Object detection method and system based on dynamic memory and motion perception

Abstract

The invention belongs to the technical field of computer vision, in particular to a target detection method and device based on dynamic memory and motion perception, aiming at solving the problem of low target detection accuracy caused by video false detection. The method comprises the following steps: obtaining a feature map corresponding to a current frame image in a target video by using a neural network, and obtaining a target candidate frame; obtaining a motion memory feature map corresponding to the current frame image according to the feature map with the highest resolution and the motion memory feature map corresponding to the previous frame image; obtaining a motion feature map of the current frame according to the motion memory feature map corresponding to the current frame imageand the feature map with the maximum resolution; The feature map with the highest resolution is fused with the motion feature map of the current frame image to obtain the fused feature map; obtainingfusion features of each target candidate frame according to the fusion feature map; The fusion feature is used for target detection. A more robust and stable target detection result can be obtained based on the above method.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a target detection method and system based on dynamic memory and motion perception. Background technique [0002] The task of object detection is to find out the objects of interest in the image or video, and detect their position and size at the same time, which is one of the core problems in the field of computer vision. With the application and development of convolutional neural networks, object detection based on single-frame images has made great progress, but object detection based on video still has certain characteristic difficulties, such as motion blur and video defocus. [0003] At present, there are mainly two ways to suppress video false detection in video object detection. The first method is to use the detector to detect each frame of the video separately, and then use the heuristic algorithm to post-process the detection results of each frame...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above problems in the prior art, that is, to solve the problem of low target detection accuracy caused by video false detection, one aspect of the present invention provides a target detection method based on dynamic memory and motion perception, include:

Images