Forest carbon reserve calculation system and method based on neural network model

Abstract

The invention discloses a forest carbon reserve calculation system and method based on a neural network model, and relates to the technical field of ecological environment evaluation. According to the forest carbon reserve calculation system and method based on the neural network model, the forest carbon reserve calculation system is arranged, after data processing operation is achieved, the neural network model is established, historical data is introduced for training operation, an output value generated by training is compared with an expected value, then parameters are adjusted for iterative optimization operation, and the forest carbon reserve calculation accuracy is improved. According to the method, a neural network model is optimized to obtain an optimized neural network model, the optimized neural network model is tested, training optimization is carried out based on the neural network model by using historical data, compensation is carried out according to a minimum error after iterative optimization, and the method is more accurate when carbon reserve calculation of an evaluation area is completed. In addition, influence factors are considered, different features are discriminated more completely, effectively and accurately, and errors between calculated values and actual conditions are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological environment assessment, and in particular to a forest carbon reserve calculation system and method based on a neural network model. Background Art

[0002] As an important ecosystem on the earth, forests have a significant impact on the global carbon cycle and climate change. Accurate estimation of forest carbon stocks is of great significance for assessing the global carbon balance, formulating emission reduction policies and monitoring climate change.

[0003] The reference patent name is: A method for calculating forest carbon stock based on a neural network model (patent publication number: CN112884088A, patent publication date: 2021-06-01), which involves the field of ecological environmental evaluation technology. The method includes: extracting the slope, aspect, altitude and vegetation index data sets of the target forest through the DEM and TM images of the target forest; establishing a BP neural network model, with the slope, aspect, altitude and vegetation index as the input of the model, and carbon density as the output of the model; dividing the data set in step S1 into a training set and a validation set, inputting the training set into the model to train the model, and verifying the accuracy of the model through the validation set; inputting the slope, aspect, altitude and vegetation index of the forest to be tested into the model to obtain the carbon density value of the forest to be tested. When this method is used to estimate forest carbon stocks, the accuracy is high and the error is small.

[0004] Based on the description in the above documents, the existing forest carbon stock calculation realizes the estimation of large areas after image collection, and the different products in the actual area will have an impact on the forest carbon stock value. The deviation of the value caused by different influencing factors is not considered, so that the forest carbon stock calculated by the designed neural network model has a large error compared with the actual situation. Therefore, the present invention provides a forest carbon stock calculation system and method based on a neural network model. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a forest carbon stock calculation system and method based on a neural network model, which solves the problem that the existing forest carbon stock calculation realizes the estimation of a large area after collecting images, while the different products in the actual area will have an impact on the forest carbon stock value, and the deviation of the value caused by different influencing factors cannot be considered, so that the forest carbon stock calculated by the designed neural network model has a large error with the actual situation.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A forest carbon storage calculation system based on a neural network model, comprising: The data pre-processing module realizes the data collection and analysis and processing operations, specifically: Data collection unit, used to obtain historical data and data of the required assessment area; The data classification and analysis unit performs classification operations on the historical data and the evaluation area data to form the required data with different category names; The data ratio transmission unit distributes historical data according to the corresponding ratio to implement the training and testing operations of subsequent models and evaluate the normal transmission of regional data; The neural network model training module establishes a neural network model, introduces historical data for training operations, compares the output value generated by the training with the expected value, and then adjusts the parameters for iterative optimization operations to obtain the optimized neural network model, and implements testing operations on the optimized neural network model; The forest carbon stock calculation module introduces the data of the required assessment area into the optimized neural network model for processing and outputs the forest carbon stock estimation results; The visualization display panel is used to realize the combined display of data results and regional terrain.

[0007] Preferably, the data classification analysis unit performs the classification operation as follows: A1. Form a category name based on the required data content, and complete the synchronous classification operation of historical data and assessment area data based on the different content of the category name; A2. Then divide the historical data into test data, factor data and result data, divide the evaluation area data into corresponding test data and factor data, and filter out other unused data; A3. Classify the historical data, the test data, the factor data and the result data into a test data set, a factor data set and a result data set, respectively, and mark them as M, N and L, and reclassify the factor data set N.

