CN116796248A - Forest health care environment assessment system and method - Google Patents

Abstract

A forest health environment assessment system and method are disclosed. Firstly, carrying out image preprocessing on a panoramic image of a forest health environment to be evaluated, then obtaining a panoramic feature matrix through a convolutional neural network model, then carrying out feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix, then carrying out feature matrix segmentation on the optimized panoramic feature matrix to obtain a plurality of local feature matrices, then expanding the plurality of local feature matrices into a plurality of local feature vectors, then obtaining classification feature vectors through a context encoder, and finally, passing the classification feature vectors through a classifier to obtain a classification result for representing whether the plant diversity of the forest health environment to be evaluated meets a preset standard. Thus, the detection of the plant diversity of the forest health environment can be realized.

Description

Forest health care environment assessment system and method

Technical field

This application relates to the field of intelligent assessment, and more specifically, to a forest health care environment assessment system and its method.

Background technique

Forest health care refers to the use of natural resources and ecological services of forests to provide protection and promotion of physical and mental health for people. Forest health care environment refers to a natural environment with certain forest coverage, biodiversity, air quality, water quality, noise level, landscape beauty and other factors, which can meet people's health care needs and expectations.

In the fast pace and high pressure of modern urban life, forest wellness has become a popular way of life. In order to effectively develop and utilize forest health resources, it is necessary to conduct a scientific assessment of the forest health environment in order to determine its health value and potential, and to formulate reasonable management and protection measures.

Therefore, an optimized forest health environment assessment system is desired.

Contents of the invention

In order to solve the above technical problems, this application is proposed. Embodiments of the present application provide a forest health care environment assessment system and a method thereof. First, image preprocessing is performed on the panorama of the forest health care environment to be evaluated, and the panorama is passed through a convolutional neural network model to obtain a panoramic feature matrix. Then, the feature distribution of the panoramic feature matrix is optimized to obtain an optimized panoramic feature matrix, and then , perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices, then expand the multiple local feature matrices into multiple local feature vectors and then pass them through the context encoder to obtain the classification feature vector, and finally , passing the classification feature vector through a classifier to obtain a classification result indicating whether the plant diversity of the forest health environment to be evaluated meets the predetermined standard. In this way, the plant diversity of the forest health care environment can be detected.

According to one aspect of the present application, a forest health care environment assessment system is provided, which includes:

The panoramic image acquisition module is used to acquire the panoramic image of the forest health environment to be evaluated collected by the drone;

An image preprocessing module, used to perform image preprocessing on the panorama of the forest health care environment to be evaluated to obtain a preprocessed panorama;

A panoramic image feature extraction module, used to pass the preprocessed panorama through a convolutional neural network model as a feature extractor to obtain a panoramic feature matrix;

A feature optimization module, used to perform feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix;

A matrix segmentation module, used to segment the optimized panoramic feature matrix into feature matrices to obtain multiple local feature matrices;

A global image semantic encoding module, used to expand the multiple local feature matrices into multiple local feature vectors and then pass the converter-based context encoder to obtain the classification feature vector; and

An environmental assessment module is used to pass the classification feature vector through a classifier to obtain a classification result. The classification result is used to indicate whether the plant diversity of the forest health environment to be evaluated meets predetermined standards.

In the above-mentioned forest health care environment assessment system, the panoramic image feature extraction module is used for:

Each layer of the convolutional neural network model as a feature extractor is used to perform convolution processing, mean pooling processing along the channel dimension and nonlinear activation processing on the input data in the forward pass of the layer to obtain the input data as described as The last layer of the convolutional neural network model of the feature extractor outputs the panoramic feature matrix, wherein the input of the first layer of the convolutional neural network model serving as the feature extractor is the preprocessed panorama.

In the above-mentioned forest health care environment assessment system, the feature optimization module includes:

An optimization factor calculation unit is used to calculate the position information pattern scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix to obtain multiple position information pattern scene attention unbiased estimation factors; and

A weighted optimization unit is configured to perform weighted optimization on each position feature value of the panoramic feature matrix using the multiple position information pattern scene attention unbiased estimation factors as weighting coefficients to obtain the optimized panoramic feature matrix.

In the above-mentioned forest health care environment assessment system, the optimization factor calculation unit is used for:

Calculate the position information graph scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix using the following optimization formula to obtain the multiple position information graph scene attention unbiased estimation factors;

Among them, the optimization formula is:

Wherein, _fi is each position feature value in the panoramic feature matrix, (xi _, y _i ) is the position coordinate of each position feature value in the panoramic feature matrix, and is the global mean of all eigenvalues of the panoramic eigenmatrix,/> and/> respectively represent different functions that map two-dimensional real numbers to one-dimensional real numbers, W and H are respectively the width and height of the panoramic feature matrix, log represents the logarithmic function with base 2, and w _i represents the multiple positions Infographic scene attention unbiased estimation factors for each position in the infographic scene attention unbiased estimation factor.

In the above-mentioned forest health environment assessment system, the global image semantic encoding module is used for:

Passing the plurality of local feature vectors through the transformer-based context encoder to obtain a plurality of local semantic feature vectors; and

The plurality of local semantic feature vectors are concatenated to obtain the classification feature vector.

