Forest information acquisition system based on hyperspectral image features

Abstract

The invention discloses a forest information acquisition system based on hyperspectral image features, and relates to the technical field of hyperspectral image.An image acquisition unit is in electrical output connection with an image feature parameter extraction module, and the image feature parameter extraction module is in electrical output connection with a feature fusion matrix construction unit; the feature fusion matrix construction unit is in electrical output connection with the original hyperspectral image unit; a multi-scale dense connection and feature aggregation network is constructed through cooperation of a multi-scale feature extraction module, a spectral spatial feature extraction module, a multi-scale feature extraction module and a multi-level feature aggregation module, the multi-scale dense connection and feature aggregation network is trained by using a generated training set, and hyperspectral images are classified. Moreover, the spectral spatial features of the hyperspectral image can be captured more comprehensively, and then the spatial features of the hyperspectral image and different hierarchical features are fully utilized, so that the problems of insufficient spectral spatial feature representation, unbalanced samples and the like in a hyperspectral image classification method are effectively solved.

Description

Technical Field

[0001] The invention relates to the technical field of hyperspectral images, and in particular to a forest information collection system based on hyperspectral image features. Background Art

[0002] Forest resource information is the main basis for forestry planning and decision-making. It can be divided into three categories: the first category is the information of the forest itself. It includes the quantity and quality of forest resources at all levels of management units, the spatial configuration of forest types and local types, the distribution, growth, and loss of forest age and diameter classes, etc.; the second category is the external information of the forest. It includes forest rights, social economy, natural conditions, engineering equipment, etc.; the third category is forestry production information, including forestry-related roads, engineering projects, and various plans, reports, costs, and implementation effects of seedling cultivation, afforestation, forest management, logging, and processing; The main features of hyperspectral images include high spectral resolution, rich spectral information, and the ability to obtain multi-dimensional information. Hyperspectral imaging technology achieves higher spectral resolution than multispectral images by dividing light into finer bands. This technology can divide the electromagnetic spectrum into dozens to hundreds of continuous and subdivided spectral bands, and simultaneously image the target area to obtain rich spectral information. Each pixel of a hyperspectral image contains the reflectance values ​​of multiple imaging bands, forming a continuous spectral curve. This feature enables hyperspectral images to reflect the changing relationship between the spectral response of the object and the wavelength, and provides sensitivity to differences in the physical structure and chemical composition of the object.

[0003] In addition, hyperspectral image technology combines imaging technology and spectral technology to achieve the simultaneous acquisition of the target's two-dimensional geometric spatial information and spectral dimension information, forming the so-called "image cube" structure and form. This structure not only provides the spatial position information of the object, but also reflects the changing relationship between the spectral response of the object and the wavelength through the spectral feature space, thereby providing a richer mathematical classification model and the possibility of processing, analyzing, and extracting data. However, the existing hyperspectral image system used for forest information collection system still has some shortcomings, such as insufficient spatial feature representation, sample imbalance, and a large amount of redundant information. At the same time, the extraction and processing of feature information is not thorough enough, resulting in poor use effect. We propose a forest information collection system based on hyperspectral image features. Summary of the invention

[0004] The purpose of the present invention is to provide a forest information collection system based on hyperspectral image features to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A forest information collection system based on hyperspectral image features, comprising an input receiving module, the input receiving module is electrically connected to the processing module through an output, the processing module is electrically connected to the hyperspectral image collection module through an output, the hyperspectral image collection module is electrically connected to the image collection unit through an output, the image collection unit is electrically connected to the image feature parameter extraction module through an output, the image feature parameter extraction module is electrically connected to the feature fusion matrix construction unit through an output, the feature fusion matrix construction unit is electrically connected to the original hyperspectral image unit, one end of the original hyperspectral image unit is electrically connected to the spectral feature extraction module through an output, the spectral feature extraction module is electrically connected to the discrete cosine transform unit through an output, the discrete cosine transform unit is electrically connected to the independent component analysis unit, the other end of the original hyperspectral image unit is electrically connected to the spatial feature extraction module through an output, and the spatial feature extraction module is electrically connected to the spatial feature extraction module through an output. The extraction module is electrically connected to the superpixel segmentation unit, the superpixel segmentation unit is electrically connected to the principal component analysis unit, the principal component analysis unit and the independent component analysis unit are both electrically connected to the image extraction module, the image extraction module is electrically connected to the stacked feature unit, the stacked feature unit is electrically connected to the multi-feature fusion modeling module, the multi-feature fusion modeling module is electrically connected to the processor module, the processor module is electrically connected to the spectral spatial feature extraction module, the spectral spatial feature extraction module is electrically connected to the multi-scale feature extraction module, the multi-scale feature extraction module is electrically connected to the multi-level feature aggregation module, the multi-level feature aggregation module is electrically connected to the evaluation prediction model, the prediction model is electrically connected to the transmission module, and the transmission module is electrically connected to the display module.