[0008] Preferably, the classification operation of the factor data set N is: a31. Divide the collection area into square areas of equal size, remove areas without forest features, and then analyze each local area; a32. Take the type of forest as the primary analysis factor, and then extract the factors involved in the corresponding carbon storage calculation based on different forest types, and summarize the factor types and corresponding numerical results to form a single factor data group marked as G.

[0009] Preferably, the data ratio transmission unit performs the corresponding ratio allocation operation as follows: B1. Divide the data set M to be tested into a training set and a test set in a ratio of 8:2, and normalize the data. The training set is used to implement the training operation of the neural network model. B2. Then scale the data linearly so that the data value is between 0 and 1. The specific expression is: ; Where M is the data to be processed, M min is the minimum value in the processed data, M max To process the maximum value in the data, M * This is the final result obtained after data normalization.

[0010] Preferably, the operations of establishing a neural network model and introducing data for training in the neural network model training module are: C1. Construct a basic three-layer neural network structure, which consists of input layer, hidden layer and output layer, and establish a transmission relationship between the input layer, hidden layer and output layer; C2, introduce the test data set M from the input layer to realize data transmission, and then the data signal acts on the output layer through the hidden layer, and after nonlinear changes, the actual output value is generated; C3. By comparing the actual output value with the known expected value in the historical data, and implementing reciprocating iterative training operations based on the comparison results, the optimal error value is found and then a reverse compensation neural network model is performed.

[0011] Preferably, the calculation operation of the actual output value in C2 is: c21, the nodes corresponding to the transmission of the input layer, hidden layer and output layer are respectively input nodes marked with X j , hidden nodes are marked as Y j and output node Z j , and determine the input node X j With hidden node Y j The network weight threshold between kn , and the hidden node Y j With output node Z j The network weight threshold between m , and the expected value corresponding to the historical data is marked as T j ; c22, corresponding calculation of hidden node Y j The output value is marked as S y , and then according to the hidden node Y j The output data of the output node Z is calculated j The output value S z ; c23. Finally, determine the output value S based on the error calculation formula z and the expected value T j The result of time.

[0012] Preferably, S in c22y The calculation formula is: ; Among them, X j represents the jth input node X, W kn Represents the input node X j With hidden node Y j The nth network weight threshold W of the same data type k , W k represents the weight threshold of different data types at the kth iteration, θ k It represents the comparison threshold of different data types corresponding to k iterations, and when k=1, it is the output summary of the three-layer neural network structure; Output value S z The calculation formula is ; And W m Represents the hidden node Y j With output node Z j The mth network weight threshold W,θ m Indicates the corresponding hidden node Y j With output node Z j The comparison threshold between And the error calculation formula in c23 is: ; R represents the corresponding error value.

[0013] Preferably, the actual output value is compared with the expected value known in the historical data in C3, and the reciprocating iterative training operation is implemented according to the comparison result: c31, when the calculated output value S z Less than the expected value T j When , the neural network structure is changed, the corresponding hidden layer parameters are adjusted, and the hidden nodes are added one by one; c32, then the actual output value is calculated through the operations of C2 and C3, and the current actual output value is compared with the expected value known in the historical data; c33, until the actual output value S of the current operation is calculated z Greater than or equal to the expected value T j When the actual output value S in the previous operation is z Calculate the minimum optimization error value.

[0014] Preferably, the formula for calculating the minimum optimization error value in c33 is: ; at this time y The calculation formula is: ; It is expressed as the network weight threshold under the corresponding number of iterations; Then through Calculate the corresponding minimum optimization error value.

[0015] The present invention also discloses a method for calculating forest carbon reserves based on a neural network model, which specifically includes the following steps: S1. Collect historical data of known forest carbon stocks and data of the areas to be assessed, process them, and classify them according to the required requirements; S2. Then, a neural network model is established, and the classified training set is introduced to realize the training optimization of the model, and then the accuracy of the model is verified through the test set; S3. The trained neural network model is applied to the required assessment area to calculate the forest carbon storage.

