Vehicle re-identification method based on channel coordinating attention

Abstract

The invention relates to a vehicle re-identification method based on channel collaborative attention, The method comprises the following steps: constructing channel joint features, calculating channel covariance vectors, constructing an identification region detector, carrying out identification region regularization based on weight density, and establishing a deep learning feature extraction model; and inputting a picture into the model to extract a re-identification feature vector, calculating a cosine distance between the re-identification feature vector of the query image and the re-identification feature vector of the candidate image, and sorting the feature distances to obtain a final re-identification matching result. Through the method, the problem of mismatching caused by low intra-class similarity and high inter-class similarity in vehicle re-identification can be effectively solved, additional local labeling information is not needed, deployment and application are easier, and the method has positive significance in the aspects of smart city construction and intelligent security management.

Description

Technical field [0001] The present invention belongs to the field of computer technology, particularly a vehicle re-identification method in accordance with the channel synergistic. Background technique [0002] Vehicles rereading is a challenging and meaningful problem in intelligent transportation tasks. Most of the current vehicle re-identification system is based on consolonine neural network technology construction, basic convolutive neural network technology, despite better completion of vehicle appearance characteristics. Extraction, but there is no higher re-identification accuracy. The main reason for the recognition error is as follows: Due to too many vehicles similar to the appearance, the overall feature cannot be effectively distinguished, and the detail characteristics that can be used to identify are too high, and the network is difficult to give too high attention; The appearance characteristic changes caused by the change of the gesture are too obvious, so the h...

AI Technical Summary

Problems solved by technology

Usually, people detect the target local area by means of adding a detection model. This process requires a large amount of local area labeling information, and the detection model needs to be added to the original model, which increases the parameter scale and calculation amount of the model.

Images