Forest carbon reserve prediction method, device and equipment based on forestry carbon sink monitoring

Abstract

The invention provides a forest carbon reserve prediction method, device and equipment based on forestry carbon sink monitoring, and relates to the technical field of data analysis. According to the method, firstly, a forest optical image and climate-related data are acquired, and the forest optical image at least comprises a red light band and a near-infrared band; secondly, extracting an optical global semantic vector from the forest optical image, and performing semantic constraint on the optical global semantic vector based on growth condition semantic information in the forest optical image to form an optical growth semantic vector; then, a climate global semantic vector is extracted from climate related data; thirdly, performing semantic fusion on the optical growth semantic vector and the climate global semantic vector to obtain a forest global semantic vector; and finally, performing semantic reduction based on the forest global semantic vector to form a carbon reserve prediction result. Based on the above content, the problem that the reliability of forest carbon reserve prediction is relatively low in the prior art can be improved.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and more specifically, to a method, apparatus, and equipment for predicting forest carbon storage based on forestry carbon sink monitoring. Background Technology

[0002] Forestry carbon sequestration monitoring refers to the use of technological means to monitor and assess the carbon storage and carbon sequestration capacity of forest ecosystems. It comprehensively considers various technologies such as sensor technology, remote sensing technology, and geographic information system technology, combined with field surveys and data analysis, to achieve accurate monitoring and assessment of forest carbon storage and carbon sequestration capacity. However, in existing technologies, carbon storage is generally calculated based on biomass and carbon density. Biomass is the total weight of trees, bark, leaves, trunks, etc., and its estimation is usually determined directly through vegetation indices. This leads to generally low accuracy in estimated biomass, resulting in relatively low reliability of the determined carbon storage. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method, apparatus and equipment for predicting forest carbon storage based on forestry carbon sink monitoring, so as to improve the problem of relatively low reliability of forest carbon storage prediction in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solution: A method for predicting forest carbon storage based on forestry carbon sink monitoring includes: Acquire forest optical images and climate-related data of the target forest area, wherein the forest optical images include at least the red light band and the near-infrared band; An optical global semantic vector is extracted from the forest optical image, and semantic constraints are applied to the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector. Extract the global climate semantic vector from the climate-related data; The optical growth semantic vector and the climate global semantic vector are semantically fused to obtain the forest global semantic vector; Semantic reconstruction is performed based on the global semantic vector of the forest to form a carbon storage prediction result, wherein the carbon storage prediction result is used to reflect the carbon storage of the target forest area.

[0005] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of extracting an optical global semantic vector from the forest optical image and semantically constraining the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector includes: Extract the global optical semantic vector from the forest optical image; The reflectance of the red light band and near-infrared band in the forest optical image is calculated to obtain the red light reflectance matrix and the near-infrared reflectance matrix, and the distribution of growth status parameters is calculated based on the red light reflectance matrix and the near-infrared reflectance matrix. A growth status semantic vector is extracted from the growth status parameter distribution, wherein the growth status semantic vector is used to characterize the growth status semantic information in the forest optical image; Based on the growth status semantic vector, the optical global semantic vector is semantically constrained to form an optical growth semantic vector.

[0006] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of extracting the optical global semantic vector from the forest optical image includes: The red light band image and the near-infrared band image corresponding to the red light band and the near-infrared band are extracted from the forest optical image, respectively. Then, the red light band image and the near-infrared band image are convolved to obtain the red light band convolution vector and the near-infrared band convolution vector. The red band convolution vector and the near-infrared band convolution vector are concatenated to form a concatenated convolution vector, and the concatenated convolution vector is compressed to form a compressed vector. The concatenated convolution vector and the compressed vector are used to represent the global semantic information of the red band convolution vector and the near-infrared band convolution vector at two different scales. Self-attention mining is performed on the concatenated convolution vector to form a self-attention vector; The self-attention vector is compressed to form a self-attention compressed vector, wherein the self-attention compressed vector and the compressed vector have the same scale; Based on the self-attention compression vector, a gating mapping is performed on the compression vector to form an optical global semantic vector.

[0007] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of semantically constraining the optical global semantic vector based on the growth status semantic vector to form an optical growth semantic vector includes: The growth status semantic vector is subjected to multiple levels of convolutional compression to form multiple levels of growth status compressed vectors, wherein the growth status compressed vector of the next level is formed by convolutional compression of the growth status compressed vector of the previous level. Each growth condition compression vector at each level is deconvolved and expanded to form multiple growth condition expansion vectors, wherein the multiple growth condition expansion vectors have the same size; Based on each of the growth status expansion vectors, cross-attention mining is performed on the optical global semantic vector to obtain multiple optical constraint semantic vectors; The multiple optical constraint semantic vectors are merged to form an optical growth semantic vector.

