Video quality diagnosis method for foreign object occlusion

Abstract

The invention relates to a method for diagnosing video quality blocked by foreign objects. First, the RGB image is bilaterally filtered, and the filtered image is converted to HSV space, then the gradient value and gradient direction of the image are calculated in HSV space, and the image straight line is extracted by using the gradient value and gradient direction. Then merge the extracted straight lines, select two adjacent and non-intersecting straight lines, and obtain the largest rectangular area between the two straight lines. Candidate regions are determined by calculating the standard deviation of the H and S components of the rectangular region and the structural similarity value of the rectangular region. Finally, foreign object detection is completed by merging candidate regions. The present invention does not need to use reference images, avoids the need to introduce a variety of scene matching management mechanisms due to the polymorphism of the reference images itself, and does not cause loss of accuracy due to subjective judgment criteria in the matching mechanism of reference images. The result is objective and accurate.

Description

technical field [0001] The invention belongs to the field of computer vision and can be used for an abnormality detection system in which a monitoring video picture is blocked by foreign objects (bigger band-like or planar objects under two-dimensional imaging). In particular, it relates to a video quality diagnosis method for foreign object occlusion. Background technique [0002] With the popularization and application of dome surveillance camera equipment in various monitoring places, while improving the operation experience and monitoring efficiency, some irregular operation behaviors also have a negative impact on the daily monitoring business. The camera is not reset to the preset point after the zoom-in operation, so that the imaging of the objects (trunks, fences, billboards, utility poles, walls, etc.) that were originally small and secondary targets in the scene occupies the camera ROI (region of interest); another example, when the dome camera is rotated, a large...

AI Technical Summary

Problems solved by technology

[0003] At present, the research on video occlusion at home and abroad is mainly divided into two categories: one is to use the comparison method between the reference image and the image to be detected to complete the occlusion detection, and the other is to extract a series of local feature information in the reference image and the image to be detected Image matching is used for occlusion detection. However, in the actual application of surveillance video occlusion detection, it is difficult to define the reference image of the original video in most cases due to the complex and changeable surveillance scenes. Therefore, the above two methods are not suitable for surveillance video occlusion. detection

Images