A method and device for target re-identification based on feature selection convolutional neural network
A convolutional neural network and feature selection technology, applied in the field of image recognition, can solve problems such as the inability to effectively identify similar targets, and achieve the effect of reducing interference and improving capabilities
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0048] This embodiment is a target re-identification method based on feature selection convolutional neural network, refer to figure 1 , including the following steps:
[0049] S1. Input the original image to be re-identified into the feature selection convolutional neural network;
[0050] S2. The feature selection convolutional neural network processes the original image, thereby extracting and outputting the feature vector of the original image;
[0051] S3. Re-identify the target according to the feature vector of the original image;
[0052] The feature selection convolutional neural network includes a plurality of convolutional layers, and each convolutional layer is used to process respective input values, thereby outputting a feature map group corresponding to the input value, and the feature map group includes multiple feature maps;
[0053] The feature selection convolutional neural network further includes at least one feature map selection layer, the feature map...
Embodiment 2
[0084] This embodiment is a method for training and testing the feature selection convolutional neural network described in Embodiment 1.
[0085] Further as a preferred embodiment, before step S1 is performed, it also includes the step of training the feature selection convolutional neural network, and the step of training the feature selection convolutional neural network specifically includes:
[0086] Use training images to train a feature selection convolutional neural network;
[0087] During the training process, the total number of iterations performed by the feature selection convolutional neural network is recorded, and the number of times each feature map is filtered out by the feature map selection layer is recorded.
[0088] Further as a preferred embodiment, after performing the step of training the feature selection convolutional neural network and before performing step S1, it also includes the step of testing the feature selection convolutional neural network,...
Embodiment 3
[0098] This embodiment provides a target re-identification system based on feature selection convolutional neural network, refer to figure 2 , the system of this embodiment includes a vehicle image acquisition module, a feature extraction module and a query matching module. The vehicle image acquisition module may be a surveillance camera, the feature extraction module is a feature selection convolutional neural network including a feature map selection layer, and the query matching module is used to match the feature vector output by the feature selection convolutional neural network and output the matching result.
[0099] Reference to the structure and principle of the feature selection convolutional neural network including the feature map selection layer in this embodiment image 3 and Table 1, image 3 In the feature selection convolutional neural network structure shown, the part pointed by the dotted circle is the feature map selection layer, and the part pointed by ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More