Lane line detection method, device, terminal and storage medium based on artificial intelligence

Abstract

The present invention relates to the field of artificial intelligence technology, and provides an artificial intelligence-based lane line detection method, device, terminal, and storage medium, including: inputting historical lane line images marked with lane line areas into a neural network for iterative training; The weight parameter output by each neuron at the end of the iteration; use the generalized normalization model to generalize the weight parameter output by each neuron to obtain a new weight parameter and use it as the current iteration result, which is the weight Parameter, is the sum of the dimensions on each order, is the balance coefficient, and is the new weight parameter; perform the next round of iterative training according to the results of the current iteration until the end condition of the training is met to obtain the lane line detection model; use the lane line detection The model detects lane lines in the image of lane lines to be tested. The invention can be applied to the field of intelligent traffic, and improves the segmentation accuracy of lane lines in lane line images.

Description

technical field [0001] The present invention relates to the technical field of deep learning in artificial intelligence, in particular to an artificial intelligence-based lane line detection method, device, terminal and storage medium. Background technique [0002] Instance segmentation is a very important one in the field of artificial intelligence and in computer vision tasks. In the traffic field, due to its inconspicuous features, the lane line requires high position accuracy, and its detection is more dependent on the instance segmentation technology to complete. The existing lane line detection technology based on instance segmentation technology can achieve a test accuracy of over 96% on public datasets. However, the performance on real data is average, and the accuracy of lane line detection results is low and the robustness is poor. [0003] The reason is that the existing instance segmentation ignores the inherent relationship between the detection process and the...

AI Technical Summary

Problems solved by technology

Therefore, parameter redundancy is caused, which leads to unnecessary expansion of instance segmentation in space, which becomes a major reason for the low robustness of the model.

Images