Obstacle early warning method and obstacle early warning device

Abstract

The present invention discloses a method and device for early warning of obstacles, wherein the method includes obtaining scene images at the current sampling time and the previous sampling time respectively, and a first relative distance map from each viewpoint to the moving device within the field of view, according to In this way, the outline of the obstacle, the calibration information and the second relative distance map from the obstacle to the moving device at the current sampling time are obtained, according to the first relative distance map at the previous sampling time, the outline of the obstacle and the calibration information at the current sampling time The motion vector from the obstacle at the current sampling moment to the corresponding obstacle at the previous sampling moment is used to calculate the second relative distance map between the obstacle and the moving device at the previous sampling moment; accordingly, the relative distance between the obstacle and the moving device at the current sampling moment is calculated speed; furthermore, predict the collision time between the obstacle and the moving device at the current sampling moment. The embodiment of the invention solves the technical problem of how to quickly and accurately predict obstacles.

Description

technical field [0001] Embodiments of the present invention relate to the technical fields of pattern recognition, machine learning, and video image processing, and in particular, to an obstacle early warning method and an obstacle early warning device. Background technique [0002] In addition to radar ranging and binocular vision ranging, the currently used algorithms for the relative distance between obstacles and moving devices such as cars, robots, and drones have low precision and missed obstacles in the ranging scheme using monocular vision. The downside of false positives. [0003] To obtain the relative speed of the obstacle and the moving device, it is necessary to track and match the two measured obstacle depths to obtain the corresponding relationship of the same obstacle in the two sampling data. The currently used obstacle tracking algorithm needs to perform feature point matching and obstacle segmentation on the twice-sampled obstacle information. Whether it ...

AI Technical Summary

Problems solved by technology

The currently used obstacle tracking algorithm needs to perform feature point matching and obstacle segmentation on the twice-sampled obstacle information. Whether it is based on a depth map or an image-based feature point matching scheme, there are disadvantages of large amount of calculation and low accuracy.

Therefore, the accuracy of the relative speed algorithm of the current moving device is low, the calculation speed is slow, and it is impossible to make accurate predictions for obstacles

Images