Vehicle detection system based on binocular stereo vision and method thereof

Abstract

The invention discloses a vehicle detection system based on binocular stereo vision and a method thereof. The system comprises binocular parallel cameras, a DSP processor, communication and other modules. The system is vertically installed above a lane in a top-down mode so as to carry out real-time detection on road vehicles. The method comprises the following steps of (1) carrying out main point difference calibration on left and right cameras in advance and acquiring an accurate main point difference parameter; (2) collecting left and right images in real time so as to carry out detection and extraction of a foreground moving object; (3) carrying out stereo matching on an extracted and acquired foreground object area and acquiring a current frame parallax image; (4) carrying out subsequent processing on the acquired parallax image and removing a false detection area, simultaneously acquiring a two-dimensional ground plane mapping image; (5) through calculation, acquiring correlation parameters of a speed, a vehicle height, a vehicle model and the like. In the invention, the binocular cameras are used; a stereo vision principle is used so as to acquire depth information of the object; problems that a monocular vision technology is sensitive to light changes and is easy to be disturbed by shadow are solved.

Description

technical field [0001] The invention relates to the application field of computer vision technology, in particular to a vehicle detection system and method based on binocular stereo vision. Background technique [0002] With the intelligent development of urban traffic system, the demand for the detection and statistics of road traffic vehicles is increasing day by day. Existing vehicle detection methods mainly include ultrasonic, infrared, and ring induction loop detection, but in the application, there are problems such as weak anti-interference ability, low detection accuracy, inability to adapt to multi-lane detection environments, and failure to obtain vehicle-related parameters such as vehicle speed and model. and other defects. At present, the vehicle detection method based on computer vision has become a research hotspot and a future development trend. This type of method is divided into: a method based on monocular vision and a method based on binocular stereo vis...

AI Technical Summary

Problems solved by technology

However, this type of method is cumbersome to calibrate the camera, and the disparity matching and target segmentation algorithms have a large amount of calculation and low precision. The resulting vehicle detector has poor real-time performance and adaptability, and low precision.

Images