[0016] The present invention provides a forest carbon storage calculation system and method based on a neural network model. Compared with the prior art, it has the following beneficial effects: (1) The forest carbon stock calculation system and method based on the neural network model are provided with a forest carbon stock calculation system, which realizes data processing operations, establishes a neural network model, introduces historical data for training operations, and compares the output value generated by the training with the expected value, and then adjusts the parameters for iterative optimization operations to obtain an optimized neural network model, and implements a test operation on the optimized neural network model, thereby using historical data to perform training optimization based on the neural network model, and then compensating according to the minimum error after iterative optimization, so that the carbon stock calculation of the assessment area is more accurate, and the influencing factors are considered more completely, different characteristics are effectively and accurately distinguished, and the error between the calculated value and the actual situation is reduced.

[0017] (2) The forest carbon storage calculation system and method based on the neural network model divides the collection area into square areas of equal size, takes the type of forest as the primary analysis factor, and then extracts the factors involved in the corresponding carbon storage calculation based on different forest types. Then, according to the corresponding neural network structure, the association between the data to be measured, the factor data and the output value is realized, and the nodes affected by the factors are obtained through iterative optimization, so as to more accurately determine the various forest characteristics in the area, reduce the training error and make the subsequent application more accurate.

[0018] (3) The forest carbon stock calculation system and method based on the neural network model establishes a neural network model, introduces the classified training set to achieve model training optimization, and feeds back the calculated minimum error to compensate the model, and then verifies the model accuracy through the test set. This operation can ensure the accuracy of subsequent application calculations, and consider multiple factors in the pre-order, while making the forest carbon stock calculation also fault-tolerant. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a principle block diagram of the forest carbon storage calculation system of the present invention; Figure 2 is an operation flow chart of the data classification and analysis unit of the present invention; Figure 3 The operation flow chart of establishing and training the neural network model of the present invention; Figure 4 It is a logic flow chart of the present invention. DETAILED DESCRIPTION

[0020] 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.

[0021] See also Figure 1-Figure 4 , the present invention provides two technical solutions: Embodiment 1: A forest carbon storage calculation system based on a neural network model, comprising: The data pre-processing module realizes the data collection and analysis and processing operations, specifically: Data collection unit, used to obtain historical data and data of the required assessment area; The data classification and analysis unit performs classification operations on the historical data and the evaluation area data to form the required data with different category names; The data ratio transmission unit distributes historical data according to the corresponding ratio to implement the training and testing operations of subsequent models and evaluate the normal transmission of regional data; The neural network model training module establishes a neural network model, introduces historical data for training operations, compares the output value generated by the training with the expected value, and then adjusts the parameters for iterative optimization operations to obtain the optimized neural network model, and implements testing operations on the optimized neural network model; The forest carbon stock calculation module introduces the data of the required assessment area into the optimized neural network model for processing and outputs the forest carbon stock estimation results; The visualization display panel is used to realize the combined display of data results and regional terrain.

[0022] Among them, after setting multiple factors for iterative training, the neural network model training module ends the iteration when the output value is less than the expected value and the error is minimum. The optimization and compensation of the neural network model are realized by reverse transmission of the minimum error data, and then the neural network model is introduced into the test set for testing. The specific test is as follows: For example, it is assumed that the actual known expected values ​​of the feature data in the test set include 0.5 of economic forest, 0.3 of bamboo forest and 0.2 of shrub forest, and the total carbon storage is 5 tons. Then the data of the test set is introduced into the neural network model for measurement, and the output values ​​are 0.489 of economic forest, 0.306 of bamboo forest and 0.205 of shrub forest, and the total carbon storage is 4.99 tons. From the above, it can be seen that the actual known expected value is almost the same as the calculated output value, so the application of the neural network model at this time is more accurate.