[0008] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of semantically fusing the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector includes: Based on the optical growth semantic vector, the climate global semantic vector is semantically constrained to form a climate global constraint vector, wherein the climate global constraint vector focuses on characterizing the semantic information in the climate global semantic vector that has a correlation with the optical growth semantic vector; Based on the global climate semantic vector, the optical growth semantic vector is semantically constrained to form an optical growth constraint vector, wherein the optical growth constraint vector focuses on characterizing the semantic information in the optical growth semantic vector that has a correlation with the global climate semantic vector. Based on the global climate constraint vector, the optical growth constraint vector is gated and mapped to form a global forest semantic vector.

[0009] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of semantically constraining the climate global semantic vector based on the optical growth semantic vector to form a climate global constraint vector includes: The optical growth semantic vector is subjected to multiple levels of convolutional compression to form multiple levels of optical growth compression vectors, wherein the optical growth compression vector of the next level is formed by convolutional compression of the optical growth compression vector of the previous level. Each optical growth compression vector at each level is deconvolved and expanded to form multiple optical growth expansion vectors, wherein the multiple optical growth expansion vectors have the same size; Based on each of the optical growth expansion vectors, cross-attention mining is performed on the global climate semantic vector to obtain multiple climate constraint semantic vectors; The multiple climate constraint semantic vectors are merged to form a global climate constraint vector.

[0010] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of gating the optical growth constraint vector based on the global climate constraint vector to form a global forest semantic vector includes: Self-attention mining is performed on the global climate constraint vector to form a climate self-attention vector; The climate self-attention vector is nonlinearly activated to form an activation parameter distribution; Based on the activation parameter distribution, the optical growth constraint vector is weighted and adjusted to form an optical growth adjustment vector. Then, the optical growth constraint vector and the optical growth adjustment vector are connected to form a forest global semantic vector.

[0011] In a preferred embodiment of this application, in the above-mentioned forest carbon storage prediction method based on forestry carbon sink monitoring, the step of extracting the global climate semantic vector from the climate-related data includes: For each climate time series data included in the climate-related data, a one-dimensional convolution is performed on the climate time series data to obtain a climate time series convolution vector, and a two-dimensional convolution is performed on the Fourier transform data corresponding to the climate time series data to obtain a climate Fourier convolution vector. Based on the cross-attention mechanism, the climate temporal convolution vector and the climate Fourier convolution vector are subjected to cross-domain semantic fusion to form a global climate semantic vector.

[0012] This application also provides a forest carbon storage prediction device based on forestry carbon sink monitoring, comprising: The data acquisition module is used to acquire forest optical images and climate-related data of the target forest area, wherein the forest optical images include at least the red light band and the near-infrared band; The image semantic extraction module is used to extract the optical global semantic vector from the forest optical image, and to perform semantic constraints on the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector. The climate semantic extraction module is used to extract a global climate semantic vector from the climate-related data; The semantic fusion module is used to perform semantic fusion on the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector. The semantic restoration module is used to perform semantic restoration based on the global semantic vector of the forest to form a carbon storage prediction result, wherein the carbon storage prediction result is used to reflect the carbon storage of the target forest area.

[0013] Based on the above, this application also provides an electronic device, including: Memory, used to store computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the above-described method for predicting forest carbon storage based on forestry carbon sink monitoring.

[0014] The forest carbon storage prediction method, apparatus, and equipment based on forestry carbon sink monitoring provided in this application first acquire forest optical images and climate-related data, wherein the forest optical images include at least the red light band and the near-infrared band; secondly, extract the global optical semantic vector from the forest optical images, and semantically constrain the global optical semantic vector based on the growth status semantic information in the forest optical images to form an optical growth semantic vector; then, extract the climate global semantic vector from the climate-related data; subsequently, semantically fuse the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector; finally, semantically restore the forest global semantic vector to form the carbon storage prediction result. Based on the above, because semantic information is extracted, the potential semantic information in the data and images can be fully utilized. Therefore, compared with conventional methods, the basis for prediction is more sufficient and reliable, resulting in higher reliability of the carbon storage prediction result. Furthermore, because semantic constraints are applied based on the growth status semantic information after extracting the global optical semantic vector, the semantic representation accuracy of the obtained optical growth semantic vector in the growth semantic space is higher, thereby further improving the reliability of the prediction. Therefore, it can improve the problem of relatively low reliability in the prediction of forest carbon storage in existing technologies. Attached Figure Description

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings.

