An improved method for detecting moving objects based on gmm

Abstract

The invention discloses an improved algorithm based on GMM moving target detection. The algorithm comprises steps that (1), during moving target detection, a moving detection feedback mechanism is added; when a moving target is determined to move quite slowly or be static, a GMM background model is not updated in a target region, and the GMM background model is updated outside the target region according to GMM background model update rules; and (2), a background image generated through modeling through utilizing newest video frames and a GMM background is utilized to do difference processing to acquire a foreground image a; a foreground image b acquired through the improved GMM is utilized to fuse the image a and the image b in a space domain, and a detection foreground image of the moving target is acquired. The algorithm is advantaged in that the large and slow moving target can be accurately detected, when the moving target suddenly stops and stays for a while, the target is not determined to be background, and the target can be further detected.

Description

technical field [0001] The invention relates to the field of video monitoring, in particular to an improved algorithm for detecting moving targets based on GMM. Background technique [0002] With the development of computer technology, video surveillance is becoming more and more intelligent. One of the main tasks of an intelligent video surveillance system is to detect, identify, and track objects or interested parts in video images. The correct detection of the target is the premise of video surveillance, and the detection effect of the target will affect the accuracy and robustness of the subsequent target recognition and tracking. In real application scenarios, unchangeable factors such as changes in lighting in the environment, disturbance of leaves in the background, and camera shake will affect the effect of moving target detection. [0003] There are many methods of target detection. Commonly used detection algorithms include inter-frame difference method, backgroun...

AI Technical Summary

Problems solved by technology

In a real application scenario, due to changes in lighting in the environment, unchangeable factors such as the disturbance of leaves in the background and the shaking of the camera itself will affect the effect of moving target detection

[0003] There are many methods of target detection. Commonly used detection algorithms include inter-frame difference method, background difference method, and optical flow method. The inter-frame difference method compares video frames at fixed intervals, which is suitable for dynamically changing environments. A large area of ​​holes is generated, and the integrity of the extracted target is poor; the optical flow method is difficult to meet the real-time performance of motion detection due to its complex calculation; This method can extract the target well and completely, and it is greatly affected by the change of illumination and background.

[0004] As a related technology, GMM background modeling can continuously update the background model. It can also accurately detect moving objects in the change of illumination and the disturbance of leaves, which improves the quality of motion detection. Object detection is not very good

Images