[0023] By setting up a forest carbon stock calculation system, after the data processing operation is realized, a neural network model is established, historical data is introduced for training operation, and the output value generated by the training is compared with the expected value, and then the parameters are adjusted for iterative optimization operation to obtain the optimized neural network model, and the optimized neural network model is tested. In this way, historical data is used to perform training optimization based on the neural network model, and then compensation is performed according to the minimum error after iterative optimization, so that the carbon stock calculation in the assessment area is completed more accurately, and the influencing factors are considered more completely, different characteristics are effectively and accurately judged, and the error between the calculated value and the actual situation is reduced.

[0024] In the embodiment of the present invention, the data classification analysis unit performs the classification operation as follows: A1. Form a category name based on the required data content, and complete the synchronous classification operation of historical data and assessment area data based on the different content of the category name; A2. Then divide the historical data into test data, factor data and result data, divide the evaluation area data into corresponding test data and factor data, and filter out other unused data; A3. Classify the historical data, the test data, the factor data and the result data into a test data set, a factor data set and a result data set, respectively, and mark them as M, N and L, and reclassify the factor data set N.

[0025] In the embodiment of the present invention, the classification operation of the factor data set N is: a31. Divide the collection area into square areas of equal size, remove areas without forest features, and then analyze each local area; a32. Take the type of forest as the primary analysis factor, and then extract the factors involved in the corresponding carbon storage calculation based on different forest types, and summarize the factor types and corresponding numerical results to form a single factor data group marked as G.

[0026] Among them, by dividing the collection area into square areas of equal size, the type of forest is used as the primary analysis factor, and then the factors involved in the corresponding carbon storage calculation are extracted based on different forest types. Then, according to the corresponding neural network structure, the association between the measured data, factor data and output values ​​is realized, and iterative optimization is achieved to obtain the nodes affected by the factors, so as to more accurately determine the various forest characteristics in the area, reduce training errors and make subsequent applications more accurate.

[0027] In the embodiment of the present invention, the data ratio transmission unit performs the corresponding ratio allocation operation as follows: B1. Divide the data set M to be tested into a training set and a test set in a ratio of 8:2, and normalize the data. The training set is used to implement the training operation of the neural network model. B2. Then scale the data linearly so that the data value is between 0 and 1. The specific expression is: ; Where M is the data to be processed, M min is the minimum value in the processed data, M max To process the maximum value in the data, M * This is the final result obtained after data normalization.

[0028] In the embodiment of the present invention, the operations of establishing a neural network model and introducing data for training in the neural network model training module are as follows: C1. Construct a basic three-layer neural network structure, which consists of input layer, hidden layer and output layer, and establish a transmission relationship between the input layer, hidden layer and output layer; C2, introduce the test data set M from the input layer to realize data transmission, and then the data signal acts on the output layer through the hidden layer, and after nonlinear changes, the actual output value is generated; C3. By comparing the actual output value with the known expected value in the historical data, and implementing reciprocating iterative training operations based on the comparison results, the optimal error value is found and then a reverse compensation neural network model is performed.

[0029] In the embodiment of the present invention, the calculation operation of the actual output value in C2 is: c21, the nodes corresponding to the transmission of the input layer, hidden layer and output layer are respectively input nodes marked with X j , hidden nodes are marked as Y j and output node Z j , and determine the input node X j With hidden node Y j The network weight threshold between kn , and the hidden node Y j With output node Z j The network weight threshold between m , and the expected value corresponding to the historical data is marked as T j ; c22, corresponding calculation of hidden node Y j The output value is marked as S y , and then according to the hidden node Y j The output data of the output node Z is calculated j The output value S z ; c23. Finally, determine the output value S based on the error calculation formula z and the expected value T j The result of time.

[0030] In the embodiment of the present invention, S in c22 y The calculation formula is: ; Among them, X j represents the jth input node X, W kn Represents the input node X j With hidden node Y j The nth network weight threshold W of the same data type k , W k represents the weight threshold of different data types at the kth iteration, θ k It represents the comparison threshold of different data types corresponding to k iterations, and when k=1, it is the output summary of the three-layer neural network structure; Output value S z The calculation formula is ; And W m Represents the hidden node Y j With output node Z j The mth network weight threshold W,θ m Indicates the corresponding hidden node Y j With output node Z j The comparison threshold between And the error calculation formula in c23 is: ; R represents the corresponding error value.