[0016] Figure 1 A structural block diagram of an electronic device provided in an embodiment of this application.

[0017] Figure 2 This is a flowchart illustrating the forest carbon storage prediction method based on forestry carbon sink monitoring provided in this application embodiment.

[0018] Figure 3 This is a schematic diagram illustrating the extraction of optical global semantic vectors as provided in an embodiment of this application.

[0019] Figure 4 A schematic diagram illustrating semantic constraints provided in the embodiments of this application.

[0020] Figure 5 This is a block diagram of a forest carbon storage prediction device based on forestry carbon sink monitoring, provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] like Figure 1 As shown in the figure, this application provides an electronic device. The electronic device may include a memory, a processor, and a forest carbon storage prediction device based on forestry carbon sink monitoring.

[0024] In detail, the memory and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, the memory and the processor can be electrically connected via one or more communication buses or signal lines. The forest carbon storage prediction device based on forestry carbon sink monitoring includes at least one software functional module stored in the memory in the form of software or firmware. The processor is used to execute executable computer programs stored in the memory, such as the software functional modules and computer programs included in the forest carbon storage prediction device based on forestry carbon sink monitoring, to implement the forest carbon storage prediction method based on forestry carbon sink monitoring provided in this application embodiment.

[0025] Optionally, the memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. Additionally, the processor may be a general-purpose processor, including a Central Processing Unit (CPU), Network Processor (NP), System on Chip (SoC), etc.; it may also be a Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0026] Understandable. Figure 1 The structure shown is for illustrative purposes only; the electronic device may also include components that are more advanced than those shown. Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown may include, for example, a communication unit for exchanging information with other devices (such as sensors).

[0027] Combination Figure 2 This application also provides a method for predicting forest carbon storage based on forestry carbon sink monitoring, applicable to the aforementioned electronic device. The method steps defined in the process of the forest carbon storage prediction method based on forestry carbon sink monitoring can be implemented by the electronic device. The following will describe... Figure 2 The specific process shown will be explained in detail.

[0028] Step S110: Obtain forest optical images and climate-related data for the target forest area.

[0029] In this embodiment, the electronic device can acquire forest optical images and climate-related data of a target forest area. The forest optical images include at least red and near-infrared bands and can be obtained by collecting information from the target forest area using appropriate remote sensing sensors. Furthermore, the climate-related data may include data on one or more dimensions such as temperature, humidity, precipitation, and sunshine duration, which can be acquired through sensors.

[0030] Step S120: Extract the optical global semantic vector from the forest optical image, and based on the growth status semantic information in the forest optical image, perform semantic constraints on the optical global semantic vector to form an optical growth semantic vector.

[0031] In this embodiment, after obtaining the forest optical image, the electronic device can extract an optical global semantic vector from the forest optical image and, based on the growth status semantic information in the forest optical image, semantically constrain the optical global semantic vector to form an optical growth semantic vector. That is, a semantic vector capable of representing global semantic information can be determined first, and then semantic constraints can be applied based on semantic information related to growth status, so that the resulting optical growth semantic vector focuses on representing semantic information related to growth status.

[0032] Step S130: Extract the global climate semantic vector from the climate-related data.

[0033] In this embodiment of the application, after obtaining the climate-related data, the electronic device can extract a global climate semantic vector from the climate-related data. That is, it can mine the overall potential semantic information within the climate-related data, thus obtaining the global climate semantic vector.

[0034] Step S140: Semantically fuse the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector.

[0035] In this embodiment, after obtaining the optical growth semantic vector and the climate global semantic vector, the electronic device can perform semantic fusion on the optical growth semantic vector and the climate global semantic vector to obtain a forest global semantic vector. That is, it fuses data from both the optical image and climate dimensions to obtain a forest global semantic vector capable of representing the potential semantic information of both dimensions.

[0036] Step S150: Semantic reconstruction is performed based on the global semantic vector of the forest to form a carbon storage prediction result.

[0037] In this embodiment of the application, after obtaining the global semantic vector of the forest, the electronic device can perform semantic reconstruction based on the global semantic vector of the forest to form a carbon storage prediction result. The carbon storage prediction result is used to reflect the carbon storage of the target forest area.

