Deep learning image recognition method, system and terminal equipment for intelligent driving

Abstract

The present invention provides a deep learning image detection and recognition method, system and terminal equipment for intelligent driving. The system includes: a shared convolution network, a region segmentation network, and a target recognition network; the region segmentation network is based on the shared convolution network to extract The feature map of the target recognition network is used to perform region classification processing based on the feature map extracted by the shared convolutional network; the region segmentation result obtained by the region segmentation network and the target recognition result obtained by the target recognition network are used. Supervising the shared convolutional network to complete the shared learning of the region segmentation network and the target recognition network. The present invention has an obvious speed advantage in multi-task learning. Compared with learning two independent networks alone, it has the characteristics of less time-consuming and high efficiency; in addition, the present invention also effectively avoids the problem of repeated operations of the convolutional layer. Can complete multi-task detection and recognition.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to an image detection and recognition method, in particular to a deep learning image recognition method, system and terminal equipment for intelligent driving. Background technique [0002] In the field of intelligent vehicles, there are mainly two existing methods for target positioning and recognition, one is based on traditional neural network training classifiers; the other is based on artificially extracted grayscale image features input into multi-class support vector machines implementation of the recognition algorithm. The traditional neural network classifier needs to train two networks for the two tasks of detection and recognition, and the calculation cost is high; while the artificial feature extraction method has great limitations, and the selection of decision parameters is a heuristic method. If the value is not selected properly, it will directly affect the recognition...

AI Technical Summary

Problems solved by technology

[0003] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a deep learning image recognition method, system and terminal equipment for intelligent driving, which are used to solve the high computational cost and feature extraction limitations of existing target recognition methods Large, or / and parameter selection is difficult to achieve the optimal value, affecting the recognition results

Images