Word bag generation method and device based on feature matching

Abstract

The invention provides a word bag generation method and device based on feature matching, and the method comprises the steps: obtaining a two-dimensional image video stream which is shot for a target environment, wherein the target environment is an environment corresponding to a map constructed by target equipment during the simultaneous positioning and mapping of SLAM; extracting a feature descriptor of a target feature key point of each image frame in the two-dimensional image video stream by using a feature extraction algorithm, the number of the target feature key points of each image frame in the extracted two-dimensional image video stream being at least one; performing similarity matching on the feature descriptors of the target feature key points of the image frames in the two-dimensional image video stream according to the sequence of the image frames in the two-dimensional image video stream to obtain leaf nodes of the word bag; taking the number of the leaf nodes and the number of the leaf nodes as input parameters and inputting th input parameters into the text clustering model to construct the word bag, and by means of the method, the time consumed for obtaining the word bag can be shortened.

Description

technical field [0001] The present application relates to the field of computer technology, in particular, to a bag-of-words generation method and device based on feature matching. Background technique [0002] When an intelligent robot performs a task in an unknown environment, it needs to use SLAM (Simultaneous Localization And Mapping, simultaneous positioning and mapping) technology to construct a map. Due to the accumulation of errors in the process of constructing the map, the constructed map is often not a closed map. In order to improve the accuracy of the constructed map, the constructed map can be corrected by using the closed-loop detection algorithm based on two-dimensional images. In the process of detecting closed-loops by the closed-loop detection algorithm based on two-dimensional images, it is necessary to construct a bag of words. In order to make the construction To produce a bag of words with high accuracy, it is necessary to match the constructed bag of ...

AI Technical Summary

Problems solved by technology

[0002] When an intelligent robot performs a task in an unknown environment, it needs to use SLAM (Simultaneous Localization And Mapping, simultaneous positioning and mapping) technology to construct a map. Due to the accumulation of errors in the process of constructing the map, the constructed map is often not a closed map. In order to improve the accuracy of the constructed map, the constructed map can be corrected by using the closed-loop detection algorithm based on two-dimensional images. In the process of detecting closed-loops by the closed-loop detection algorithm based on two-dimensional images, it is necessary to construct a bag of words. In order to make the construction To produce a bag of words with high accuracy, it is necessary to match the constructed bag of words with the feature descriptors of the feature key points of the acquired continuous environmental image frames. In the prior art, different construction parameters are usually used to construct the bag of words. After obtaining multiple word bags, an optimal word bag is selected from the multiple word bags as the basis for correction through experiments. However, the above method requires repeated calculations to obtain the final word bag, resulting in the consumption longer time

Images