[0038] Based on the above, the extraction of semantic information allows for the full utilization of latent semantic information in data and images. Therefore, compared to conventional methods, the basis for prediction is more comprehensive and reliable, resulting in higher reliability of the carbon storage prediction results. Furthermore, after extracting the global optical semantic vector, corresponding semantic constraints are applied based on growth status semantic information, leading to higher accuracy in the semantic representation of the resulting optical growth semantic vector in the growth semantic space, further improving prediction reliability. Therefore, this approach can address the relatively low reliability of existing forest carbon storage prediction technologies.

[0039] Firstly, regarding step S110, it should be noted that the specific method for acquiring forest optical images and climate-related data of the target forest area is not limited and can be selected according to actual needs.

[0040] For example, in an alternative implementation, in order to predict forest carbon storage in real time, data collected by relevant sensors can be acquired in real time to obtain corresponding forest optical images and climate-related data.

[0041] Secondly, it should be noted that the specific method of forming the optical growth semantic vector is not limited and can be selected according to actual needs.

[0042] For example, in an alternative implementation, in order to enable the formed optical growth semantic vector to fully represent the semantic information related to the growth status, the above step S120 may further include steps S121, S122, S123 and S124, wherein the specific contents of each step are as follows.

[0043] Step S121: Extract the optical global semantic vector from the forest optical image.

[0044] In this embodiment of the application, an optical global semantic vector can be extracted from the forest optical image. That is, the potential semantic information of the forest optical image as a whole can be mined and represented in the form of a vector to obtain the optical global semantic vector.

[0045] Step S122: The reflectance of the red light band and the near-infrared band in the forest optical image is calculated to obtain the red light reflectance matrix and the near-infrared reflectance matrix, and the distribution of growth status parameters is calculated based on the red light reflectance matrix and the near-infrared reflectance matrix.

[0046] In this embodiment, reflectance can be calculated for the red and near-infrared bands of the forest optical image to obtain a red reflectance matrix and a near-infrared reflectance matrix, respectively. Furthermore, the distribution of growth status parameters can be calculated based on the red reflectance matrix and the near-infrared reflectance matrix. In other words, reflectance can be calculated for the red band data in the forest optical image to obtain the corresponding red reflectance, and reflectance can be calculated for the near-infrared band data in the forest optical image to obtain a near-infrared reflectance matrix. The resolution (size) of the red band data in the forest optical image is the same as the resolution (size) of the red reflectance matrix, and the resolution (size) of the near-infrared band data in the forest optical image is the same as the resolution (size) of the near-infrared reflectance matrix. The reflectance calculation process may include: for any pixel, calculating the difference between the pixel value (radiation value received by the sensor) and the minimum pixel value in the corresponding band to obtain a first difference; calculating the difference between the maximum and minimum pixel values ​​in the corresponding band to obtain a second difference; and then calculating the ratio between the first and second differences to obtain the reflectance corresponding to the pixel. The specific principle can be found in existing technologies and will not be elaborated here. That is, the reflectance calculation belongs to existing technologies and is not an improvement of this application. Additionally, the process of calculating the growth status parameter distribution may include: for each near-infrared reflectance in the near-infrared reflectance matrix, calculating the difference and sum between the near-infrared reflectance and the red reflectance at the same position in the red reflectance matrix; and then calculating the ratio between the difference and the sum to obtain the growth status parameter corresponding to the near-infrared reflectance. A growth status parameter close to 1 indicates healthy vegetation (e.g., strong near-infrared reflectance and low red reflectance). A growth status parameter close to 0 indicates a non-vegetated area (e.g., water body or bare soil). Negative growth status parameters may represent water bodies or urban areas, etc. The specific principles can be found in relevant existing technologies, which will not be elaborated here. Referring to the calculation of the normalized vegetation index is existing technology and does not constitute an improvement of this application.

[0047] Step S123: Extract the growth status semantic vector from the growth status parameter distribution.

[0048] In this embodiment, after obtaining the growth status parameter distribution, a growth status semantic vector can be extracted from the growth status parameter distribution. The growth status semantic vector is used to characterize the growth status semantic information in the forest optical image. For example, the process of extracting the growth status semantic vector may include: processing the growth status parameter distribution through a convolutional network layer to obtain the corresponding growth status semantic vector. The convolutional network layer may include convolutional layers, pooling layers, fully connected layers, etc.

[0049] Step S124: Based on the growth status semantic vector, semantic constraints are applied to the optical global semantic vector to form an optical growth semantic vector.