[0031] In the embodiment of the present invention, C3 compares the actual output value with the expected value known in the historical data, and implements the reciprocating iterative training operation according to the comparison result: c31, when the calculated output value S z Less than the expected value T j When , the neural network structure is changed, the corresponding hidden layer parameters are adjusted, and the hidden nodes are added one by one; c32, then the actual output value is calculated through the operations of C2 and C3, and the current actual output value is compared with the expected value known in the historical data; c33, until the actual output value S of the current operation is calculated z Greater than or equal to the expected value T j When the actual output value S in the previous operation is z Calculate the minimum optimization error value.

[0032] In the embodiment of the present invention, the formula for calculating the minimum optimization error value in c33 is: ; At this time, the preceding operation S y The calculation formula is: ; It is expressed as the network weight threshold under the corresponding number of iterations; Then through Calculate the corresponding minimum optimization error value.

[0033] Among them, by establishing a neural network model, the classified training set is introduced to realize the training optimization of the model, and the calculated minimum error is fed back to compensate the model, and then the model accuracy is verified through the test set. This operation can ensure the accuracy of subsequent application calculations, and consider multiple factors in the front sequence, while making the forest carbon stock calculation also fault-tolerant.

[0034] Embodiment 2 is different from Embodiment 1 in that: the present invention further discloses a method for calculating forest carbon reserves based on a neural network model, which specifically includes the following steps: S1. Collect historical data of known forest carbon stocks and data of the areas to be assessed, process them, and classify them according to the required requirements; S2. Then, a neural network model is established, and the classified training set is introduced to realize the training optimization of the model, and then the accuracy of the model is verified through the test set; S3. The trained neural network model is applied to the required assessment area to calculate the forest carbon storage.

[0035] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0036] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0037] 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 carbon storage calculation system based on a neural network model, characterized in that: include: The data pre-processing module realizes the data collection and analysis and processing operations, specifically: Data collection unit, used to obtain historical data and data of the required assessment area; The data classification and analysis unit performs classification operations on the historical data and the evaluation area data to form the required data with different category names; The data ratio transmission unit distributes historical data according to the corresponding ratio to implement the training and testing operations of subsequent models and evaluate the normal transmission of regional data; The neural network model training module establishes a neural network model, introduces historical data for training operations, compares the output value generated by the training with the expected value, and then adjusts the parameters for iterative optimization operations to obtain the optimized neural network model, and implements testing operations on the optimized neural network model; The forest carbon stock calculation module introduces the data of the required assessment area into the optimized neural network model for processing and outputs the forest carbon stock estimation results; The visualization display panel is used to realize the combined display of data results and regional terrain.

2. The forest carbon stock calculation system based on the neural network model according to claim 1 is characterized in that: The data classification analysis unit performs the classification operation as follows: A1. Form a category name based on the required data content, and complete the synchronous classification operation of historical data and assessment area data based on the different content of the category name; A2. Then divide the historical data into test data, factor data and result data, divide the evaluation area data into corresponding test data and factor data, and filter out other unused data; A3. Classify the historical data, the test data, the factor data and the result data into a test data set, a factor data set and a result data set, respectively, and mark them as M, N and L, and reclassify the factor data set N.

3. A forest carbon stock calculation system based on a neural network model according to claim 2, characterized in that: The classification operation of the factor data set N is: a31. Divide the collection area into square areas of equal size, remove areas without forest features, and then analyze each local area; a32. Take the type of forest as the primary analysis factor, and then extract the factors involved in the corresponding carbon storage calculation based on different forest types, and summarize the factor types and corresponding numerical results to form a single factor data group marked as G.

