A roadside edge detection method based on deep learning

Abstract

A roadside edge detection method based on deep learning comprises the following steps of (1) collecting image data including a roadside edge on a real road, marking the position and category information of a target related to roadside edge detection through a manual marking method, and constructing a roadside edge detection data set; (2) constructing a multi-task convolutional neural network suitable for roadside edge detection and a corresponding loss function; and (3) inputting the acquired image and the labeled data into the convolutional neural network constructed in the step (2), updatingparameter values in the neural network according to a loss value between an output value and a target value, and finally obtaining ideal network parameters. The method has good detection capability on various visible and invisible roadside edges with clear geometrical characteristics and height difference, has cost advantage compared with detection modes such as 3D laser radar and the like, is beneficial to large-scale popularization and application, and promotes the development of an automatic driving technology.

Description

technical field [0001] The invention belongs to the technical field of intelligent driving, and relates to a target detection method in a specific scene combining computer vision with deep learning. Background technique [0002] Road edge detection is one of the important components in the field of autonomous driving and active safety. It can help autonomous vehicles identify the current drivable area and judge path information. [0003] Due to the significance of road edge detection, many institutions at home and abroad have provided some detection methods for this kind of problem. However, most of the existing road edge detection technologies are based on 3D lidar. The height changes between roads are used to detect road edges. The limitations of this detection method are as follows: [0004] (1) The price of 3D lidar is relatively high, and it is difficult to apply it on a large scale; [0005] (2) The edge of the road without clear geometric features and height differe...

AI Technical Summary

Problems solved by technology

[0004] (1) The price of 3D lidar is relatively high, and it is difficult to apply it on a large scale;

[0005] (2) The edge of the road without clear geometric features and height difference cannot be recognized;

[0006] In fact, with the development of deep learning technology, many efficient target detection algorithms have emerged, but unfortunately, these target detection algorithms are not suitable for roadsides that occupy few pixels, have no clear geometric features, and are continuous lines. target does not apply

Images