In the above-mentioned forest health environmental assessment system, the environmental assessment module includes:

a fully connected encoding unit, configured to perform fully connected encoding on the classification feature vector using multiple fully connected layers of the classifier to obtain a coded classification feature vector; and

A classification unit, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

According to another aspect of the present application, a forest health care environment assessment method is provided, which includes:

Obtain a panoramic view of the forest health environment to be assessed collected by drone;

Perform image preprocessing on the panorama of the forest health care environment to be evaluated to obtain a preprocessed panorama;

Pass the preprocessed panorama through a convolutional neural network model as a feature extractor to obtain a panoramic feature matrix;

Perform feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix;

Perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices;

Expand the multiple local feature matrices into multiple local feature vectors and then pass them through a context encoder based on a converter to obtain a classification feature vector; and

The classification feature vector is passed through a classifier to obtain a classification result. The classification result is used to indicate whether the plant diversity of the forest health environment to be evaluated meets the predetermined standard.

In the above forest health care environment assessment method, the preprocessed panorama is passed through the convolutional neural network model as a feature extractor to obtain a panoramic feature matrix, including:

Each layer of the convolutional neural network model as a feature extractor is used to perform convolution processing, mean pooling processing along the channel dimension and nonlinear activation processing on the input data in the forward pass of the layer to obtain the input data as described as The last layer of the convolutional neural network model of the feature extractor outputs the panoramic feature matrix, wherein the input of the first layer of the convolutional neural network model serving as the feature extractor is the preprocessed panorama.

In the above forest health care environment assessment method, feature distribution optimization is performed on the panoramic feature matrix to obtain an optimized panoramic feature matrix, including:

Calculate the position information pattern scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix to obtain multiple position information pattern scene attention unbiased estimation factors; and

Each position feature value of the panoramic feature matrix is weighted and optimized using the multiple position information graph scene attention unbiased estimation factors as weighting coefficients to obtain the optimized panoramic feature matrix.

In the above forest health care environment assessment method, the position information schema scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix is calculated to obtain multiple position information schema scene attention unbiased estimation factors, include:

Calculate the position information graph scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix using the following optimization formula to obtain the multiple position information graph scene attention unbiased estimation factors;

Among them, the optimization formula is:

Wherein, _fi is each position feature value in the panoramic feature matrix, (xi _, y _i ) is the position coordinate of each position feature value in the panoramic feature matrix, and is the global mean of all eigenvalues of the panoramic eigenmatrix,/> and/> respectively represent different functions that map two-dimensional real numbers to one-dimensional real numbers, W and H are respectively the width and height of the panoramic feature matrix, log represents the logarithmic function with base 2, and w _i represents the multiple positions Infographic scene attention unbiased estimation factors for each position in the infographic scene attention unbiased estimation factor.

Compared with the existing technology, the forest health care environment assessment system and method provided by this application first perform image preprocessing on the panorama of the forest health care environment to be evaluated, and then the panorama passes through the convolutional neural network model to obtain the panoramic feature matrix. , then perform feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix, then perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices, and then divide the multiple local feature matrices into The feature matrix is expanded into multiple local feature vectors and then passed through the context encoder to obtain the classification feature vector. Finally, the classification feature vector is passed through the classifier to obtain whether the plant diversity of the forest health environment to be evaluated meets the predetermined standards. classification results. In this way, the plant diversity of the forest health care environment can be detected.

Description of the drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. Those of ordinary skill in the art can also obtain other drawings based on these drawings without any creative effort. The following drawings are not intentionally scaled to actual sizes, but are focused on illustrating the gist of the present application.

Figure 1 is an application scenario diagram of the forest health care environment assessment system according to an embodiment of the present application.

Figure 2 is a schematic block diagram of a forest health care environment assessment system according to an embodiment of the present application.

Figure 3 is a schematic block diagram of the feature optimization module in the forest health care environment assessment system according to an embodiment of the present application.

Figure 4 is a schematic block diagram of the environmental assessment module in the forest health care environmental assessment system according to an embodiment of the present application.

Figure 5 is a flow chart of a forest health care environment assessment method according to an embodiment of the present application.

Figure 6 is a schematic diagram of the system architecture of the forest health care environment assessment method according to an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts also fall within the scope of protection of this application.

As shown in this application and claims, words such as "a", "an", "an" and/or "the" do not specifically refer to the singular and may include the plural unless the context clearly indicates an exception. Generally speaking, the terms "comprising" and "comprising" only imply the inclusion of clearly identified steps and elements, and these steps and elements do not constitute an exclusive list. The method or apparatus may also include other steps or elements.

Although this application makes various references to certain modules in systems according to embodiments of the application, any number of different modules may be used and run on user terminals and/or servers. The modules described are illustrative only, and different modules may be used by different aspects of the systems and methods.

Flowcharts are used in this application to illustrate operations performed by systems according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps can be processed in reverse order or simultaneously, as appropriate. At the same time, you can add other operations to these processes, or remove a step or steps from these processes.

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described here.

As mentioned above, in the fast pace and high pressure of modern urban life, forest wellness has become a popular way of life. In order to effectively develop and utilize forest health resources, it is necessary to conduct a scientific assessment of the forest health environment in order to determine its health value and potential, and to formulate reasonable management and protection measures. Therefore, an optimized forest health environment assessment system is desired.

