Remote sensing scene image classification method based on channel multi-group fusion
A scene image and classification method technology, applied in the field of remote sensing scene image classification, can solve the problems of single grouping form of channel grouping and low extraction accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
specific Embodiment approach 1
[0049] Specific implementation mode 1: The specific process of the remote sensing scene image classification method based on channel multi-group fusion in this implementation mode is as follows:
[0050] Step 1, acquiring hyperspectral images (labeled);
[0051] Step 2. Establishing a lightweight convolution neural network based on channel multigroup fusion (LCNN-CMGF) model;
[0052] Step 3: Input the hyperspectral image (labeled) into the established lightweight convolution neural network based on channel multigroup fusion (LCNN-CMGF) model for training, and obtain the trained A lightweight convolution neural network based on channel multigroupfusion (LCNN-CMGF) model based on channel multigroup fusion;
[0053] The lightweight convolutional neural network model based on channel multi-packet fusion includes input layer, first group (Group1), second group (Group2), third group (Group3), fourth group (Group4), fifth group Group (Group5), sixth group (Group6), seventh group (...
specific Embodiment approach 2
[0083] Specific embodiment two: the difference between this embodiment and specific embodiment one is that the
[0084] The first three groups are used to extract shallow information of remote sensing images;
[0085] Among them, the first and second groups adopt the proposed three-branch shallow downsampling structure;
[0086] The first group and the second group respectively include branch 1, branch 2 and branch 3;
[0087] Branch 1 sequentially includes the convolution layer with the first step size of 2 convolution kernel size 3×3, batch normalization, Rule activation function, the second step size of 1 convolution layer with convolution kernel size of 3×3, batch Standardization, Rule activation function;
[0088] Branch 2 sequentially includes a maximum pooling layer with a step size of 2 and a pooling kernel size of 2, and a third convolution layer with a step size of 1 and a convolution kernel size of 3×3;
[0089] Branch 3 includes a convolutional layer with a four...
specific Embodiment approach 3
[0111] Specific implementation mode 3: The difference between this implementation mode and specific implementation mode 1 or 2 is that the third group sequentially includes a convolution layer with a fifth step size of 1 and a convolution kernel size of 1×1, and a sixth step size of 1 Convolution layer with a convolution kernel size of 3×3 and a depth-separable convolution layer with a convolution kernel size of 2×3 in the first step; the specific process is:
[0112] The third group uses a hybrid convolution method combining standard convolution and depthwise separable convolution to extract features.
[0113] Compared with standard convolution, depthwise separable convolution has a large reduction in the number of parameters.
[0114] Suppose the input feature size is H×W×C 1 , the convolution kernel size is H 1 ×W 1 ×C 1 , the output feature size is H×W×C 2 , then the parameter volume of the standard convolutional layer convolution is:
[0115] params conv =(H 1 ×W ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


