Scene change detection method and scene change detection system

Abstract

The invention discloses a scene change detection method. The scene change detection method includes the following steps. Firstly, two images of a monitored scene are captured. Secondly, characteristics of the two images are abstracted with a rapid robust feature calculation method. Thirdly, the characteristics of the two images are matched through a K-dimension data space tree algorithm. Fourthly, estimation of transparent transformational matrixes are performed on matched characteristics of the two images through a modified random sampling consistency algorithm. Fifthly, according to estimated transparent transformational matrixes, space offset amount of the two images is worked out so as to obtain a scene change degree. The rapid robust feature calculation method, the K-dimension data space tree algorithm and the modified random sampling consistency algorithm are applied in the scene change detection method which has the advantages of being rapid, good in stability and high in precision. The invention further discloses a scene change detection system which comprises an image capturing module, an image characteristics abstracting module, an image characteristics matching module, a transparent change matrix estimating module and a scene change degree calculation module. The scene change detection method and the scene change detection system are widely applied in the field of video monitoring.

Description

technical field [0001] The invention relates to the field of video monitoring, in particular to a method and system for detecting scene changes. Background technique [0002] In the current video surveillance application, due to the impact of some external force on the camera (such as artificial movement of the camera, displacement of the camera fixing frame, wind or changes in the focal length of the camera, etc.), the actual monitored video scene will deviate from the original intended monitoring scene. In addition, if the camera is blocked by foreign objects, the original scene to be monitored will also be lost. Due to the above-mentioned various reasons, the scene monitored by the camera is partially or completely changed, and the scene originally wanted to be monitored is lost, which may seriously affect the entire video surveillance system. Therefore, we need to detect this abnormal situation in time and accurately and prompt an alarm so that the staff can respond qu...

AI Technical Summary

Problems solved by technology

[0003] Most of the current scene change detection methods are based on grayscale histogram or brightness change methods. These methods will seriously affect the accuracy when the scene illumination changes greatly or there are many moving objects in the scene.

In addition, these methods cannot detect small changes in the scene (such as small camera rotations), nor can they accurately evaluate the degree of scene change.

Images