Accordingly, considering that in the actual assessment process of the forest health care environment, the key is to analyze and detect the plant diversity of the forest health care environment to ensure that the plant diversity meets the predetermined requirements. Based on this, in the technical solution of this application, it is expected to use drones to collect a panoramic view of the forest health environment to be evaluated, and to evaluate the panoramic view by performing image analysis and processing based on machine vision on the panoramic view. Whether the plant diversity of the forest health environment meets predetermined standards. However, since there is a large amount of useless interference information in the panorama of the forest health environment to be evaluated collected by drones, and the plant category feature information about the forest health environment is small-scale implicit feature information in the image, it is difficult to do so. Effectively captured through traditional feature extraction methods. Therefore, in this process, the difficulty lies in how to fully express the implicit characteristic information about the plant categories of the forest health care environment in the panoramic view of the forest health care environment to be evaluated, so as to evaluate the diversity of plants in the forest health care environment. To determine their health value and potential, appropriate management and protection measures can be formulated.

In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. The development of deep learning and neural networks provides new solutions and solutions for mining the hidden feature information of plant categories of the forest health environment in the panoramic view of the forest health environment to be evaluated.

Specifically, in the technical solution of this application, first, a panoramic view of the forest health environment to be evaluated collected by a drone is obtained. It should be understood that during the actual collection process of the panorama of the forest health environment to be evaluated, the original image collected may be affected by factors such as noise, deformation or edge distortion, which will affect the subsequent processing algorithm. accuracy and stability are affected. Therefore, before performing feature extraction of the panorama of the forest health environment to be evaluated to detect the plant diversity of the forest health environment to be evaluated, it is necessary to image the panorama of the forest health environment to be evaluated. Preprocessing to obtain a preprocessed panorama to remove noise, smooth edges, and correct nonlinear effects in the original image. In other words, through such a preprocessing process, a clearer, more accurate and stable panorama can be obtained, which helps to improve the accuracy of subsequent classification.

Then, a convolutional neural network model as a feature extractor that has excellent performance in extracting implicit features of images is used to perform feature mining of the preprocessed panorama, thereby extracting the preprocessed panorama. The implicit feature distribution information about the plant categories in the forest health care environment is obtained to obtain the panoramic feature matrix.

Further, considering that although the convolutional neural network model as a feature extractor can extract the local implicit feature distribution information about the plant categories of the forest health care environment in the preprocessed panorama, due to the convolution Due to the inherent limitations of operations, it is difficult for pure CNN methods to learn clear global and long-range semantic information interactions. Therefore, in the technical solution of the present application, after further performing feature matrix segmentation on the panoramic feature matrix to obtain multiple local feature matrices, the multiple local feature matrices are expanded into multiple local feature vectors and then transformed based on Encoding is performed in the context encoder of the device to extract each local implicit feature of the plant category of the forest health care environment in the preprocessed panorama based on the global context image semantic association feature information, thereby obtaining a classification feature vector.

Then, the classification feature vector is further processed through a classifier to obtain a classification result indicating whether the plant diversity of the forest health environment to be evaluated meets the predetermined standard. That is to say, in the technical solution of the present application, the labels of the classifier include that the plant diversity of the forest health environment to be evaluated meets the predetermined standard (the first label), and the plant diversity of the forest health environment to be evaluated does not meet the predetermined standard (the first label). The predetermined criterion (second label) is met, wherein the classifier determines which classification label the classification feature vector belongs to through a soft maximum function. It is worth noting that the first label p1 and the second label p2 here do not contain artificially set concepts. In fact, during the training process, the computer model does not have the diversity of plants to be evaluated in the forest health environment. The concept of "whether the property meets the predetermined standards" is just the probability that there are two classification labels and the output feature is under these two classification labels, that is, the sum of p1 and p2 is one. Therefore, the classification result to be evaluated to see whether the plant diversity of the forest health environment meets the predetermined standards is actually converted into a binary class probability distribution that conforms to the laws of nature through classification labels. In essence, the natural probability distribution of the labels is used. The physical meaning, rather than the linguistic textual meaning of "whether the plant diversity of the forest health environment to be assessed meets the predetermined standards". It should be understood that in the technical solution of the present application, the classification label of the classifier is a detection and evaluation label for whether the plant diversity of the forest health environment to be evaluated meets the predetermined standards. Therefore, after obtaining the classification result, the Based on the classification results, the plant diversity of the forest health environment is detected to determine its health value and potential.

In particular, in the technical solution of the present application, considering that the source image semantics of each pixel of the panorama is extracted through the image semantic correlation feature of the convolutional neural network model as a feature extractor, the resulting panoramic features The eigenvalues of each position of the matrix also have corresponding position attributes. Therefore, the multiple local feature matrices obtained by performing feature matrix segmentation on the panoramic feature matrix have position-related expression attributes both within the matrix and between matrices.

However, when expanding the multiple local feature matrices into multiple local feature vectors, it involves position-wise aggregation of feature values of the local feature matrices, and when the transformer-based context encoder performs contextual semantic encoding, The position attribute is basically ignored. Therefore, it is necessary to improve the expression effect of each feature value of the panoramic feature matrix on the original feature manifold of the panoramic feature matrix when aggregated by position.

