Vehicle brand model fine identification method and system based on depth learning

Abstract

The invention discloses a vehicle brand model fine identification method and system based on depth learning. The vehicle brand model fine identification method based on depth learning includes the steps: acquiring an original vehicle image; performing spatial pyramid dividing on the original vehicle image to divide the original vehicle image into three layers and 21 image blocks, wherein the number of the image blocks in the three layers is respectively 1, 4 and 16; utilizing an improved convolution neural network to perform characteristic extraction on each divided image block to obtain the characteristic vector of each image block, wherein the improved convolution neural network includes a convolution layer, a maximum pooling layer, a configuration layer and an average pooling layer; according to the characteristic vector of each image block, utilizing the method of weight space pyramid to obtain the final expression vector of the vehicle image; and sending the final expression vector of the vehicle image to a pre-trained multi-class linear support vector machine classifier to perform vehicle brand model identification. The vehicle brand model fine identification method and system based on depth learning have the advantages of high robustness and high identification accuracy, and can be widely applied to the image processing field.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method and system for finely identifying vehicle brand models based on deep learning. Background technique [0002] In recent years, with the gradual improvement of road monitoring and security deployment, the video image data that traffic management departments need to process every day has also increased sharply. Much attention. The main work of fine image-based vehicle brand and model recognition is to identify unknown vehicle types, brands, models, years, etc. from an image. Different from the traditional vehicle type identification, there are two difficulties in the fine identification of vehicle brand models: the first is that the appearance of the vehicles has a high similarity, especially the vehicles of different years of the same brand; the second is due to the Even the same vehicle may show different image characteristics due to the influence of lighting, weather, a...

AI Technical Summary

Problems solved by technology

The advantage of the recognition algorithm based on deep learning is that the image features extracted adaptively can well describe the internal information of the vehicle image, and the generalization ability is good, but this method does not consider the spatial structure information of the vehicle, which affects the subsequent recognition accuracy. cause a certain impact

Images