[0050] In this embodiment, after obtaining the optical global semantic vector and the growth status semantic vector, semantic constraints can be applied to the optical global semantic vector based on the growth status semantic vector to form an optical growth semantic vector. That is, during further mining of the optical global semantic vector, corresponding constraints can be applied through the growth status semantic vector to make the direction of further mining clearer, ensuring that the obtained optical growth semantic vector focuses on representing semantic information related to the growth status, or has higher semantic representation accuracy in the semantic space of the growth status.

[0051] It is understood that the specific method for extracting the optical global semantic vector from the forest optical image in step S121 above is not limited. For example, in an alternative implementation, in order to fully extract the potential semantic information in the forest optical image, step S121 above may further include steps S121a, S121b, S121c, S121d and S121e, wherein the specific contents of each step are as follows.

[0052] Step S121a: Extract the red light band image and the near-infrared band image corresponding to the red light band and the near-infrared band respectively from the forest optical image, and perform convolution processing on the red light band image and the near-infrared band image respectively to obtain the red light band convolution vector and the near-infrared band convolution vector.

[0053] In the embodiments of this application, combined with Figure 3 The red and near-infrared band images corresponding to the red and near-infrared bands can be extracted from the forest optical image, respectively. Convolutional processing is then performed on the red and near-infrared band images to obtain red-band convolution vectors and near-infrared band convolution vectors, respectively. In other words, red-band data can be extracted from the forest optical image to obtain the corresponding red-band image, which can then be processed through a convolutional network layer to obtain a red-band convolution vector. Similarly, near-infrared data can be extracted from the forest optical image to obtain the corresponding near-infrared band image, which can then be processed through a convolutional network layer to obtain a near-infrared band convolution vector. It should be noted that while the architecture of each convolutional network layer can be the same, different priorities may exist, thus requiring different parameters.

[0054] Step S121b involves concatenating the red band convolution vector and the near-infrared band convolution vector to form a concatenated convolution vector, and then compressing the concatenated convolution vector to form a compressed vector.

[0055] In this embodiment, after obtaining the red-band convolution vector and the near-infrared band convolution vector, the two convolution vectors can be concatenated to form a concatenated convolution vector, and then compressed to form a compressed vector. The concatenated convolution vector and the compressed vector are used to represent the global semantic information of the red-band and near-infrared band convolution vectors at two different scales. It should be noted that compression can be achieved through convolution, pooling, or other downsampling operations.

[0056] Step S121c: Perform self-attention mining on the concatenated convolution vector to form a self-attention vector.

[0057] In this embodiment of the application, after obtaining the spliced ​​convolution vector, self-attention mining can be performed on the spliced ​​convolution vector to form a self-attention vector. In this way, the interaction between the red light band convolution vector and the near-infrared band convolution vector can be realized, so that important semantic information related to each other can be captured or highlighted.

[0058] Step S121d: Compress the self-attention vector to form a self-attention compressed vector.

[0059] In this embodiment of the application, after obtaining the self-attention vector, the self-attention vector can be compressed (using the same compression method as in step S121b, such as convolution compression, pooling compression, etc.) to form a compressed self-attention vector. The compressed self-attention vector and the compressed vector have the same scale.

[0060] Step S121e: Based on the self-attention compression vector, perform gated mapping on the compression vector to form an optical global semantic vector.

[0061] In this embodiment of the application, after obtaining the self-attention compression vector, the compression vector can be gated and mapped to form an optical global semantic vector. The specific method of gated mapping can be referred to the relevant description below.

[0062] It is understood that the specific method of semantically constraining the optical global semantic vector in step S124 above is not limited. For example, in an alternative implementation, in order to ensure the sufficiency of semantic constraints, step S124 above may further include steps S124a, S124b, S124c and S124d, wherein the specific contents of each step are as follows.

[0063] Step S124a: Perform multi-level convolutional compression on the growth status semantic vector to form a multi-level growth status compressed vector.

[0064] In the embodiments of this application, combined with Figure 4 The growth status semantic vector can be compressed using multiple levels of convolution, resulting in multiple levels of compressed growth status vectors. Each subsequent level of compressed growth status vector is formed by convolving the previous level. For example, performing a first level of convolution on the growth status semantic vector yields a first-level compressed growth status vector, and performing a second level of convolution on the first-level vector yields a second-level compressed growth status vector. Thus, through multiple levels of convolutional compression, abstract semantic information of the growth status semantic vector at different depths can be obtained. Furthermore, convolutional compression can be implemented using convolutional layers, and edge padding can be omitted during the convolution process, with a stride greater than or equal to 2.