Based on this, the applicant of this application calculates the position information graphical scene attention unbiased estimation factor of the eigenvalue of each position of the panoramic feature matrix, which is expressed as:

in and/> Respectively represent different functions that map two-dimensional real numbers to one-dimensional real numbers, for example, implemented as a nonlinear activation function activation weighted and biased representation, W and H are respectively the width and height of the panoramic feature matrix, (x _i , _yi ) are the coordinates of each eigenvalue f _i of the panoramic feature matrix. For example, any vertex or center of the feature matrix can be used as the origin of the coordinates, and/> is the global mean of all eigenvalues of the panoramic feature matrix.

Here, the position information schema scene attention unbiased estimation factor is expressed by using schema information that fuses the relative geometric direction and relative geometric distance of the high-dimensional spatial position of the feature value with respect to the overall feature distribution and the information of the high-dimensional feature itself. The higher-order feature expression is used to further aggregate the shape information of the feature manifold when the feature value is aggregated by position on the overall feature distribution, so as to realize the set shape of each sub-manifold based on the feature manifold in the high-dimensional space. Unbiased estimation of distributed scene geometry to accurately express the geometric properties of the manifold shape of the feature matrix. In this way, by weighting the feature values of each position of the panoramic feature matrix with the position information graph scene attention unbiased estimation factor, it is possible to improve the performance of each feature value of the panoramic feature matrix when aggregating by position. The expression effect of the original feature manifold of the panoramic feature matrix improves the feature expression effect of the classification feature vector obtained from the global feature matrix, so as to improve the accuracy of the classification result obtained by the classifier. In this way, the plant diversity of the forest health environment can be more accurately detected and evaluated, thereby determining its health value and potential to formulate reasonable management and protection measures.

Figure 1 is an application scenario diagram of the forest health care environment assessment system according to an embodiment of the present application. As shown in Figure 1, in this application scenario, first, a panoramic view of the forest health environment to be evaluated (for example, N illustrated in Figure 1) collected by a drone (for example, N illustrated in Figure 1) is obtained D), then input the panoramic view of the forest health environment to be evaluated into a server (for example, S shown in Figure 1) where the forest health environment assessment algorithm is deployed, where the server can use the The forest health care environment assessment algorithm processes the panorama of the forest health care environment to be evaluated to obtain a classification result indicating whether the plant diversity of the forest health care environment to be evaluated meets predetermined standards.

After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be specifically introduced below with reference to the accompanying drawings.

Figure 2 is a schematic block diagram of a forest health care environment assessment system according to an embodiment of the present application. As shown in Figure 2, the forest health care environment assessment system 100 according to the embodiment of the present application includes: a panorama acquisition module 110, used to obtain a panorama of the forest health care environment to be evaluated collected by a drone; image preprocessing Module 120 is used to perform image preprocessing on the panorama of the forest health environment to be evaluated to obtain a preprocessed panorama; the panoramic image feature extraction module 130 is used to extract the preprocessed panorama as a feature The convolutional neural network model of the processor is used to obtain the panoramic feature matrix; the feature optimization module 140 is used to perform feature distribution optimization on the panoramic feature matrix to obtain the optimized panoramic feature matrix; the matrix segmentation module 150 is used to optimize the panoramic feature matrix The feature matrix is divided into feature matrices to obtain multiple local feature matrices; the global image semantic encoding module 160 is used to expand the multiple local feature matrices into multiple local feature vectors and then pass them through the context encoder based on the converter to obtain Classification feature vector; and, the environmental assessment module 170 is used to pass the classification feature vector through a classifier to obtain a classification result, which is used to indicate whether the plant diversity of the forest health environment to be evaluated meets predetermined standards.

More specifically, in this embodiment of the present application, the panorama collection module 110 is used to obtain a panorama of the forest health environment to be evaluated collected by a drone. In the actual assessment process of the forest health care environment, the key is to analyze and detect the plant diversity of the forest health care environment to ensure that the plant diversity meets the predetermined requirements. Based on this, in the technical solution of this application, a panoramic view of the forest health environment to be evaluated can be collected by drone, and the panoramic view can be evaluated by performing image analysis and processing based on machine vision on the panoramic view. Whether the plant diversity of the forest health environment meets predetermined standards.

More specifically, in this embodiment of the present application, the image preprocessing module 120 is used to perform image preprocessing on the panorama of the forest health care environment to be evaluated to obtain a preprocessed panorama. During the actual acquisition process of the panorama of the forest health environment to be evaluated, the original image collected may be affected by factors such as noise, deformation or edge distortion, which will affect the accuracy and stability of the subsequent processing algorithm. Sex has an impact. Therefore, before performing feature extraction of the panorama of the forest health environment to be evaluated to detect the plant diversity of the forest health environment to be evaluated, it is necessary to image the panorama of the forest health environment to be evaluated. Preprocessing to obtain a preprocessed panorama to remove noise, smooth edges, and correct nonlinear effects in the original image. In other words, through such a preprocessing process, a clearer, more accurate and stable panorama can be obtained, which helps to improve the accuracy of subsequent classification.

More specifically, in this embodiment of the present application, the panoramic image feature extraction module 130 is used to pass the preprocessed panorama through a convolutional neural network model as a feature extractor to obtain a panoramic feature matrix. The convolutional neural network model as a feature extractor with excellent performance in extracting implicit features of images is used to perform feature mining of the pre-processed panorama, so as to extract the relevant information in the pre-processed panorama. Hidden feature distribution information of plant categories in the forest health care environment, thereby obtaining a panoramic feature matrix.

