Automobile rearview pedestrian detection and alarm method based on fisheye image correction

Abstract

The invention discloses an automobile rearview pedestrian detection and alarm method based on fisheye image correction. The method comprises the steps of using a fisheye lens for collecting information behind a vehicle; extracting an effective area from the acquired fisheye image; re-processing the distortion of the edge of the image; detecting pedestrians in the image, identifying and segmentingthe pedestrians, and predicting the moving pedestrians; and judging whether to give an alarm according to the positions of the pedestrian and the tail and the pedestrian movement trend. The fisheye lens is used for collecting vehicle driving information, and the defects that a traditional camera is limited in visual range and large in installation number are overcome; de-noising and distortion processing are performed on the fisheye image through combined use of multiple image processing algorithms so that accurate pedestrian identification is enabled to be possible. A pedestrian recognition algorithm is used for analyzing the pedestrian position and the movement trend, and finally the effects of prompting and protecting the vehicle driving safety are achieved.

Description

technical field [0001] The invention relates to the fields of computer vision and video image processing, in particular to a method for detecting and alarming pedestrians in rear view of automobiles based on fisheye image correction. Background technique [0002] Pedestrian Detection is the use of computer vision technology to determine whether there are pedestrians in an image or video sequence and give precise positioning. This technology can be combined with pedestrian tracking, pedestrian re-identification and other technologies to be applied in artificial intelligence systems, vehicle assisted driving systems, intelligent robots, intelligent video surveillance, human behavior analysis, intelligent transportation and other fields. As one of the pioneer technologies of unmanned driving, the pedestrian detection system has also become a research and development hotspot in recent years. It is usually integrated into the collision prevention system, which uses cameras and se...

AI Technical Summary

Problems solved by technology

[0009] 1. The detection object of the sound notification system is not only pedestrians. When immovable objects (such as trees and buildings) are detected, the sound emitted by the vehicle is likely to cause noise and affect the surrounding environment;

[0010] 2. In many cities in my country, there are regulations prohibiting honking. This automatic honking system can easily punish the driver;

[0011] 3. Non-fisheye cameras generally do not have a 180° viewing angle, and the accuracy is low

If you want to observe the driving environment of the vehicle in all directions, you generally need to install multiple cameras, which not only increases the cost, but also affects the appearance to a certain extent;

[0012] 4. Installing multiple cameras will increase the lateral volume of the module, and the system design and calculation complexity will increase

Images