Detection model optimization method for remote sensing image building, detection method and detection device
A technology for detecting models and remote sensing images, which is applied in the field of remote sensing image processing, can solve problems such as limited modeling ability and detection result errors, and achieve the effects of inhibiting learning, preventing overfitting, and strengthening learning
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0070] Such as figure 1 As shown, a method for optimizing a detection model of a remote sensing image building provided by an embodiment of the present invention includes the following steps:
[0071] Optimize the U-Net network model: replace the 3×3 standard convolution with an asymmetric convolution block in the feature extraction link of the U-Net network, and introduce an attention mechanism to adjust the feature weight in the step connection part of the U-Net network;
[0072] Make sample data;
[0073] Input the sample data into the modified U-Net network model for model training;
[0074] Calculate the accuracy of the model training results.
[0075] The U-Net network has gone through 5 encoding blocks and 10 times of 3×3 standard convolution operations in the feature extraction part. Repeated convolution operations will cause a large amount of information loss in the feature extraction part of the network, resulting in the model being prone to failure during training...
Embodiment 2
[0112] Such as Figure 4 As shown, a kind of building change detection method based on U-NET model that the embodiment of the present invention provides, comprises the following steps:
[0113] Preprocessing the remote sensing images to be detected;
[0114] Input the preprocessed remote sensing images to be detected into any of the above-mentioned optimized U-Net network models for building change detection.
[0115] As a possible implementation of this embodiment, the preprocessing of the remote sensing image to be detected includes:
[0116] Perform color balance and stretch processing on the remote sensing image to be detected;
[0117] Carry out secondary fine correction processing on the remote sensing image to be detected;
[0118] The remote sensing image to be detected is normalized, and the pixel value of the image is normalized to between 0 and 1.
[0119] As a possible implementation of this embodiment, the color difference equalization and stretch processing o...
Embodiment 3
[0125] Such as Figure 5 As shown, a U-NET model-based building change detection device provided in an embodiment of the present invention includes:
[0126] The image preprocessing module is used for preprocessing the remote sensing image to be detected;
[0127] The building change detection module is used to input the preprocessed remote sensing images to be detected into the optimized U-Net network model for building change detection.
[0128] As a possible implementation of this embodiment, the building change detection module includes:
[0129] The color difference equalization and stretching processing module is used for color difference equalization and stretching processing of the remote sensing image to be detected: the Canny edge detection algorithm is used to extract the band splicing boundary of the image, and the image is segmented according to the detected boundary; each time a Part is used as the region of interest, and the information of other parts is maske...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More