[0065] In step S124b, the growth status compression vector of each level is deconvolved and expanded to form multiple growth status expansion vectors.

[0066] In this embodiment, after obtaining the growth condition compression vector, deconvolution expansion can be performed on the growth condition compression vector at each level to form multiple growth condition expansion vectors. These multiple growth condition expansion vectors have the same size. That is, by performing deconvolution processing, the size of the vectors can be expanded, thereby restoring them to the size of the growth condition semantic vector (which is consistent with the size of the optical global semantic vector), or adjusting them to the size of the optical global semantic vector.

[0067] Step S124c: Based on each of the growth status expansion vectors, perform cross-attention mining on the optical global semantic vector to obtain multiple optical constraint semantic vectors.

[0068] In this embodiment of the application, after obtaining the growth status expansion vector, cross-attention mining can be performed on the optical global semantic vector (e.g., mapped to key vector and value vector) based on each growth status expansion vector (e.g., mapped to query vector) to obtain multiple optical constraint semantic vectors, such as one optical constraint semantic vector corresponding to one growth status expansion vector.

[0069] Step S124d: Merge the multiple optical constraint semantic vectors to form an optical growth semantic vector.

[0070] In this embodiment of the application, after obtaining the plurality of optical constraint semantic vectors, the plurality of optical constraint semantic vectors can be merged to form an optical growth semantic vector. For example, the plurality of optical constraint semantic vectors can be concatenated, averaged, summed, etc., to obtain the optical growth semantic vector.

[0071] Thirdly, regarding step S130, it should be noted that the specific method for extracting the global semantic vector of climate from the climate-related data is not limited and can be selected according to actual needs.

[0072] For example, in an alternative implementation, in order to ensure that the mined potential semantic information has a high degree of richness, the above-mentioned step S130 may further include steps S131 and S132, wherein the specific contents of each step are as follows.

[0073] Step S131: For each climate time series data included in the climate-related data, perform a one-dimensional convolution on the climate time series data to obtain a climate time series convolution vector, and perform a two-dimensional convolution on the Fourier transform data corresponding to the climate time series data to obtain a climate Fourier convolution vector.

[0074] In this embodiment of the application, for each climate time series data (such as temperature time series data for a recent period) included in the climate-related data, a one-dimensional convolution is performed on the climate time series data to obtain a climate time series convolution vector. In addition, a two-dimensional convolution is performed on the Fourier transform data (such as the spectrum data obtained by Fourier transform) corresponding to the climate time series data to obtain a climate Fourier convolution vector. The one-dimensional convolution and the two-dimensional convolution can be implemented through the corresponding convolutional network layers respectively.

[0075] Step S132: Based on the cross-attention mechanism, cross-domain semantic fusion is performed on the climate temporal convolution vector and the climate Fourier convolution vector to form a climate global semantic vector.

[0076] In this embodiment of the application, after obtaining the climate temporal convolution vector and the climate Fourier convolution vector, cross-domain semantic fusion can be performed on the climate temporal convolution vector and the climate Fourier convolution vector based on a cross-attention mechanism to form a global climate semantic vector. For example, cross-attention processing can be performed on the climate temporal convolution vector based on the climate Fourier convolution vector to obtain the global climate semantic vector.

[0077] Fourthly, regarding step S140, it should be noted that the specific method for semantic fusion of the optical growth semantic vector and the climate global semantic vector is not limited and can be selected according to actual needs.

[0078] For example, in an alternative implementation, in order to ensure the sufficiency of semantic fusion and make the effective semantic information in the obtained global climate semantic vector richer, the above step S140 may further include steps S141, S142 and S143, wherein the specific contents of each step are as follows.

[0079] Step S141: Based on the optical growth semantic vector, semantic constraints are applied to the climate global semantic vector to form a climate global constraint vector.

[0080] In this embodiment, the global climate semantic vector can be semantically constrained based on the optically grown semantic vector to form a global climate constraint vector. The global climate constraint vector focuses on characterizing semantic information within the global climate semantic vector that has a correlation with the optically grown semantic vector.

[0081] Step S142: Based on the global climate semantic vector, semantic constraints are applied to the optical growth semantic vector to form an optical growth constraint vector.

[0082] In this embodiment, the optical growth semantic vector can also be semantically constrained based on the global climate semantic vector (in the same way as the semantic constraint in step S141) to form an optical growth constraint vector. The optical growth constraint vector focuses on characterizing the semantic information in the optical growth semantic vector that has a correlation with the global climate semantic vector.