[0006] Further: the hyperspectral image acquisition module includes an image acquisition unit, the image acquisition unit is electrically connected to the image classification unit, and the image classification unit is electrically connected to the data analysis unit.

[0007] Further: the image feature parameter extraction module includes an image preprocessing module, the image preprocessing module electrically output is connected to the spectral feature extraction and analysis module, the spectral feature extraction and analysis module electrically output is connected to the feature extraction model module, the feature extraction model module electrically output is connected to the cloud model module, and the cloud model module electrically output is connected to the prediction and response module.

[0008] Further: the original hyperspectral image unit includes a data acquisition module, the data acquisition module is electrically connected to the data analysis module, the data analysis module is electrically connected to the classification module, and the classification module is electrically connected to the data storage module.

[0009] Further: the spectral feature extraction module includes a spectral preprocessing module, the spectral preprocessing module electrically output is connected to the spectral feature extraction module, the spectral feature extraction module electrically output is connected to the spatiotemporal variation law analysis module, and the spatiotemporal variation law analysis module electrically output is connected to the similarity analysis module.

[0010] Further: the data analysis unit includes a component analysis module, the component analysis module is electrically connected to the pixel purity analysis module, and the pixel purity analysis module is electrically connected to the minimum noise separation module.

[0011] Further: the superpixel segmentation unit includes a target image determination unit, the target image determination unit is electrically connected to the segmentation unit, the segmentation unit is electrically connected to the labeling unit, and the labeling unit is electrically connected to the training sample determination unit.

[0012] Further: the image extraction module includes an image preprocessing module, the image preprocessing module is electrically connected to the color feature extraction module, the color feature extraction module is electrically connected to the texture feature extraction module, the texture feature extraction module is electrically connected to the shape feature extraction module, the shape feature extraction module is electrically connected to the spatial relationship feature extraction module, and the spatial relationship feature extraction module is electrically connected to the feature comparison module.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The forest information collection system based on hyperspectral image features constructs a spectral spatial feature extraction module, a multi-scale feature extraction module and a multi-level feature aggregation module to construct a multi-scale dense connection and feature aggregation network. The generated training set is used to train the multi-scale dense connection and feature aggregation network to classify the hyperspectral image, and the spectral spatial features of the hyperspectral image can be captured more comprehensively. Then, the hyperspectral image spatial features and different levels of features are fully utilized to effectively solve the problems of insufficient spectral spatial feature representation and sample imbalance in the hyperspectral image classification method. At the same time, an image feature parameter extraction module, a spectral feature extraction and analysis module, a feature fusion matrix construction unit, a spectral feature extraction module, a stacked feature unit, and a multi-feature fusion modeling module are set to comprehensively and effectively process the spectral image data features. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the system of the present invention; Figure 2 This is a block diagram of the hyperspectral image acquisition module subsystem of the present invention; Figure 3 This is a block diagram of the image feature parameter extraction module subsystem of the present invention; Figure 4 This is a block diagram of the original hyperspectral image unit subsystem of the present invention.