It should be understood that Convolutional Neural Network (CNN) is an artificial neural network that is widely used in image recognition and other fields. A convolutional neural network can include an input layer, a hidden layer, and an output layer. The hidden layer can include a convolution layer, a pooling layer, an activation layer, a fully connected layer, etc. The upper layer performs corresponding processing based on the input data. Operation, the operation result is output to the next layer. The input initial data undergoes multiple layers of operations to obtain a final result.

Correspondingly, in a specific example, the panoramic image feature extraction module 130 is configured to use each layer of the convolutional neural network model as a feature extractor to respectively convolute the input data in the forward pass of the layer. product processing, mean pooling processing along the channel dimension and non-linear activation processing to output the panoramic feature matrix from the last layer of the convolutional neural network model as a feature extractor, where the The input of the first layer of the convolutional neural network model is the preprocessed panorama.

More specifically, in this embodiment of the present application, the feature optimization module 140 is used to perform feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix. In particular, in the technical solution of the present application, considering that the source image semantics of each pixel of the panorama is extracted through the image semantic correlation feature of the convolutional neural network model as a feature extractor, the resulting panoramic features The eigenvalues of each position of the matrix also have corresponding position attributes. Therefore, the multiple local feature matrices obtained by performing feature matrix segmentation on the panoramic feature matrix have position-related expression attributes both within the matrix and between matrices. However, when expanding the multiple local feature matrices into multiple local feature vectors, it involves position-wise aggregation of feature values of the local feature matrices, and when the transformer-based context encoder performs contextual semantic encoding, The position attribute is basically ignored. Therefore, it is necessary to improve the expression effect of each feature value of the panoramic feature matrix on the original feature manifold of the panoramic feature matrix when aggregated by position. Based on this, the applicant of the present application calculates the position information schema scene attention unbiased estimation factor of the feature value of each position of the panoramic feature matrix.

Correspondingly, in a specific example, as shown in Figure 3, the feature optimization module 140 includes: an optimization factor calculation unit 141, used to calculate the position information schema scene attention of each position feature value in the panoramic feature matrix. Use unbiased estimation factors to obtain multiple location information pattern scene attention unbiased estimation factors; and, a weighted optimization unit 142 for using the multiple location information pattern scene attention unbiased estimation factors as weighting coefficient pairs. Each position feature value of the panoramic feature matrix is weighted and optimized to obtain the optimized panoramic feature matrix.

Correspondingly, in a specific example, the optimization factor calculation unit 141 is configured to calculate the position information schema scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix using the following optimization formula to obtain the The multiple location information pattern scene attention unbiased estimation factors; wherein, the optimization formula is:

Wherein, _fi is each position feature value in the panoramic feature matrix, (xi _, y _i ) is the position coordinate of each position feature value in the panoramic feature matrix, and is the global mean of all eigenvalues of the panoramic eigenmatrix,/> and/> respectively represent different functions that map two-dimensional real numbers to one-dimensional real numbers, W and H are respectively the width and height of the panoramic feature matrix, log represents the logarithmic function with base 2, and w _i represents the multiple positions Infographic scene attention unbiased estimation factors for each position in the infographic scene attention unbiased estimation factor.

Here, the position information schema scene attention unbiased estimation factor is expressed by using schema information that fuses the relative geometric direction and relative geometric distance of the high-dimensional spatial position of the feature value with respect to the overall feature distribution and the information of the high-dimensional feature itself. The higher-order feature expression is used to further aggregate the shape information of the feature manifold when the feature value is aggregated by position on the overall feature distribution, so as to realize the set shape of each sub-manifold based on the feature manifold in the high-dimensional space. Unbiased estimation of distributed scene geometry to accurately express the geometric properties of the manifold shape of the feature matrix. In this way, by weighting the feature values of each position of the panoramic feature matrix with the position information graph scene attention unbiased estimation factor, it is possible to improve the performance of each feature value of the panoramic feature matrix when aggregating by position. The expression effect of the original feature manifold of the panoramic feature matrix improves the feature expression effect of the classification feature vector obtained from the global feature matrix, so as to improve the accuracy of the classification result obtained by the classifier. In this way, the plant diversity of the forest health environment can be more accurately detected and evaluated, thereby determining its health value and potential to formulate reasonable management and protection measures.

More specifically, in this embodiment of the present application, the matrix segmentation module 150 is used to segment the optimized panoramic feature matrix into feature matrices to obtain multiple local feature matrices. Although the convolutional neural network model as a feature extractor can extract the local implicit feature distribution information about the plant categories of the forest health care environment in the preprocessed panorama, due to the inherent limitations of the convolution operation , it is difficult for pure CNN methods to learn explicit global and long-range semantic information interactions. Therefore, in the technical solution of the present application, after further performing feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices, the multiple local feature matrices are expanded into multiple local feature vectors and then are processed based on Encoding is performed in the context encoder of the converter to extract each local implicit feature of the plant category of the forest health care environment in the pre-processed panorama based on the global context image semantic association feature information, thereby obtaining a classification feature vector .

More specifically, in this embodiment of the present application, the global image semantic encoding module 160 is used to expand the multiple local feature matrices into multiple local feature vectors and then pass them through a converter-based context encoder to obtain classification features. vector.

