Noise point filtering method for obstacle avoidance of robot laser radar

Abstract

The invention provides a noise point filtering method for obstacle avoidance of robot laser radar. Specifically, the method includes the following steps of: arranging a distance threshold; reading allscanning points of the laser radar and distances thereof from a robot; taking three consecutive points as samples from a certain point of all the scanning points, and determining outliers according to the distances between each two of the three points in the samples; if there is the outliner, calculating the distances between the connection lines of the outliner and the other two points, and performing comparison of the distances and the distance threshold obtained in the step 1; if the distances are larger than the distance threshold, considering that the point is a noise point, and allowingthe distance between the noise point and the robot is set as 0 or infinity; if the distances are smaller than the distance threshold, allowing processing not to be carried out; and if there is no anoutliner, allowing the sampling not to be processed, and continuously performing sampling until all the scanning points at least complete once comparison and return back an initial point. The noise point filtering method is good in timeliness and facilitates promotion of artificial intelligence productization.

Description

technical field [0001] The present invention relates to the technical field of laser radar, in particular to a noise filtering method for robot laser radar obstacle avoidance. Background technique [0002] Existing low-cost laser radars applied to robots will have jumping noises when scanning the surrounding environment. The noises will cause the robot to mistakenly think that there are obstacles at the noises and avoid them, resulting in deviation from the walking route and even affecting the robot's vision. Walk normally. The filtering algorithms in the prior art cannot meet the requirements of robot obstacle avoidance in terms of real-time performance or correctness. Because: 1. These filtering algorithms need to be sampled to a certain extent, and the obstacle information required by the robot obstacle avoidance must be real-time, and the lag of obstacle information will cause problems in obstacle avoidance. For example, if the lidar sees that the obstacle is 1m away f...

AI Technical Summary

Problems solved by technology

[0002] Existing low-cost laser radars applied to robots will have jumping noises when scanning the surrounding environment. The noises will cause the robot to mistakenly think that there are obstacles at the noises and avoid them, resulting in deviation from the walking route and even affecting the robot's vision. walk normally

The filtering algorithms in the prior art cannot meet the requirements of robot obstacle avoidance in terms of real-time performance or correctness.

Because: 1. These filtering algorithms need to be sampled to a certain extent, and the obstacle information required by the robot obstacle avoidance must be real-time, and the lag of obstacle information will cause problems in obstacle avoidance.

For example, if the lidar sees that the obstacle is 1m away from the robot, the filtering time is 0.5s. After 0.5s, the robot is actually only 0.5m away from the obstacle, but the filtering time is before 0.5s, and the distance of the obstacle is considered to be 1m, there is time delay and space error

2. The existing filtering algorithm does not consider the information related to the characteristics of obstacles, and mistakenly filters out the jumping points

For example, if there is indeed an object jumping in front of the robot, the robot will think that the obstacle does not exist after being filtered by the existing filtering algorithm, resulting in a collision with the obstacle

Images