Obstacle detection method and device based on monocular camera

Abstract

The invention provides an obstacle detection method and device based on a monocular camera. The method comprises the steps of obtaining a target frame image shot by the monocular camera and the adjacent frame images, shot by the monocular camera, of the target frame image; deleting unstable feature points from an initial feature point set of the adjacent frame images to obtain an optimized feature point set of the adjacent frame image; carrying out sub set partitioning on a target feature point set to obtain multiple target feature point sub sets; for each target feature point sub set, judging whether the target feature point sub set belongs to an obstacle or not according to distance change values, between the target frame moment and the adjacent frame moment, of the points in a corresponding ground projection point set of the target feature point sub set; determining the sum set of all the target feature point sub sets which are judged to belong to obstacles as an obstacle point set in the target frame image. According to the obstacle detection method and device based on the monocular camera, the detection accuracy of obstacle points and the detection efficiency of the obstacles can be improved.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to an obstacle detection method and device based on a monocular camera. Background technique [0002] With the advancement of science and technology, the application of cars and navigation in people's lives is becoming more and more popular, and obstacle detection based on vision is also important in the fields of car assisted driving and robot navigation. Sensors that can be used for obstacle detection include monocular cameras, stereo cameras, laser, radar or infrared sensors, etc. Among them, monocular cameras have broader application prospects in the field of obstacle detection due to their advantages such as easy installation, small size, and low cost. [0003] There are many obstacle detection methods based on monocular cameras in the prior art, such as methods based on appearance features. When the method based on appearance features is implemented, the te...

AI Technical Summary

Problems solved by technology

[0004] However, this type of method is only suitable for detecting specific types of obstacles, such as pedestrians or vehicles. If multiple types of obstacles need to be detected, multiple classifiers need to be trained, and the process of training multiple classifiers is cumbersome and complicated, making The existing technology is inefficient and error-prone when detecting obstacles

Images