Image feature recognition method, related device and storage medium
A recognition method and image feature technology, applied in the electronic field, can solve the problems of not making full use of 3D, reducing the robustness of the algorithm, etc., and achieve the effect of accurate SSIP feature points, accurate positions, and abundant quantities
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0058] see Figure 1-a , the embodiment of the present application provides an image feature recognition method, including:
[0059] 101. Input the image to be processed into the feature extraction model, and extract the SSIP feature points of interest points in the hyperspectral space spectrum domain;
[0060] The image to be processed is input into the feature extraction model, and the hyperspectral space spectrum domain interest point SSIP feature point is extracted; the feature extraction model includes: a geometric algebra sub-model and a scale space processing sub-model; the geometric algebra sub-model is used for Obtain the geometric algebraic vector with spectral gradient information of the image to be processed in the hyperspectral space-spectral domain; the scale space processing sub-model realizes the corresponding Euclidean space operation through the geometric algebraic space operation of inner product and outer product.
[0061] Wherein, the image to be processed...
Embodiment 2
[0071] For ease of understanding, the embodiment of the present application introduces the geometric algebra sub-model and the scale-space processing sub-model in the feature extraction model in detail, including:
[0072] The settings of the geometric algebra submodel:
[0073] In order to better combine the geometric algebraic model with the hyperspectral image, and conduct an in-depth analysis of the spectral curve to propose a unified model of the spectral value and gradient change information of the hyperspectral image (HSI, Hyperspectral image) (that is, the embodiment of this application The geometric algebra submodel in ) firstly gives the general expression of the hyperspectral image in the geometric algebraic space; then, analyzes the spectral curve and its changing law of the hyperspectral image, and obtains the approximate changing law of the spectral curve of the same substance Almost unanimous conclusions, and then on the basis of the above, the concept of spectr...
Embodiment 3
[0160] The embodiment of the present application introduces the feature description model in detail, including:
[0161] In order to describe the extracted feature points, the local image information of the SSIP feature points is formed into a reference direction. For this purpose, a UMSGC-SIFT descriptor suitable for hyperspectral images is proposed in combination with the framework model of geometric algebra. Use the image gradient method to obtain the reference direction.
[0162] Assume and p 0 =x i e 1 +y j e 2 +λ k e 3 ,
[0163] p 1 =(x i -1)e 1 +y j e 2 +λ k e 3 ,p 2 =(x i +1)e 1 +y j e 2 +λ k e 3 ,
[0164] p 3 =x i e 1 +(y j -1)e 2 +λ k e 3 ,p 4 =x i e 1 +(y j +1)e 2 +λ k e 3 ,
[0165] p 5 =x i e 1 +y j e 2 +(λ k -1)e 3 ,p 6 =x i e 1 +y j e 2 +(λ k +1)e 3 , then p 0 The modulus value of is expressed as follows:
[0166]
[0167] For the convenience of expression, G(p 0 ,σ) means p 0 The Gaussian filter in t...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More