Vehicle fine-grained identification method and device based on multiple attention mechanisms

Abstract

The embodiment of the invention provides a vehicle fine-grained identification method and device based on a multiple attention mechanism, and the method comprises the steps: carrying out the pre-training of a channel clustering layer and a classification network layer through an obtained vehicle image, and constructing a multiple attention convolution neural network model; determining a loss function of the multi-attention convolutional neural network model, and adjusting weight parameter matrixes and offset values in the channel clustering layer and the classification network layer by learning joint loss of the loss function to obtain adjustment values of the weight parameter matrixes and offset values of the channel clustering layer and the classification network layer; and training themulti-attention convolutional neural network model by using a vehicle data set containing fine-grained image classifications of different vehicle attributes to obtain determined values of weight parameter matrixes and offset values of the channel clustering layer and the classification network layer for vehicle feature extraction and vehicle multi-attribute identification. Through the embodiment of the invention, the accuracy of vehicle identification can be improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision and intelligent transportation, and in particular relates to a vehicle fine-grained recognition method, device, terminal equipment and computer-readable medium based on a multi-attention mechanism. Background technique [0002] With the rapid development of modern transportation, security and other industries, more and more target recognition technology has been applied in various fields, which is one of the important research topics of computer vision and pattern recognition technology in the field of intelligent transportation in recent years. [0003] Vehicle fine-grained recognition is an important research direction in the field of computer vision. Vehicle identification of the same model is more difficult for traditional methods, because the differences between similar vehicles are often very small, and the difference may only lie in the annual inspection mark on it, or some small de...

AI Technical Summary

Problems solved by technology

Aiming at the shortcomings of the existing algorithms, such as the low recognition rate and inaccurate recognition of the same type of vehicles, it is necessary to improve the existing vehicle image recognition method so that it can accurately detect and identify illegal vehicles (such as detecting whether the vehicle has an annual inspection label or not). , whether the driver is making a phone call and not wearing a seat belt, etc.), and realize the functions of intelligent retrieval of vehicles through multi-dimensional feature combinations such as tissue boxes, ornaments, pendants, etc.

Images