[0083] Step S143: Based on the climate global constraint vector, perform gated mapping on the optical growth constraint vector to form a forest global semantic vector.

[0084] In this embodiment, after obtaining the global climate constraint vector and the optical growth constraint vector, a gated mapping can be performed on the optical growth constraint vector based on the global climate constraint vector to form a global forest semantic vector. That is, since the global climate constraint vector focuses on characterizing semantic information related to the optical growth semantic vector within the global climate semantic vector, and the optical growth constraint vector focuses on characterizing semantic information related to the global climate semantic vector within the optical growth semantic vector, growth-related semantic information is more important for forest carbon storage prediction. Therefore, the optical growth constraint vector can be used as the primary basis, and the global climate constraint vector as an auxiliary basis. In this way, a gated mapping can be performed on the optical growth constraint vector based on the global climate constraint vector, so that the resulting global forest semantic vector also focuses on characterizing growth-related semantic information.

[0085] It is understood that the specific method of semantically constraining the global climate semantic vector in step S141 above is not limited. For example, in an alternative implementation, in order to ensure the sufficiency of semantic constraints, step S141 above may further include steps S141a, S141b, S141c and S141d, wherein the specific contents of each step are as follows.

[0086] Step S141a: Perform multi-level convolutional compression on the optical growth semantic vector to form a multi-level optical growth compressed vector.

[0087] In this embodiment, the optical growth semantic vector can be compressed using multiple levels of convolution to form multiple levels of compressed optical growth vectors. Each subsequent level of compressed optical growth vector is formed by convolving and compressing the preceding level. For example, the optical growth semantic vector is compressed using a first level of convolution to form a first-level compressed optical growth vector, and then the first-level compressed optical growth vector is compressed using a second level of convolution to form a second-level compressed optical growth vector. Thus, through multiple levels of convolutional compression, abstract semantic information of the optical growth semantic vector at different depths can be obtained. Furthermore, convolutional compression can be implemented using convolutional layers, and edge padding can be omitted during the convolution process, with a stride greater than or equal to 2.

[0088] In step S141b, the optical growth compression vector of each level is deconvolved and expanded to form multiple optical growth expansion vectors.

[0089] In this embodiment, the optical growth compression vector at each level can be deconvolved and expanded to form multiple optical growth expansion vectors. These multiple optical growth expansion vectors have the same size. That is, by performing deconvolution processing, the size of the vectors can be expanded, thereby restoring them to the size of the optical growth semantic vector (which is consistent with the size of the climate global semantic vector), or adjusting them to the size of the climate global semantic vector.

[0090] Step S141c: Based on each of the optical growth expansion vectors, perform cross-attention mining on the global climate semantic vector to obtain multiple climate constraint semantic vectors.

[0091] In this embodiment of the application, after obtaining the optical growth expansion vector, cross-attention mining (e.g., mapping to key vector and value vector) can be performed on the climate global semantic vector based on each optical growth expansion vector (e.g., mapped to query vector) to obtain multiple climate constraint semantic vectors, such as one optical growth expansion vector corresponding to one climate constraint semantic vector.

[0092] Step S141d: Merge the multiple climate constraint semantic vectors to form a global climate constraint vector.

[0093] In this embodiment of the application, after obtaining the plurality of climate constraint semantic vectors, the plurality of climate constraint semantic vectors can be merged to form a global climate constraint vector. For example, the plurality of climate constraint semantic vectors can be concatenated, averaged, summed, etc., to obtain the global climate constraint vector.

[0094] It is understood that the specific method of gating the optical growth constraint vector in step S143 above is not limited. For example, in an alternative implementation, in order to improve the accuracy of the gating mapping and make the semantic representation accuracy of the resulting forest global semantic vector higher, so as to give full attention to important semantic information, step S143 above may further include steps S143a, S143b and S143c. The specific contents of each step are as follows.

[0095] Step S143a: Perform self-attention mining on the global climate constraint vector to form a climate self-attention vector.

[0096] In this embodiment, self-attention mining can be performed on the global climate constraint vector to form a climate self-attention vector. This allows important semantic information within the global climate constraint vector to be extracted and represented in a focused manner.

[0097] Step S143b: Nonlinear activation is performed on the climate self-attention vector to form an activation parameter distribution.