[0015] In the figure: 1. Input receiving module; 2. Processing module; 3. Hyperspectral image acquisition module; 301. Image acquisition unit; 302. Image classification unit; 303. Data analysis unit; 4. Image acquisition unit; 5. Image feature parameter extraction module; 501. Image preprocessing module; 502. Spectral feature extraction and analysis module; 503. Feature extraction model module; 504. Cloud model module; 505. Prediction and response module; 6. Feature fusion matrix construction unit; 7. Original hyperspectral image unit; 701. Data acquisition module; 702. Data analysis module; 703. Classification module; 704, data storage module; 8, spectral feature extraction module; 9, discrete cosine transform unit; 10, independent component analysis unit; 11, spatial feature extraction module; 12, superpixel segmentation unit; 13, principal component analysis unit; 14, image extraction module; 15, stacked feature unit; 16, multi-feature fusion modeling module; 17, processor module; 18, build spectral spatial feature extraction module; 19, build multi-scale feature extraction module; 20, build multi-level feature aggregation module; 21, evaluation and prediction model; 22, transmission module; 23, display module. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Embodiment 1: See also Figure 1-4The present invention provides a technical solution: a forest information collection system based on hyperspectral image features, comprising an input receiving module 1, the input receiving module 1 is used to receive the hyperspectral image data of the forest, the input receiving module 1 is electrically connected to the processing module 2, the processing module 2 can pre-process the hyperspectral image data, the processing module 2 is electrically connected to the hyperspectral image acquisition module 3, the hyperspectral image acquisition module 3 is electrically connected to the image acquisition unit 4, the image acquisition unit 4 is used to collect hyperspectral image data information, the image acquisition unit 4 is electrically connected to the image feature parameter extraction module 5, the image feature parameter extraction module 5 is used to extract image feature data, the image feature parameter extraction module 5 is electrically connected to the feature fusion matrix construction unit 6, the feature fusion matrix construction unit 6 can establish an estimation model, the feature fusion matrix construction unit 6 is electrically connected to the original hyperspectral image unit 7, one end of the original hyperspectral image unit 7 is electrically connected to the spectral feature extraction module 8, the spectral feature The extraction module 8 is electrically connected to the discrete cosine transform unit 9, which is a transform related to the Fourier transform and is often used in signal processing and image processing for lossy data compression of signals and images (including still images and moving images). The discrete cosine transform unit 9 is electrically connected to the independent component analysis unit 10. Independent component analysis is a principal component decomposition method. Its basic idea is to separate independent signals from a set of mixed observation signals, where the observation signal is a linear combination of several statistically independent components. Independent component analysis is an unmixing process, assuming that the data is a linear mixture of a set of independent sources, and these signal sources can be decomposed based on the statistical independence of these signal sources. The other end of the original hyperspectral image unit 7 is electrically connected to the spatial feature extraction module 11. The spatial feature extraction module 11 is electrically connected to the super pixel segmentation unit 12. The super pixel segmentation unit 12 subdivides the digital image into multiple image sub-regions (a collection of pixels) (also called super pixels). In the process, a superpixel is a small area composed of a series of pixel points with adjacent positions and similar features such as color, brightness, and texture. Most of these small areas retain effective information for further image segmentation and generally do not destroy the boundary information of objects in the image. The superpixel segmentation unit 12 is electrically connected to the main component analysis unit 13 through its output. The main component analysis unit 13 and the independent component analysis unit 10 are both electrically connected to the image extraction module 14 through their outputs. The image extraction module 14 is electrically connected to the stacked feature unit 15 through its output. The stacked feature unit 15 is a way of feature fusion. Feature fusion can usually be achieved by a simple stacking method (serial method) or a feature superposition method (parallel method).The serial method will greatly increase the data dimension and computational complexity, and will also cause the classification process to further fall into an ill-posed problem. The feature stacking of the present invention refers to the superposition of layers. The stacked feature unit 15 is electrically output connected to the multi-feature fusion modeling module 16, and the multi-feature fusion modeling module 16 is electrically output connected to the processor module 17. The processor module 17 is electrically output connected to build a spectral space feature extraction module 18, and the spectral space feature extraction module 18 is electrically output connected to build a multi-scale feature extraction module 19. The multi-scale feature extraction module 19 is electrically output connected to build a multi-level feature aggregation module 20, and a multi-scale dense connection and feature aggregation network is constructed. The multi-scale dense connection and feature aggregation network is trained using the generated training set to classify the hyperspectral image, and the multi-level feature aggregation module 20 is electrically output connected to the evaluation prediction model 21. The prediction model 21 is electrically output connected to the transmission module 22. The transmission module 22 is used to transmit all data. The transmission module 22 is electrically output connected to the display module 23, and the display module 23 is used to display all data.

