A loop detection method based on global image barcode in street scene scene

Abstract

The invention discloses a loopback detection method based on a global image bar code in a street view scene, comprising the following steps: encoding all image frames of an existing map, and storing the encoding in a hierarchical data structure; The key frames are also binary coded, compared with the existing binary description codes, and the first K pictures with the smallest Hamming distance are obtained; the feature point extraction and comparison operation is performed on the K pictures and the newly collected image key frames, and the comparison The feature similarity between the two images is used to obtain the image with the highest score, so as to know the position of the robot at this time. The present invention only needs to carry out simple encoding on the image, and the existing algorithm can be used to match the encoding to achieve the purpose, and the operability is high; it does not need to rely on the feature points of the image and advance training in advance, which reduces the amount of calculation; and Compared with the traditional method, it solves the problems of inability to perform loopback detection, high work intensity, and low efficiency.

Description

technical field [0001] The invention relates to an image processing method, in particular to a loop detection method based on a global image bar code in a street view scene. Background technique [0002] Instant Localization and Mapping (SLAM) is the key to autonomous navigation and control of mobile robots in unknown unstructured environments. A complete SLAM system includes four parts: sensor data processing, pose estimation, map building, and loop detection. Among them, the loopback detection mechanism is an important link to solve the closed-loop relocation of the mobile robot and improve the robustness of the SLAM system. Existing pixel-level SLAM positioning methods cannot perform loop-closure detection unless they rely on feature points. The bag-of-words model is a typical method of loop-closure detection using feature points. However, the use of feature points in the SLAM system based on the direct method violates the original intention of using the direct method, a...

AI Technical Summary

Problems solved by technology

The machine learning algorithm does not need to extract feature points for loop closure detection, but it needs to be trained in advance, which takes too long and the effect is unstable

Images