Fine-grained vehicle type recognition method based on weak surveillance localization and subclass similarity measurement

Abstract

The invention discloses a fine-grained vehicle type identification method based on weak supervision positioning and subclass similarity measurement, which comprises the following steps: 1) weak supervision positioning: using a pre-trained VGG-NET network to locate the image object, processing the mask map of convolution layer, and obtaining the boundary frame of the object. 2) constructing a fuzzysimilarity matrix: features of the pictures in the training set after the positioning of the picture are extracted by using B-CNN, and a fuzzy similarity matrix to measure the similarity of each subclass is obtained according to softmax classification results; 3) performing sampling to form a triple set: sampling to form a triple set on that basis of a fuzzy similarity matrix; 4) joint learning of the improved loss function: the improved triplet loss and the weighted softmax loss are used to restrict the distance between samples of the same sub-category and increase the distance between samples of different sub-categories. Compared with the original model, the invention is more accurate in positioning, and the classification accuracy is obviously improved, so that the vehicle target can be positioned well.

Description

technical field [0001] The invention relates to a fine-grained car model recognition method based on weakly supervised positioning and subcategory similarity measurement. Background technique [0002] Fine-grained model recognition is a sub-problem of fine-grained image classification. On the basis of traditional model recognition, which only recognizes the vehicle manufacturer, it is also necessary to distinguish different models of the same vehicle brand, such as Audi S5 and Audi S4. [0003] The purpose of fine-grained vehicle model recognition is to identify vehicle appearance images at any angle and in any scene to determine vehicle manufacturer, vehicle model and other information, which is of great significance in the fields of smart transportation, security, and automobile sales. In the problem of fine-grained vehicle identification, the differences between different vehicle categories are often very subtle, and the information with significant discrimination only ex...

AI Technical Summary

Problems solved by technology

Most of the current mainstream target detection and localization algorithms rely on additional labeling information, resulting in greatly reduced practicability

Although some weakly supervised target detection and positioning methods have appeared, they usually focus on the positioning of a certain local area and lose other local details. Therefore, it is necessary to accurately target the vehicle first.

[0008] 2. Another issue worthy of attention is that the subcategories are similar

Existing fine-grained car model recognition methods treat all subcategories equally, resulting in low discrimination between subcategories with high similarity. Therefore, it is necessary to measure the overall similarity between each category and expand Between-class variance and reduced intra-class variance

Images