Crowd counting method and system based on feature pyramid, medium and electronic equipment

Abstract

The invention provides a crowd counting method and a crowd counting system based on a feature pyramid, a medium and electronic equipment, which belong to the technical field of computer vision. The crowd counting method comprises the following steps of: preprocessing an acquired image to obtain an initial crowd density map corresponding to the image, inputting the initial crowd density map into apreset feature pyramid network model, and acquiring a feature map fused with multi-scale context information on each level; and carrying out information transmission updating layer by layer from the bottom layer to the top layer, then carrying out reverse information transmission until reaching the bottom layer, fusing the feature map of each layer obtained by bidirectional information transmission to obtain a final feature map of each layer, carrying out reverse layer-by-layer side edge connection to obtain a final crowd density map, and further acquiring a final crowd counting value. According to the crowd counting method and the crowd counting system, the multi-layer features are integrated through bidirectional message transmission, the problem of scale change in crowd pictures is solved by fusing feature maps of different scales in a network layer by layer, and more detail information is reserved, so that counting of multi-scale dense crowds with a better effect is realized.

Description

technical field [0001] The present disclosure relates to the technical field of computer vision, and in particular to a feature pyramid-based crowd counting method, system, medium and electronic equipment. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. [0003] Crowd analysis has important geopolitical and civic applications. Large mass gatherings are commonplace at candlelight vigils, pro-democracy protests, religious gatherings and presidential rallies. Civic agencies and planners rely on crowd counting to regulate access points and plan disaster responses for such events. Accurately estimating total crowd size from images has become increasingly important for crowd management and public safety. Crowd counting is a challenging task. Crowd images, especially high-density crowd images, have serious problems of crowd overlap and scale change. ...

AI Technical Summary

Problems solved by technology

[0006] The inventors of the present disclosure found that since a single-column convolutional neural network (Convolutional Neural Networks, CNN) only contains a receptive field of one scale, it is difficult to deal with the problem of scale change. The researchers proposed a series of multi-column CNN, multi-input CNN and multi- The task learns the CNN structure, but these multi-column CNNs generally use the highest-level feature map to regress to generate a density map, and the high-level feature map will lose more detailed information after layer-by-layer abstract expression and downsampling of the pooling layer. Even some small-scale targets are filtered out, resulting in poor counting ability of the algorithm for small targets

Images