Road structure prediction and target detection method based on multi-task neural network

Abstract

The invention provides a road structure prediction and target detection method based on a multi-task neural network, and relates to the fields of automatic driving, deep learning, computer vision and the like. The method comprises the steps: constructing a multi-task neural network of a context parameter sharing mechanism, wherein the multi-task neural network has the functions of predicting a road structure and detecting a target at the same time; building a loss function mathematical model through the loss between the road structure predicted value and the vehicle layout predicted value and the real value; making a data set through an image and a map, and carrying out closed-loop training on a prediction part of the network; and finally, deploying the method on an automobile, and applying the method to road structure prediction and target detection. According to the method, the multi-task neural network can complete road structure prediction and target detection functions only through image information, and road structure and target prediction can be carried out on invisible and shielded areas in the image.

Description

technical field [0001] The invention belongs to the fields of automatic driving, deep learning, computer vision and the like, and is a road structure prediction and target detection method. Background technique [0002] With the development of deep learning, autonomous driving technology has become more and more mature, and the car's ability to perceive the environment has gradually improved. Most of today's autonomous driving platforms still use some powerful sensors (such as lidar, precision GPS, etc.) for environmental perception solutions, but most of these sensors are expensive and bulky. If only visual sensors are used to complete environmental perception tasks, it will be greatly cut costs. Most of the neural networks currently used for road structure prediction and object detection can only use the information presented in the image to make predictions, but cannot take advantage of the occluded clues in the image. During the training process, most of the networks m...

AI Technical Summary

Problems solved by technology

[0003] To sum up, the problem of the existing technology is that the cost of sensors such as laser radar and precision GPS is relatively high for environmental perception

In this patent, the neural network can only complete the target detection task, and cannot predict the distance of the target relative to the vehicle itself and the distribution of targets ahead. This patent discards the spatial relationship between the road and the target in the input image, and can only use the image surface presented information for reasoning

Images