CN-117173540-B - Anomaly detection method, anomaly detection system, anomaly detection equipment and anomaly detection storage medium for hyperspectral remote sensing image

CN117173540BCN 117173540 BCN117173540 BCN 117173540BCN-117173540-B

Abstract

The invention relates to the technical field of remote sensing image detection, and discloses an anomaly detection method, an anomaly detection system, anomaly detection equipment and an anomaly detection storage medium for a hyperspectral remote sensing image, wherein the hyperspectral remote sensing image to be detected is acquired; and inputting the hyperspectral remote sensing image to be detected into a pre-trained anomaly detection model for anomaly detection to obtain a detection result of the hyperspectral remote sensing image to be detected, wherein the anomaly detection model comprises a first convolution module, a second convolution module, a first transform module, a first Resize layer and a first full connection layer which are sequentially connected. According to the invention, the background and the abnormal sample are detected through clustering heuristic, so that the dependence of the network model on the manual labeling sample is effectively reduced, the abnormal information and the background information are distinguished for the training of the abnormal detection model, the training effect of the model is improved by utilizing the reconstructed image of the hyperspectral image, the whole model has local and global receptive fields, and the detection precision of the abnormal detection model is effectively improved.

Inventors

HE ZHI
ZHOU CHENGLE

Assignees

中山大学

Dates

Publication Date: 20260508
Application Date: 20230831

Claims (8)

1. An anomaly detection method for hyperspectral remote sensing images is characterized by comprising the following steps: acquiring a hyperspectral remote sensing image to be detected; Inputting the hyperspectral remote sensing image to be detected into a pre-trained anomaly detection model for anomaly detection to obtain a detection result of the hyperspectral remote sensing image to be detected, wherein the anomaly detection model comprises a first convolution module, a second convolution module, a first transform module, a first Resize layer and a first full connection layer which are sequentially connected; The training step of the anomaly detection model comprises the following steps: detecting a pseudo background sample and a pseudo abnormal sample of the hyperspectral remote sensing image set to generate a sample set; Constructing a decoder model symmetrical to the structure of the anomaly detection model; Inputting the sample set into the anomaly detection model for encoding to generate potential features, and inputting the potential features into the decoder model for decoding to generate reconstructed samples; Adjusting network parameters of the anomaly detection model according to a first loss function to obtain a trained anomaly detection model; the potential features are input into the decoder model for decoding, the step of generating the reconstructed sample comprises: adjusting network parameters of the decoder model according to a second loss function; Decoding the decoder model after the potential feature input is adjusted to generate a reconstructed sample; The first loss function is expressed using the following formula: where x (i) represents the hyperspectral sample of the ith input, Representing a reconstructed sample corresponding to the i-th input hyperspectral sample; the second loss function is expressed using the following formula: Where L (i) represents the i-th potential representation learned by the anomaly detection model, C represents the average of potential features learned from the pseudo-background sample by the anomaly detection model, and L (i) represents the label of the i-th sample.
2. The anomaly detection method of hyperspectral remote sensing images of claim 1, wherein the first convolution module comprises at least one first convolution unit, the first convolution unit comprising a first convolution layer and a batch normalization layer connected in sequence; The second convolution module comprises at least one second convolution unit, wherein the second convolution unit comprises a first sub-convolution module, a second sub-convolution module and a second convolution layer which are sequentially connected, and output data of the second convolution unit are input data of the second convolution unit and output data of the second convolution layer; the first transducer module includes at least one transducer unit.
3. The anomaly detection method for hyperspectral remote sensing images as claimed in claim 1, wherein the step of performing pseudo background sample and pseudo anomaly sample detection on the hyperspectral remote sensing image set, generating a sample set includes: performing cluster analysis on the hyperspectral remote sensing image set to obtain a specific class, wherein the specific class is a class with the maximum variance in a clustering result and the number of samples accords with a sample threshold value; calculating a first distance between each sample in the specific class and the class center, selecting a maximum distance from each first distance, dividing the first distance by the maximum distance, and obtaining a quotient corresponding to each sample; And classifying samples in the specific class according to the quotient to obtain a pseudo background sample set and a pseudo abnormal sample set.
4. The anomaly detection method of hyperspectral remote sensing images as claimed in claim 3 wherein the step of classifying samples in the specific class according to the quotient value to obtain a pseudo-background sample set and a pseudo-anomaly sample set comprises: Judging whether the quotient is smaller than a first threshold value, if so, taking a sample corresponding to the quotient as a pseudo background sample, otherwise, judging whether the quotient is larger than a second threshold value; if the sample is larger than a second threshold value, taking the sample corresponding to the quotient as a pseudo-abnormal sample; and generating a pseudo background sample set and a pseudo abnormal sample set according to the pseudo background sample and the pseudo abnormal sample.
5. The anomaly detection method of hyperspectral remote sensing images of claim 2, wherein the decoder model comprises a second full-connection layer, a second resolution layer, a second transducer module, a third convolution module, a fourth convolution module and an activation layer which are sequentially connected; the third convolution module has the same structure as the second convolution module, the fourth convolution module has the same structure as the first convolution module, and the second convolution module has the same structure as the first convolution module.
6. An anomaly detection system for hyperspectral remote sensing images, the system being applied to the method as claimed in any one of claims 1 to 5, comprising: The image acquisition module is used for acquiring hyperspectral remote sensing images to be detected; the image detection module is used for inputting the hyperspectral remote sensing image to be detected into a pre-trained anomaly detection model to perform anomaly detection, so as to obtain a detection result of the hyperspectral remote sensing image to be detected, and the anomaly detection model comprises a first convolution module, a second convolution module, a first transform module, a first Resize layer and a first full connection layer which are sequentially connected.
7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 5 when the computer program is executed by the processor.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.