Correspondingly, in a specific example, the global image semantic encoding module 160 is used to: pass the multiple local feature vectors through the converter-based context encoder to obtain multiple local semantic feature vectors; and, The plurality of local semantic feature vectors are concatenated to obtain the classification feature vector.

More specifically, in the embodiment of the present application, the environmental assessment module 170 is used to pass the classification feature vector through a classifier to obtain a classification result. The classification result is used to represent the plant diversity of the forest health environment to be evaluated. Whether the sex meets predetermined standards. After obtaining the classification results, the plant diversity of the forest health care environment can be detected based on the classification results, thereby determining its health care value and potential.

It should be understood that the role of a classifier is to use a given category and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistics, SVM, etc. are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are needed to form a multi-classification problem, but this is prone to errors. And the efficiency is not high. Commonly used multi-classification methods include Softmax classification function.

Correspondingly, in a specific example, as shown in Figure 4, the environment assessment module 170 includes: a fully connected coding unit 171, configured to use multiple fully connected layers of the classifier to perform coding on the classification feature vector. Fully connected encoding to obtain the encoded classification feature vector; and, a classification unit 172, used to pass the encoded classification feature vector through the Softmax classification function of the classifier to obtain the classification result.

In summary, the forest health care environment assessment system 100 based on the embodiment of the present application is clarified. It first performs image preprocessing on the panorama of the forest health care environment to be evaluated, and then passes the panorama through the convolutional neural network model to obtain the panoramic feature matrix, and then , perform feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix, then perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices, and then divide the multiple local feature matrices into After expanding into multiple local feature vectors, the classification feature vector is obtained through the context encoder. Finally, the classification feature vector is passed through the classifier to obtain a classification indicating whether the plant diversity of the forest health environment to be evaluated meets the predetermined standards. result. In this way, the plant diversity of the forest health care environment can be detected.

As mentioned above, the forest health care environment assessment system 100 based on the embodiment of the present application can be implemented in various terminal devices, such as a server with the forest health care environment assessment algorithm based on the embodiment of the present application, etc. In one example, the forest health care environment assessment system 100 based on the embodiment of the present application can be integrated into a terminal device as a software module and/or a hardware module. For example, the forest health care environment assessment system 100 based on the embodiment of the present application can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the forest health care environment assessment system 100 based on the embodiment of the present application can The forest health care environment assessment system 100 of the application embodiment can also be one of the many hardware modules of the terminal device.

Alternatively, in another example, the forest health care environment assessment system 100 based on the embodiment of the present application and the terminal device can also be separate devices, and the forest health care environment assessment system 100 can be connected via wired and/or wireless devices. The network is connected to the terminal device and transmits interactive information according to the agreed data format.

Figure 5 is a flow chart of a forest health care environment assessment method according to an embodiment of the present application. As shown in Figure 5, the forest health care environment assessment method according to the embodiment of the present application includes: S110, obtaining a panoramic view of the forest health care environment to be assessed collected by the drone; S120, performing the forest health care environment to be assessed. Perform image preprocessing on the panorama of the breeding environment to obtain the preprocessed panorama; S130, pass the preprocessed panorama through the convolutional neural network model as a feature extractor to obtain the panoramic feature matrix; S140, apply the preprocessed panorama to the panorama. The feature matrix performs feature distribution optimization to obtain an optimized panoramic feature matrix; S150, perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices; S160, expand the multiple local feature matrices into multiple local feature matrices The feature vector is then passed through the context encoder based on the converter to obtain the classification feature vector; and, S170, the classification feature vector is passed through the classifier to obtain the classification result, which is used to represent the plants of the forest health environment to be evaluated. Whether diversity meets predetermined criteria.

Figure 6 is a schematic diagram of the system architecture of the forest health care environment assessment method according to an embodiment of the present application. As shown in Figure 6, in the system architecture of the forest health care environment assessment method, first, a panoramic view of the forest health care environment to be assessed collected by a drone is obtained; then, the forest health care environment to be assessed is Perform image preprocessing on the panorama to obtain a preprocessed panorama; then, pass the preprocessed panorama through a convolutional neural network model as a feature extractor to obtain a panoramic feature matrix; then, the panoramic feature matrix is Perform feature distribution optimization to obtain an optimized panoramic feature matrix; then perform feature matrix segmentation on the optimized panoramic feature matrix to obtain multiple local feature matrices; then, expand the multiple local feature matrices into multiple local feature vectors Finally, the classification feature vector is obtained through the context encoder based on the converter; finally, the classification feature vector is passed through the classifier to obtain the classification result, which is used to indicate whether the plant diversity of the forest health care environment to be evaluated meets the Predetermined standards.

In a specific example, in the above forest health care environment assessment method, the preprocessed panorama is passed through a convolutional neural network model as a feature extractor to obtain a panoramic feature matrix, including: using the preprocessed panorama as a feature extractor Each layer of the convolutional neural network model performs convolution processing, mean pooling processing along the channel dimension and nonlinear activation processing on the input data in the forward pass of the layer to be processed by the convolutional neural network as a feature extractor. The last layer of the network model outputs the panoramic feature matrix, wherein the input of the first layer of the convolutional neural network model serving as a feature extractor is the preprocessed panorama.

