Target detection method fused with multi-angle information

Abstract

The invention discloses a target detection method fused with multi-angle information, and relates to the field of human body detection. The method comprises the following steps of: obtaining a moving foreground to reduce false detection rate, then performing human body detection within the foreground range and directly associating the information among vision angles by affine transformation to make up a miss detection situation of human body detection under single vision angle by a human body detected out via various vision angles, so as to achieve the purpose of improving the detection rate. By adopting the method, the accuracy of space position mapping is improved, the target matching difficulty is reduced, and the situation of sharply increased position deviation is avoided, in this way, a target human body can be mapped accurately, and meanwhile, the shield problem caused by single vision angle is solved to certain extent by using rapid matching information fusing.

Description

technical field [0001] The invention relates to the field of human body detection, in particular to a target detection method for fusing multi-view information. Background technique [0002] Human body detection has always been an important and basic problem in many human-computer interaction systems and intelligent monitoring systems. Existing human detection directions are mainly carried out under single-view, multi-view and recent hot depth cameras. The pedestrian detection method under single-view has always been the focus of many scholars. See references [1] and [2]. This method generally finds the characteristics of the human body different from the background and other objects in a single picture. The advantage is that The human target can be obtained through a single screen, but due to the limited single-view information, many algorithms will have missed and false detections. [0003] In order to overcome the lack of single-view information, Zhengzhi Zhan and K.R.S...

AI Technical Summary

Problems solved by technology

[0006] The multi-view method indirectly obtains the position information of the horizontal plane through the information between the views. Therefore, when the angle of the mapping line is relatively small, the position deviation increases sharply, resulting in inaccurate spatial position mapping and increasing the difficulty of target matching.

Images