Arbitrary distortion image line segment detection method based on Bezier curve

Abstract

The invention relates to an arbitrary distortion image line segment detection method based on a Bezier curve. The method comprises the steps of firstly constructing a training data set; constructing aline segment detection network; constructing a loss function of a line segment detection network through intersection prediction loss, line segment prediction loss and classifier loss, inputting dataof the training set into the line segment detection network in a loop iteration mode, and according to the constructed loss function of the line segment detection network, using a self-adaptive moment estimation optimizer to optimize and update the network; obtaining a trained line segment detection network; and enabling the to-be-detected image to pass through the trained line segment detectionnetwork to obtain a line segment detection result in the to-be-detected image. According to the invention, the distortion line segment in the distortion image is directly detected without pre-distortion removal processing, so that the operation is simpler and more convenient; the method provided by the invention does not depend on camera parameters, so that compared with methods depending on camera distortion parameters, such as Hough transform, the method provided by the invention is wider in application range and better in robustness.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a method for detecting line segments of arbitrary distorted images based on Bezier curves. Background technique [0002] Line segment detection is one of the most basic tasks in the field of computer vision and robotics. It can help many upper-level vision tasks, such as image matching, camera calibration, SfM (Structure from Motion), visual SLAM (Simultaneous Localization And Mapping), etc. . However, most current line segment detection methods model line segments as straight lines, so they can only detect straight line segments without distortion, but cannot detect line segments bent due to distortion, especially distorted line segments in fisheye camera images and spherical camera images . However, the existing distorted line segment detection methods basically depend on the camera distortion parameters. One of them is to use the camera distortion parameters to corr...

AI Technical Summary

Problems solved by technology

Based on the above-mentioned line segment representation, the present invention designs an end-to-end line segment detection network. At the same time, we construct two fisheye image data sets and one spherical image data set by synthesizing data sets and manual labeling methods to solve the problem of network training. The Problem of Lack of Datasets

Images