In a specific example, in the above forest health care environment assessment method, performing feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix includes: calculating the location information map of each position feature value in the panoramic feature matrix Formula scene attention unbiased estimation factors to obtain multiple position information graph scene attention unbiased estimation factors; and, use the multiple position information graph scene attention unbiased estimation factors as weighting coefficients to weight the panoramic features Each position feature value of the matrix is weighted and optimized to obtain the optimized panoramic feature matrix.

In a specific example, in the above forest health care environment assessment method, the position information schema scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix is calculated to obtain multiple position information schema scene attentions. The unbiased estimation factor includes: calculating the position information pattern scene attention unbiased estimation factor of each position feature value in the panoramic feature matrix using the following optimization formula to obtain the multiple position information pattern scene attention unbiased estimates. Factor; where, the optimization formula is:

Wherein, _fi is each position feature value in the panoramic feature matrix, (xi _, y _i ) is the position coordinate of each position feature value in the panoramic feature matrix, and is the global mean of all eigenvalues of the panoramic eigenmatrix,/> and/> respectively represent different functions that map two-dimensional real numbers to one-dimensional real numbers, W and H are respectively the width and height of the panoramic feature matrix, log represents the logarithmic function with base 2, and w _i represents the multiple positions Infographic scene attention unbiased estimation factors for each position in the infographic scene attention unbiased estimation factor.

In a specific example, in the above forest health care environment assessment method, the multiple local feature matrices are expanded into multiple local feature vectors and then passed through the context encoder based on the converter to obtain the classification feature vector, including: The plurality of local feature vectors are passed through the transformer-based context encoder to obtain a plurality of local semantic feature vectors; and the plurality of local semantic feature vectors are concatenated to obtain the classification feature vector.

In a specific example, in the above forest health care environment assessment method, the classification feature vector is passed through a classifier to obtain a classification result. The classification result is used to indicate whether the plant diversity of the forest health care environment to be evaluated meets the predetermined The standard includes: using multiple fully connected layers of the classifier to perform fully connected encoding on the classification feature vector to obtain a coded classification feature vector; and passing the coded classification feature vector through the Softmax classification function of the classifier. to obtain the classification results.

Here, those skilled in the art can understand that the specific operations of each step in the above forest health care environment assessment method have been introduced in detail in the description of the forest health care environment assessment system 100 above with reference to FIGS. 1 to 4 , and therefore , its repeated description will be omitted.

According to another aspect of the present application, a non-volatile computer-readable storage medium is also provided, on which computer-readable instructions are stored. When a computer is used to execute the instructions, the method as described above can be performed. .

The program part in the technology can be considered as a "product" or "artifact" in the form of executable code and/or related data, which is participated in or implemented through computer-readable media. Tangible, permanent storage media may include the memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives, or similar devices that provide storage capabilities for software.

This application uses specific words to describe embodiments of the application. For example, "first/second embodiment", "an embodiment", and/or "some embodiments" means a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that “one embodiment” or “an embodiment” or “an alternative embodiment” mentioned twice or more at different places in this specification does not necessarily refer to the same embodiment. . In addition, certain features, structures or characteristics in one or more embodiments of the present application may be appropriately combined.

Furthermore, those skilled in the art will appreciate that aspects of the present application may be illustrated and described in several patentable categories or circumstances, including any new and useful process, machine, product, or combination of matter, or combination thereof. any new and useful improvements. Accordingly, various aspects of the present application may be executed entirely by hardware, may be entirely executed by software (including firmware, resident software, microcode, etc.), or may be executed by a combination of hardware and software. The above hardware or software may be referred to as "data block", "module", "engine", "unit", "component" or "system". Additionally, aspects of the present application may be embodied as a computer product including computer-readable program code located on one or more computer-readable media.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It should also be understood that terms such as those defined in ordinary dictionaries should be construed to have meanings consistent with their meanings in the context of the relevant technology and should not be interpreted in an idealized or highly formalized sense unless expressly stated herein Ground is defined this way.

The above is a description of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of this invention have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without departing from the novel teachings and advantages of this invention. Accordingly, all such modifications are intended to be included within the scope of the invention as defined by the claims. It is to be understood that the above is a description of the invention and should not be construed as limited to the particular embodiments disclosed, and that modifications to the disclosed embodiments as well as other embodiments are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.

Claims (10)

1. A forest health environment assessment system, comprising:

the panorama acquisition module is used for acquiring a panorama of the forest health environment to be evaluated acquired by the unmanned aerial vehicle;

the image preprocessing module is used for preprocessing the panoramic image of the forest health environment to be evaluated to obtain a preprocessed panoramic image;

the panoramic image feature extraction module is used for enabling the preprocessed panoramic image to pass through a convolutional neural network model serving as a feature extractor to obtain a panoramic feature matrix;

the feature optimization module is used for carrying out feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix;

the matrix segmentation module is used for carrying out feature matrix segmentation on the optimized panoramic feature matrix to obtain a plurality of local feature matrices;

the global image semantic coding module is used for expanding the local feature matrixes into a plurality of local feature vectors and then obtaining classification feature vectors through a context encoder based on a converter; and

the environment evaluation module is used for enabling the classification feature vectors to pass through a classifier to obtain classification results, and the classification results are used for indicating whether plant diversity of the forest health environment to be evaluated meets a preset standard or not.