Description

Anomaly detection method, anomaly detection system, anomaly detection equipment and anomaly detection storage medium for hyperspectral remote sensing image Technical Field The invention relates to the technical field of remote sensing image detection, in particular to an anomaly detection method, an anomaly detection system, anomaly detection equipment and an anomaly detection storage medium for a hyperspectral remote sensing image. Background The anomaly detection of the remote sensing image refers to extracting discrete points from the multispectral image to indicate that the pixels have anomaly values, and each pixel in the hyperspectral image contains spectral information or spectral characteristics of specific feature attributes, and different material components have different spectral information. Thus, in general, regions containing the same species have very strong similarities in their spectra, while regions containing different features have spectra that are typically combinations of different distributions that are statistically independent locally. Therefore, the hyperspectral remote sensing image anomaly detection plays a very important role in remote sensing image processing and other applications, such as being applied to various fields of mineral crystal detection, petroleum and natural gas exploration and the like. Along with the rapid development of artificial intelligence, the deep learning-based method is increasingly applied to the field of anomaly detection of the remote sensing images. The problem of the existing anomaly detection model based on deep learning is that the excellent effect of model detection is required to be at the cost of a large amount of training data, but in anomaly detection, an anomaly labeling sample is often insufficient, so that background information cannot be effectively restrained and the anomaly information cannot be enhanced, and a single anomaly detection model is difficult to extract local and global distinguishing features of anomaly and background information in a hyperspectral image at the same time, so that the accuracy of an anomaly detection result is influenced to a certain extent. Disclosure of Invention In order to solve the technical problems, the invention provides an anomaly detection method, an anomaly detection system, anomaly detection equipment and an anomaly detection storage medium for hyperspectral remote sensing images, which can solve the problems that the annotation data is insufficient and local and global distinguishing features of anomalies and background information in hyperspectral images are difficult to extract simultaneously, and achieve the technical effects of reducing dependence on manual annotation samples and improving anomaly detection accuracy. In a first aspect, the present invention provides a method for detecting anomalies in hyperspectral remote sensing images, the method comprising: acquiring a hyperspectral remote sensing image to be detected; And inputting the hyperspectral remote sensing image to be detected into a pre-trained anomaly detection model for anomaly detection to obtain a detection result of the hyperspectral remote sensing image to be detected, wherein the anomaly detection model comprises a first convolution module, a second convolution module, a first transform module, a first Resize layer and a first full connection layer which are sequentially connected. Further, the first convolution module comprises at least one first convolution unit, and the first convolution unit comprises a first convolution layer and a batch normalization layer which are sequentially connected; The second convolution module comprises at least one second convolution unit, wherein the second convolution unit comprises a first sub-convolution module, a second sub-convolution module and a second convolution layer which are sequentially connected, and output data of the second convolution unit are input data of the second convolution unit and output data of the second convolution layer; the first transducer module includes at least one transducer unit. Further, the training step of the anomaly detection model includes: detecting a pseudo background sample and a pseudo abnormal sample of the hyperspectral remote sensing image set to generate a sample set; Constructing a decoder model symmetrical to the structure of the anomaly detection model; Inputting the sample set into the anomaly detection model for encoding to generate potential features, and inputting the potential features into the decoder model for decoding to generate reconstructed samples; and adjusting network parameters of the anomaly detection model according to the first loss function to obtain a trained anomaly detection model. Further, the step of detecting the pseudo background sample and the pseudo abnormal sample of the hyperspectral remote sensing image set and generating the sample set includes: performing cluster analysis on the hyperspectral remote s