A Fast Pedestrian and Vehicle Detection Method Based on Object Consistency

Abstract

The invention discloses a method for quickly detecting pedestrians and vehicles based on a road scene of a vehicle-mounted binocular camera. The method builds a road scene columnar model, and searches and detects pedestrian and vehicle targets within the limited range of the columnar model. At the same time, the detection results of the front and rear frames and the left and right frames are matched, and the detection results of the left image of the current frame are optimized by using the front and rear matching pairs and the left and right matching pairs. The invention aims at the problem that the detection of pedestrians and vehicles based on the monocular method is not accurate enough, and optimizes the detection result by using the matching results of the front and rear frames and the left and right frames. Aiming at the problem that pedestrian and vehicle detection based on binocular images cannot be real-time, the columnar model of the road scene is used to reduce the detection area and speed up the detection speed. The invention integrates rich information of binocular images, can increase the accuracy and detection speed of pedestrian and vehicle detection, and also proposes a rapid pedestrian and vehicle detection framework based on binocular image road scenes.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and specifically relates to a vehicle-mounted camera-based fast pedestrian and vehicle detection method based on target consistency. Pedestrian and vehicle detection in automatic driving assistance systems, etc. Background technique [0002] Pedestrian and vehicle detection is widely used in applications such as intelligent driving assistance. The vision-based intelligent vehicle is to install a camera lens on the car, and use computer vision technology to obtain guidance information by processing the images captured by the camera lens. Compared with other guidance technologies, visual guidance does not need to add facilities on the road, and is suitable for passing vehicles, pedestrians, temporary facilities and other random targets that cannot be modeled and determined in advance. With the increase in the number of vehicles in our country, the increasingly complex road traffic conditi...

AI Technical Summary

Problems solved by technology

Under this premise, first of all, the matching algorithm required to construct the depth map often takes a long time, and the matching effect is not necessarily accurate; secondly, for the detection of pedestrians and vehicles, although the target detection method based on image processing is fast in terms of detection speed It meets the needs of road scenes, but the detection accuracy is often not enough, and the pedestrian detection method based on learning often has a large amount of calculation, which cannot meet the needs of real-time detection in road scenes

Images