A Monocular Visual Odometry Method Fused with Edge Features and Deep Learning

Abstract

The invention discloses a monocular visual odometry method for fusing edge features and deep learning, and relates to the technical field of visual odometry. Since there are few features that can be extracted in an image in a low-texture motion scene, the visual odometry is in progress. The problem of missing feature data will occur during feature matching, and cause the accuracy of pose estimation to decrease. The innovation of the present invention is a monocular visual odometry method that combines edge features and deep learning. First, the edge enhancement algorithm is designed based on the Canny edge detection algorithm. The image data set after edge enhancement is used as the input of the convolutional neural network and features Extraction, the output of the convolutional neural network is then input into the cyclic neural network for calculation, and finally the entire model is output to estimate the camera pose and optimize the feature extraction. Experimental results show that the algorithm can learn more image features during model training, improve the accuracy of pose estimation, and exhibit superior performance in low-texture scenes.

Description

technical field [0001] The invention relates to the technical field of visual odometry, in particular to a monocular visual odometry method that combines edge features and deep learning. Background technique [0002] Visual odometry is a method for estimating ego-motion from input images, which is a core module of instant localization and map building systems. Since the monocular visual odometry (Visual Odometry, VO) can determine the current position according to the feed signal of the camera, it has become a hot research field in the field of computer vision. It has a wide range of applications in autonomous driving, robotics and other fields. In recent years, visual odometry from stereo cameras has achieved great progress and has been widely used due to its reliable depth map estimation capability. However, once the distance between the camera and the baseline differs from the scale of the scene, it degenerates to a monocular distance. [0003] Different from stereo vi...

AI Technical Summary

Problems solved by technology

However, once the distance between the camera and the baseline differs from the scale of the scene, it degenerates to the monocular distance

[0003] Different from stereo vision odometry, monocular VO cannot obtain the real-scale environment map and robot motion, so it needs to use prior knowledge or information such as camera height to estimate the absolute scale, which makes monocular VO easier to produce than stereo VO Bigger drifts, more challenging

Images