Feature detection method and device

Abstract

The invention provides a feature detection method and a feature detection device. The traffic environment feature detection method comprises the following steps: acquiring an environment image in the advancing direction of a vehicle; and inputting the environment image into an environment feature extraction model to extract environment features in the environment image, wherein the environment feature extraction model adopts a combined network architecture of a cavity space convolution pooling pyramid and a lightweight network, and the environment features are objects or signs influencing a driving strategy of the vehicle.

Description

technical field [0001] The invention relates to the field of automatic driving, in particular to a feature detection method and device thereof. Background technique [0002] As the future research direction of vehicles, autonomous driving has a profound impact on the vehicle industry and even the transportation and express delivery industries. The advent of unmanned vehicles will liberate human hands, reduce the frequency of traffic accidents, and ensure people's safety. [0003] At the same time, with the continuous breakthrough and advancement of core technologies such as artificial intelligence and sensor detection, unmanned driving will become more intelligent, and at the same time, it will be able to realize the industrialization of unmanned vehicles. [0004] Lane detection is an important part of autonomous driving. It is one of the most important research topics driving scene understanding. Once the lane position is obtained, the vehicle will know where to go and ...

AI Technical Summary

Problems solved by technology

Lane line detection based on traditional methods has many shortcomings: it cannot achieve high detection robustness; it cannot perform fast lane detection; at the same time, it is necessary to modify the threshold parameters for different road materials; especially in curved and intersection roads, it is easy to miss Inspection, the detection effect is extremely poor, etc.

[0007] Due to the limited receptive field of convolutional neural network, lane line detection based on deep learning technology leads to the extraction of lane line features is basically local information, especially on curved roads, lane lines are prone to missed detection, and the detection speed is so slow that it is difficult to achieve real intelligent driving. requirements

Images