Picture feature extraction method and device, target re-identification method and device and electronic equipment

Abstract

The invention provides a picture feature extraction method and device, a target re-identification method and device and electronic equipment, and the method comprises the steps: inputting a current picture into a preset neural network model for feature fusion, and obtaining a first type of feature map outputted by a low convolution layer and a second type of feature map outputted by a high convolution layer of the neural network model; performing pooling operation on the first-class feature map and the second-class feature map to obtain a first-class feature sub-vector corresponding to the first-class feature map and a second-class feature sub-vector corresponding to the second-class feature map; performing vector splicing on the first type of feature sub-vectors and the second type of feature sub-vectors to obtain a feature splicing vector of the current picture; multiplying the feature splicing vector by a pre-stored attention vector point by point to obtain a feature vector corresponding to the current picture; wherein the attention vector is consistent with the feature splicing vector in length. According to the method and the device, the extracted picture features are more accurate, so that the accuracy of subsequent picture re-identification is improved.

Description

technical field [0001] The present application relates to the technical field of pedestrian re-identification, and in particular to a picture feature extraction method, object re-identification method, device and electronic equipment. Background technique [0002] Pedestrian re-identification (Person Re-Identification, ReID) technology is mainly used in the field of security, which is a technology that uses computer vision technology to determine whether a specific pedestrian exists in an image or video sequence. The current mainstream person re-identification technology generally adopts a two-stage scheme: the first stage uses labeled pedestrian pictures to train a deep classification model, and the second stage uses the trained classification model to extract features from the test pictures, and calculates the distance between pictures based on the features. distance to complete the retrieval task. The most commonly used classification model is the convolutional neural ne...

AI Technical Summary

Problems solved by technology

The most commonly used classification model is the convolutional neural network, and the existing technology mainly uses the last layer features of the network for classification learning. The potential problem is that the last layer features usually only contain the high-level abstract features of the image, while losing the underlying basic information , and for different test cases, the underlying features may be more important. Therefore, the existing person re-identification methods are not accurate enough to extract the features of the target object, which will affect the accuracy of the final retrieval results.

Images