Video occlusion detection method for automobile data recorder

Abstract

The invention relates to a video occlusion detection method for an automobile data recorder. The method comprises an occlusion detection process for obtaining an image sequence composed of a pluralityof video frame images based on continuous sampling, and the occlusion detection process for obtaining the image sequence composed of the plurality of video frame images based on continuous sampling is used for judging whether the sampled image sequence is continuously occluded or not so as to judge whether the video is continuously occluded or not. Whether the whole image is shielded or not is judged by counting the total number of the shielded sub-blocks, so that the misjudgment rate of image shielding is reduced; whether the video is occluded in a time period or not is judged by calculatingand comparing similarity indexes between video frame images before and after a video image sequence, alarm caused by instantaneous occlusion is avoided, meanwhile, false alarm caused by misjudgment of image occlusion is reduced, the accuracy is high, and the robustness is high.

Description

technical field [0001] The invention relates to the technical field of digital video image processing, in particular to a video occlusion detection method for a driving recorder. Background technique [0002] The car driving recorder is used to collect the real-time video of the monitoring camera in the car, and encode the video into digital video and store it in the hard disk. The effectiveness of digital video images is very important, whether it is for post-processing analysis, real-time analysis, or network forwarding and transmission of stored digital video images. If the camera is blocked, the stored digital video images will be obscured images. , the blocked videos are useless data to be excluded. [0003] Therefore, it is necessary to judge whether the monitoring camera is blocked, and when the monitoring camera is blocked, the video blocking fault can be eliminated in time by generating a blocking alarm, so as to ensure that the video stored in the hard disk is a r...

AI Technical Summary

Problems solved by technology

[0004] Because the crowd density in the car varies from time to time, and the human body in the car often moves, the background image inside the car is mostly blocked by foreground objects such as the human body, so it is impossible to use the background comparison method to determine whether the camera is blocked

[0005] When the camera is blocked by an object very close to the lens, the texture edge of the overall video image is blurred. Whether the image is blocked can be judged by the texture feature of the image edge, but it is greatly affected by the ripple and salt and pepper noise of the video image, and only through The method of analyzing the edge ambiguity of a frame of video image to judge whether the camera is blocked cannot judge the continuous blocking state of the video, and it is easy to generate false alarms

Images