Mobile target in complex background automatic testing method and device

Abstract

A method for automatically detecting movement object under complicated background includes picking up character block, estimating movement, making inter-frame registration difference and carrying out post treatment. The detection device used for realizing said method consists of character block picking up module, movement estimating module, inter-frame registration differencing module and post treating module.

Description

technical field [0001] The invention relates to a method and a device for automatically detecting a moving object in an image captured by a video capture unit. Background technique [0002] In real life, a large amount of meaningful visual information is contained in motion. In order to monitor or track moving targets in the video images captured by the video capture unit, such as airplanes, cars, people, etc., the targets in the video must be detected in advance. Automatic detection of moving objects has broad application prospects in traffic monitoring, parking lot scheduling monitoring, accident detection, automatic navigation and other systems. [0003] In the complex background, there is usually a lot of noise, and besides the target motion, the background also participates in the motion, all of which bring difficulties to the automatic detection of the target in the complex background. Each frame of image is composed of background and target, and the target is nested ...

AI Technical Summary

Problems solved by technology

The optical flow method based on the optical flow field is poor in real-time and practicability, and the block matching method is better for motion estimation, but it is unreliable to perform block matching on a large flat area, because the area around the matching template They are all too similar to templates, and only matching in feature space can not only ensure the reliability of motion estimation, but also minimize redundant operations, so the extraction of background feature blocks in video frames has become a research hotspot.

Some scholars have proposed the Barnard operator, which first extracts feature points, and then forms feature blocks around the feature points. This method extracts a large number of feature points and is dense, which brings trouble to the block matching operation.

Images