4. The forest carbon stock calculation system based on a neural network model according to claim 2 is characterized in that: The data ratio transmission unit performs the corresponding ratio allocation operation as follows: B1. Divide the data set M to be tested into a training set and a test set in a ratio of 8:2, and normalize the data. The training set is used to implement the training operation of the neural network model. B2. Then scale the data linearly so that the data value is between 0 and 1. The specific expression is: ； Where M is the data to be processed, M min is the minimum value in the processed data, M max To process the maximum value in the data, M * This is the final result obtained after data normalization.

5. The forest carbon stock calculation system based on a neural network model according to claim 2 is characterized in that: The operations of establishing a neural network model and introducing data for training in the neural network model training module are: C1. Construct a basic three-layer neural network structure, which consists of input layer, hidden layer and output layer, and establish a transmission relationship between the input layer, hidden layer and output layer; C2, introduce the test data set M from the input layer to realize data transmission, and then the data signal acts on the output layer through the hidden layer, and after nonlinear changes, the actual output value is generated; C3. By comparing the actual output value with the known expected value in the historical data, and implementing reciprocating iterative training operations based on the comparison results, the optimal error value is found and then a reverse compensation neural network model is performed.

6. A forest carbon stock calculation system based on a neural network model according to claim 5, characterized in that: The calculation operation of the actual output value in C2 is: c21, the nodes corresponding to the transmission of the input layer, hidden layer and output layer are respectively input nodes marked with X j , hidden nodes are marked as Y j and output node Z j , and determine the input node X j With hidden node Y j The network weight threshold between kn , and the hidden node Y j With output node Z j The network weight threshold between m , and the expected value corresponding to the historical data is marked as T j ; c22, corresponding calculation of hidden node Y j The output value is marked as S y , and then according to the hidden node Y j The output data of the output node Z is calculated j The output value S z ; c23. Finally, determine the output value S based on the error calculation formula z and the expected value T j The result of time.

7. A forest carbon stock calculation system based on a neural network model according to claim 6, characterized in that: The c22S y The calculation formula is: ； Among them, X j represents the jth input node X, W kn Represents the input node X j With hidden node Y j The nth network weight threshold W of the same data type k , W k represents the weight threshold of different data types at the kth iteration, θ k It represents the comparison threshold of different data types corresponding to k iterations, and when k=1, it is the output summary of the three-layer neural network structure; Output value S z The calculation formula is ； And W m Represents the hidden node Y j With output node Z j The mth network weight threshold W,θ m Indicates the corresponding hidden node Y j With output node Z j The comparison threshold between And the error calculation formula in c23 is: ； R represents the corresponding error value.

8. The forest carbon storage calculation system based on a neural network model according to claim 7, characterized in that: In C3, the actual output value is compared with the expected value known in the historical data, and the reciprocating iterative training operation is implemented according to the comparison result: c31, when the calculated output value S z Less than the expected value T j When , the neural network structure is changed, the corresponding hidden layer parameters are adjusted, and the hidden nodes are added one by one; c32, then the actual output value is calculated through the operations of C2 and C3, and the current actual output value is compared with the expected value known in the historical data; c33, until the actual output value S of the current operation is calculated z Greater than or equal to the expected value T j When the actual output value S in the previous operation is z Calculate the minimum optimization error value.

9. A forest carbon stock calculation system based on a neural network model according to claim 8, characterized in that: The formula for calculating the minimum optimization error value in c33 is: ； At this time, the preceding operation S y The calculation formula is: ； It is expressed as the network weight threshold under the corresponding number of iterations; Then through Calculate the corresponding minimum optimization error value.

10. A method for calculating forest carbon reserves based on a neural network model, characterized in that: The forest carbon stock calculation system based on the neural network model according to any one of claims 1 to 9 specifically comprises the following steps: S1. Collect historical data of known forest carbon stocks and data of the areas to be assessed, process them, and classify them according to the required requirements; S2. Then, a neural network model is established, and the classified training set is introduced to realize the training optimization of the model, and then the accuracy of the model is verified through the test set; S3. The trained neural network model is applied to the required assessment area to calculate the forest carbon storage.

Images