deep learning methods based on neu...

AI Technical Summary

Problems solved by technology

In terms of target recognition and detection, although network models such as Faster-RCNN, YOLO, SSD, and R-FCN all have good performance, these network models have low detection and recognition rates for objects with small targets due to the network structure. status

At the same time, for high-resolution crowd images and videos, because the image size is large, in order to improve the detection efficiency, the above methods will reduce the image to a certain extent, resulting in further loss of image details.

Images