Vehicle model recognition model construction method based on depth learning and vehicle model recognition method based on depth learning

Abstract

The invention provides a vehicle model recognition model construction method based on depth learning. According to the method, a deep convolution neural network structure is designed for vehicle model recognition; softmax classification loss and sorting loss which is built through a vehicle model type hierarchical structure and is based on tetrad are used together to supervise the deep convolution neural network training; and intra-class differences and inter-class differences of vehicle models are constrained at the same time to learn character representation with rich semantic information and a discriminating classifier. In order to speed up network convergence, an online difficult sample mining policy and an improved gradient backtracking optimization algorithm are used, which greatly reduces the network training time. The invention further provides a vehicle model recognition method based on depth learning. The vehicle model recognition method uses the model constructed by the model construction method to recognize the vehicle type in a vehicle picture, which effectively improves the vehicle model classification accuracy.

Description

technical field [0001] The invention belongs to the field of computer vision and intelligent transportation, and in particular relates to a method for constructing a vehicle identification model based on deep learning and a vehicle identification method. Background technique [0002] Vehicle type recognition belongs to the problem of fine-grained classification in computer vision, and it is also an important research direction of intelligent traffic monitoring system. It needs to realize the automatic analysis and identification of the vehicles in the monitoring screen, which are specific to fine categories such as vehicle manufacturers and car series. Therefore, the feature representation of vehicles needs to have strong expressive ability and discrimination. However, traditional hand-designed features based on global or local images are difficult to achieve better results on the issue of vehicle recognition. [0003] Deep learning is a popular research direction in the fi...

AI Technical Summary

Problems solved by technology

[0004] The existing deep learning-based vehicle recognition technology still has the following deficiencies: First, the deep neural networks they design or adopt are still relatively shallow, and generally speaking, when there are enough training data, the discriminative ability of deeper neural networks And the generalization ability is stronger, and it can distinguish more subtle differences between different models; secondly, the network training supervision function directly uses the softmax classification loss. Although it can effectively increase the inter-class differences between different models, it lacks the same model. The restraint effect of the difference within the class, resulting in the pictures of vehicles with large differences within the class are likely to be classified into adjacent vehicle categories; finally, the categories of vehicle types actually have a hierarchical structure, for example, two vehicles may belong to off-road vehicles (SUV), But belong to different vehicle manufacturers, or belong to the same vehicle manufacturer but not the same car series

However, the above-mentioned deep learning-based model recognition technology does not take into account the hierarchical structure of the model category. It may be easy to distinguish coarse-grained model categories (such as off-road vehicles and sedans), but it is difficult to distinguish fine-grained model categories ( Such as BMW 5 Series 2002 and BMW 5 Series 2008), the discrimination is still insufficient

Images