Pedestrian detection method and device

Abstract

The embodiment of the invention provides a pedestrian detection method and device. The pedestrian detection method comprises the steps that an image to be processed is acquired; the pedestrians in theimage to be processed are detected so as to acquire the pedestrian information, wherein the pedestrian information includes multiple pedestrian boxes; the heads of the pedestrians in the image to beprocessed are detected so as to acquire the head information, wherein the heat information includes the head positions corresponding to at least a part of the pedestrian boxes of multiple pedestrian boxes in a one-to-one way; the shoulders of the pedestrians in the image to be processed are detected so as to acquire the shoulder information, wherein the shoulder information includes the left shoulder positions and the right shoulder positions which are corresponding to at least a part of the pedestrian boxes in a one-to-one way; the overlapping situation of the head and shoulder parts corresponding to at least a part of the pedestrian boxes in a one-to-one way is calculated according to the head positions and the left shoulder positions and the right shoulder positions which are corresponding to at least a part of the pedestrian boxes in a one-to-one way; and the pedestrian boxes are filtered according to the overlapping situation so as to acquire the pedestrian detection result. According to the method and the device, the problem that the pedestrian boxes of other pedestrians are filtered by using the pedestrian box of a certain pedestrian can be avoided.

Description

technical field [0001] The present invention relates to the field of computers, and in particular to a pedestrian detection method and device. Background technique [0002] In the field of surveillance, pedestrian detection plays a very important role. Traditional pedestrian detection methods often have the following problems. Traditional pedestrian detection methods often use the sliding-window method to extract a variety of windows of different scales from the image to be processed (each window is a rectangular frame, which can also be called a pedestrian frame), and judge each window. Whether there are pedestrians in . In addition, traditional pedestrian detection methods usually use non-maximum suppression (NMS) to filter multiple windows on the same pedestrian. This often leads to the filtering out of windows containing other pedestrians in crowded conditions, resulting in serious false negatives (that is, pedestrians, but not detected). Contents of the invention ...

AI Technical Summary

Problems solved by technology

This often leads to the filtering of windows containing other pedestrians in crowded situations

Images