[0018] Among them, preferably, the hyperspectral image acquisition module 3 includes an image acquisition unit 301, the image acquisition unit 301 is used to acquire the captured forest hyperspectral image data, the image acquisition unit 301 is electrically output connected to the image classification unit 302, the image classification unit 302 is used to classify the captured forest hyperspectral image data, the image classification unit 302 is electrically output connected to the data analysis unit 303, and the analysis unit 303 is used to analyze and process the forest hyperspectral image data.

[0019] Preferably, the image feature parameter extraction module 5 includes an image preprocessing module 501, the image preprocessing module 501 is electrically connected to the spectral feature extraction and analysis module 502 through an output, the spectral feature extraction and analysis module 502 is electrically connected to the feature extraction model module 503 through an output, the feature extraction model module 503 utilizes an algorithm model to extract image data features, the feature extraction model module 503 is electrically connected to the cloud model module 504 through an output, the cloud model module 504 is electrically connected to the prediction and response module 505 through an output, the prediction and response module 505 predicts alarm results based on data features using an algorithm model, and sends the data and results to the cloud model module 504, the cloud model module 504 receives data sent by the feature extraction model module 503 and the prediction and response module 505, updates the model based on the latest data, and sends the updated parameters to the feature extraction model module 503 and the prediction and response module 505.

[0020] Preferably, the original hyperspectral image unit 7 includes a data acquisition module 701, which is used to acquire original hyperspectral image data. The data acquisition module 701 is electrically connected to a data analysis module 702, which analyzes and processes the original hyperspectral image. The data analysis module 702 is electrically connected to a classification module 703, which is used to classify the data. The classification module 703 is electrically connected to a data storage module 704, which is used to store data information.

[0021] Embodiment 2: Reference Figure 1 , this embodiment is different from the first embodiment in that: the spectral feature extraction module 8 includes a spectral preprocessing module, the spectral preprocessing module is used to preprocess spectral data, the spectral preprocessing module is electrically connected to the spectral feature extraction module, the spectral feature extraction module is used to extract spectral features, the spectral feature extraction module is electrically connected to the spatiotemporal change law analysis module, the spatiotemporal change law analysis module is electrically connected to the similarity analysis module, the similarity analysis module grades the similarity between the target and the background according to the size of the weighted difference value, and outputs the similarity analysis result; the data analysis unit 303 includes a component analysis module, the component analysis module is used to analyze the main components in the image, the component analysis module is electrically connected to the pixel purity analysis module, the pixel purity analysis module is used to analyze the pixel purity, the pixel purity analysis module is electrically connected to the minimum noise separation module, and the minimum noise separation module is used to separate the minimum noise; the superpixel segmentation unit 12 includes a target image determination unit, the target image determination unit is used to determine the target image to be labeled, the target image determination unit is electrically connected to the segmentation unit The segmentation unit is used to segment the target image into multiple super-pixel blocks through a super-pixel segmentation algorithm. The segmentation unit is electrically output connected to a labeling unit. The labeling unit labels the labeling objects in the target image based on the divided super-pixel blocks. The labeling unit is electrically output connected to a training sample determination unit. The training sample determination unit determines the training samples according to the labeled target image; the image extraction module 14 includes an image preprocessing module. The image preprocessing module is electrically output connected to a color feature extraction module. The color feature extraction module is used to extract color features in the image. The color feature extraction module is electrically output connected to a texture feature extraction module. The texture feature extraction module is used to extract texture features in the image. The texture feature extraction module is electrically output connected to a shape feature extraction module. The shape feature extraction module is used to extract shape features in the image. The shape feature extraction module is electrically output connected to a spatial relationship feature extraction module. The spatial relationship feature extraction module is used to extract spatial relationship features of the image. The spatial relationship feature extraction module is electrically output connected to a feature comparison module. The feature comparison module compares all features.

[0022] The electrical appliances mentioned in this article are all connected to an external power source through wires.