2. The forest health environment assessment system of claim 1, wherein the panoramic image feature extraction module is configured to:

and respectively carrying out convolution processing, mean pooling processing along a channel dimension and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model serving as the feature extractor to output the panoramic feature matrix from the last layer of the convolutional neural network model serving as the feature extractor, wherein the input of the first layer of the convolutional neural network model serving as the feature extractor is the preprocessed panoramic image.

3. The forest health environment assessment system of claim 2, wherein the feature optimization module comprises:

the optimization factor calculation unit is used for calculating the position information schema scene attention unbiased estimation factors of each position feature value in the panoramic feature matrix to obtain a plurality of position information schema scene attention unbiased estimation factors; and

and the weighted optimization unit is used for weighted optimization of each position characteristic value of the panoramic characteristic matrix by taking the unbiased estimation factors of the attention of the multiple position information schema scenes as weighting coefficients so as to obtain the optimized panoramic characteristic matrix.

4. A forest health environment assessment system according to claim 3, wherein said optimization factor calculation unit is configured to:

calculating the position information schema scene attention unbiased estimation factors of each position feature value in the panoramic feature matrix by using the following optimization formula to obtain the plurality of position information schema scene attention unbiased estimation factors;

wherein, the optimization formula is:

wherein f _i Is the feature value of each position in the panoramic feature matrix, (x) _i ,y _i ) Position coordinates for each position feature value of the panoramic feature matrix, andis the global mean value of all eigenvalues of the panoramic eigenvalue matrix,/for all eigenvalues of the panoramic eigenvalue matrix>Andrepresenting different functions of mapping two-dimensional real numbers into one-dimensional real numbers, W and H are the width and the height of the panoramic feature matrix, log represents a logarithmic function based on 2, and W _i Each of the plurality of location information schema scene attention unbiased estimation factors is represented.

5. The forest health environment assessment system according to claim 4, wherein the global image semantic coding module is configured to:

passing the plurality of local feature vectors through the converter-based context encoder to obtain a plurality of local semantic feature vectors; and

And cascading the plurality of local semantic feature vectors to obtain the classification feature vector.

6. The forest health environment assessment system of claim 5, wherein the environment assessment module comprises:

the full-connection coding unit is used for carrying out full-connection coding on the classification characteristic vectors by using a plurality of full-connection layers of the classifier so as to obtain coded classification characteristic vectors; and

and the classification unit is used for passing the coding classification feature vector through a Softmax classification function of the classifier to obtain the classification result.

7. A forest health environment assessment method, comprising:

acquiring a panoramic view of a forest health and maintenance environment to be evaluated, which is acquired by an unmanned aerial vehicle;

performing image preprocessing on the panoramic image of the forest health environment to be evaluated to obtain a preprocessed panoramic image;

the preprocessed panoramic image is passed through a convolutional neural network model serving as a feature extractor to obtain a panoramic feature matrix;

performing feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix;

performing feature matrix segmentation on the optimized panoramic feature matrix to obtain a plurality of local feature matrices;

The local feature matrixes are unfolded to be a plurality of local feature vectors, and then the local feature vectors are processed by a context encoder based on a converter to obtain classified feature vectors; and

and the classification feature vector is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the plant diversity of the forest health and maintenance environment to be evaluated meets a preset standard.

8. The forest health environment assessment method according to claim 7, wherein passing the preprocessed panorama through a convolutional neural network model as a feature extractor to obtain a panorama feature matrix, comprising:

and respectively carrying out convolution processing, mean pooling processing along a channel dimension and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model serving as the feature extractor to output the panoramic feature matrix from the last layer of the convolutional neural network model serving as the feature extractor, wherein the input of the first layer of the convolutional neural network model serving as the feature extractor is the preprocessed panoramic image.

9. The forest health environment assessment method according to claim 8, wherein performing feature distribution optimization on the panoramic feature matrix to obtain an optimized panoramic feature matrix comprises:

Calculating the position information schema scene attention unbiased estimation factors of each position feature value in the panoramic feature matrix to obtain a plurality of position information schema scene attention unbiased estimation factors; and

and carrying out weighted optimization on each position characteristic value of the panoramic characteristic matrix by taking the unbiased estimation factors of the attention of the multiple position information schema scenes as weighting coefficients to obtain the optimized panoramic characteristic matrix.

10. The method of forest health environment assessment according to claim 9, wherein calculating the location information schema scene attention unbiased estimation factors for each location feature value in the panoramic feature matrix to obtain a plurality of location information schema scene attention unbiased estimation factors, comprises:

calculating the position information schema scene attention unbiased estimation factors of each position feature value in the panoramic feature matrix by using the following optimization formula to obtain the plurality of position information schema scene attention unbiased estimation factors;

wherein, the optimization formula is:

wherein f _i Is the feature value of each position in the panoramic feature matrix, (x) _i ,y _i ) Position coordinates for each position feature value of the panoramic feature matrix, and Is the global mean value of all eigenvalues of the panoramic eigenvalue matrix,/for all eigenvalues of the panoramic eigenvalue matrix>Andrepresenting different functions of mapping two-dimensional real numbers into one-dimensional real numbers, W and H are the width and the height of the panoramic feature matrix, log represents a logarithmic function based on 2, and W _i Each of the plurality of location information schema scene attention unbiased estimation factors is represented.