Cross-modal image matching method based on coupled convolution sparse coding
A convolutional sparse coding and matching method technology, applied in neural learning methods, character and pattern recognition, biological neural network models, etc., can solve problems such as low sparsity and inaccurate image features
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0085] Such as figure 1 , figure 2 , image 3 As shown, the present invention provides a cross-modal image matching method based on coupled convolutional sparse coding, comprising the following steps:
[0086] S1: Construct a cross-modal image matching algorithm model based on coupled convolutional sparse coding;
[0087] S2: Set the dimension and number of filters; set the number of samples of the two modal training sets and preprocess to obtain training set X and training set Y;
[0088] S3: Initialize the projection matrix T according to the dimension and number X , T Y ; Initialize the local dictionary D according to the training set X and training set Y XL ,D YL , combined with training set X, training set Y and local dictionary D XL ,D YL Initialize local sparse vector and Complete the parameter setting;
[0089] S4: Optimizing and updating the parameter D through continuous cross-iteration XL ,D YL , T X , T Y , when updating a pair of parameters, se...
Embodiment 2
[0155] More specifically, on the basis of Example 1, the process and function of the matching algorithm are illustrated by using the face sketch and face photo matching simulation to assist the police in tracking suspects, but the function of the algorithm is not limited to this. Specific embodiment flow chart sees Figure 4 .
[0156] The data set used in this embodiment adopts the public data set CUHK such as Figure 5 As shown, the data set has a total of 188 pairs of face pictures, including 188 face photos and 188 face portraits, of which 88 face photos and 88 face portraits are used as training sets, and the remaining 100 face photos And 100 face portraits as a test set. The face sketch map and face photo matching algorithm are listed as follows:
[0157]
[0158] In the specific implementation process, parameter setting and initial modeling are carried out. Among them, X is a training set of 88 face photos, Y is a training set of 88 face portraits, and each 88 pi...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More