[0023] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A forest information collection system based on hyperspectral image features, comprising an input receiving module (1), characterized in that: The input receiving module (1) is electrically connected to the processing module (2); the processing module (2) is electrically connected to the hyperspectral image acquisition module (3); the hyperspectral image acquisition module (3) is electrically connected to the image acquisition unit (4); the image acquisition unit (4) is electrically connected to the image feature parameter extraction module (5); the image feature parameter extraction module (5) is electrically connected to the feature fusion matrix construction unit (6); the feature fusion matrix construction unit (6) is electrically connected to the original hyperspectral image unit (7); one end of the original hyperspectral image unit (7) is electrically connected to the spectral feature extraction module (8); the spectral feature extraction module (8) is electrically connected to the discrete cosine transform unit (9); the discrete cosine transform unit (9) is electrically connected to the independent component analysis unit (10); the other end of the original hyperspectral image unit (7) is electrically connected to the spatial feature extraction module (11); the spatial feature extraction module (11) is electrically connected to the super pixel segmentation unit (12); the super pixel segmentation unit The element (12) is electrically connected to the main component analysis unit (13), the main component analysis unit (13) and the independent component analysis unit (10) are both electrically connected to the image extraction module (14), the image extraction module (14) is electrically connected to the stacked feature unit (15), the stacked feature unit (15) is electrically connected to the multi-feature fusion modeling module (16), the multi-feature fusion modeling module (16) is electrically connected to the processor module (17), the processor module (17) is electrically connected to the constructed spectral space feature extraction module (18), the constructed spectral space feature extraction module (18) is electrically connected to the constructed multi-scale feature extraction module (19), the constructed multi-scale feature extraction module (19) is electrically connected to the constructed multi-level feature aggregation module (20), the constructed multi-level feature aggregation module (20) is electrically connected to the evaluation prediction model (21), the prediction model (21) is electrically connected to the transmission module (22), and the transmission module (22) is electrically connected to the display module (23).

2. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The hyperspectral image acquisition module (3) comprises an image acquisition unit (301), the image acquisition unit (301) having an electrical output connected to an image classification unit (302), and the image classification unit (302) having an electrical output connected to a data analysis unit (303).

3. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The image feature parameter extraction module (5) comprises an image preprocessing module (501), the image preprocessing module (501) electrically outputs a spectral feature extraction and analysis module (502), the spectral feature extraction and analysis module (502) electrically outputs a feature extraction model module (503), the feature extraction model module (503) electrically outputs a cloud model module (504), and the cloud model module (504) electrically outputs a prediction and response module (505).

4. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The original hyperspectral image unit (7) comprises a data acquisition module (701), the data acquisition module (701) electrically outputs and is connected to a data analysis module (702), the data analysis module (702) electrically outputs and is connected to a classification module (703), and the classification module (703) electrically outputs and is connected to a data storage module (704).

5. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The spectral feature extraction module (8) comprises a spectral preprocessing module, the spectral preprocessing module is electrically connected to the spectral feature extraction module, the spectral feature extraction module is electrically connected to the spatiotemporal variation law analysis module, and the spatiotemporal variation law analysis module is electrically connected to the similarity analysis module.

6. The forest information collection system based on hyperspectral image features according to claim 2 is characterized in that: The data analysis unit (303) comprises a component analysis module, the component analysis module is electrically connected to a pixel purity analysis module, and the pixel purity analysis module is electrically connected to a minimum noise separation module.

7. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The superpixel segmentation unit (12) comprises a target image determination unit, the target image determination unit has an output electrically connected to the segmentation unit, the segmentation unit has an output electrically connected to the labeling unit, and the labeling unit has an output electrically connected to the training sample determination unit.

8. The forest information collection system based on hyperspectral image features according to claim 1 is characterized in that: The image extraction module (14) comprises an image preprocessing module, the image preprocessing module is electrically connected to a color feature extraction module, the color feature extraction module is electrically connected to a texture feature extraction module, the texture feature extraction module is electrically connected to a shape feature extraction module, the shape feature extraction module is electrically connected to a space relationship feature extraction module, and the space relationship feature extraction module is electrically connected to a feature comparison module.

Images