Loop detection method and device, electronic equipment and storage medium

Abstract

The invention discloses a loopback detection method and device, electronic equipment and a computer readable storage medium. The method comprises steps of obtaining a to-be-detected multi-channel image, inputting the multi-channel image into a pre-trained CNN neural network, and extracting a target feature map and a high-dimensional vector outputted by a preset layer of the CNN neural network; ifthe historical key frame corresponding to the multi-channel image exists, determining cosine similarity between a historical vector of the historical key frame and a current high-dimensional vector; if the cosine similarity is greater than a first threshold, determining the matching degree of the historical feature map corresponding to the historical key frame and the target feature map; and if the matching degree is greater than a second threshold value, determining that a final detection result is loopback. According to the method, the CNN neural network is utilized to realize the output ofthe feature map and the high-dimensional vector, and the feature map and the high-dimensional vector are combined to jointly complete loop detection confirmation, so a multi-frame matching process isavoided, loop detection efficiency is improved, and resource waste caused by storage, migration and matching processes is avoided.

Description

technical field [0001] The present application relates to the technical field of image processing, and more specifically, to a loop detection method and device, an electronic device, and a computer-readable storage medium. Background technique [0002] The traditional loop closure detection scheme usually includes the following process: use the feature extraction method to generate feature data for key frames, use a certain structural model to model the feature points, such as spanning trees, generating dictionaries, etc., and then perform feature trees or The feature dictionary is matched, and the similarity comparison is performed. If multiple frames of the same key frame are confirmed, the loopback detection result is updated. [0003] However, in the above traditional detection schemes, different feature extraction operators will affect the complexity and memory overhead of the generated query dictionary or feature tree, and if the generated dictionary is large in size,...

AI Technical Summary

Problems solved by technology

[0003] However, in the above traditional detection schemes, different feature extraction operators will affect the complexity and memory overhead of the generated query dictionary or feature tree, and if the generated dictionary is large, it is not conducive to storage and migration

However, due to the need to compare a large amount of data during matching, the efficiency is not high

Images