[0098] In this embodiment, after obtaining the climate self-attention vector, it can be nonlinearly activated to form an activation parameter distribution. For example, nonlinear activation can be achieved using the sigmoid function. Since the climate self-attention vector has undergone sub-attention mining, important semantic information can be emphasized. Therefore, through nonlinear activation, the resulting activation parameters can serve as weight information representing importance; that is, the parameters at different positions may differ to represent the varying importance of those positions.

[0099] Step S143c: Based on the activation parameter distribution, the optical growth constraint vector is weighted and adjusted to form an optical growth adjustment vector; and the optical growth constraint vector and the optical growth adjustment vector are connected to form a forest global semantic vector.

[0100] In this embodiment, after obtaining the activation parameter distribution, the optical growth constraint vector can be weighted and adjusted based on the activation parameter distribution (e.g., by multiplying by position to weight the parameters at corresponding positions) to form an optical growth adjustment vector. Furthermore, the optical growth constraint vector and the optical growth adjustment vector can be concatenated (e.g., by addition, to avoid overemphasizing important semantic information in the climate global constraint vector while ignoring important semantic information in the optical growth constraint vector during weighted adjustment) to form a forest global semantic vector. Fifthly, regarding step S150, it should be noted that the specific method of semantic restoration based on the global semantic vector of the forest is not limited and can be selected according to actual needs.

[0101] For example, in an alternative implementation, the forest global semantic vector can be fully connected to obtain a forest fully connected vector, and then an identity mapping or linear mapping can be performed on the forest fully connected vector to obtain the carbon storage prediction result. The size of the forest fully connected vector can be 1*1.

[0102] Combination Figure 5 This application also provides a forest carbon storage prediction device based on forestry carbon sink monitoring, applicable to the aforementioned electronic equipment. The forest carbon storage prediction device based on forestry carbon sink monitoring may include a data acquisition module, an image semantic extraction module, a climate semantic extraction module, a semantic fusion module, and a semantic restoration module.

[0103] The data acquisition module is used to acquire forest optical images and climate-related data of the target forest area, wherein the forest optical images include at least the red light band and the near-infrared band. In this embodiment, the data acquisition module can be used to perform... Figure 2 The relevant content regarding the data acquisition module in step S110 shown can be found in the preceding description of step S110.

[0104] The image semantic extraction module is used to extract an optical global semantic vector from the forest optical image, and to apply semantic constraints to the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector. In this embodiment, the image semantic extraction module can be used to perform... Figure 2 The relevant content regarding the image semantic extraction module in step S120 shown can be found in the previous description of step S120.

[0105] The climate semantic extraction module is used to extract a global climate semantic vector from the climate-related data. In this embodiment, the climate semantic extraction module can be used to perform... Figure 2 The relevant content regarding the climate semantic extraction module in step S130 shown can be found in the previous description of step S130.

[0106] The semantic fusion module is used to semantically fuse the optical growth semantic vector and the climate global semantic vector to obtain a forest global semantic vector. In this embodiment, the semantic fusion module can be used to perform... Figure 2 The relevant content regarding the semantic fusion module in step S140 shown can be found in the previous description of step S140.

[0107] The semantic reconstruction module is used to perform semantic reconstruction based on the global semantic vector of the forest to form a carbon storage prediction result, wherein the carbon storage prediction result is used to reflect the carbon storage of the target forest area. In this embodiment of the application, the semantic reconstruction module can be used to perform... Figure 2 The relevant content regarding the semantic restoration module in step S150 shown can be found in the previous description of step S150.

[0108] In this embodiment of the application, corresponding to the above-described method for predicting forest carbon storage based on forestry carbon sink monitoring applied to the electronic device, a computer-readable storage medium is also provided. The computer-readable storage medium stores a computer program, which executes the various steps of the method for predicting forest carbon storage based on forestry carbon sink monitoring when it is run.

[0109] The steps executed by the aforementioned computer program during runtime will not be described in detail here, but can be found in the explanation of the forest carbon storage prediction method based on forestry carbon sink monitoring mentioned above.

[0110] In summary, the forest carbon storage prediction method, apparatus, and equipment based on forestry carbon sink monitoring provided in this application first acquire forest optical images and climate-related data, wherein the forest optical images include at least the red light band and the near-infrared band; second, extract the global optical semantic vector from the forest optical images, and semantically constrain the global optical semantic vector based on the growth status semantic information in the forest optical images to form an optical growth semantic vector; then, extract the climate global semantic vector from the climate-related data; subsequently, semantically fuse the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector; finally, semantically restore the forest global semantic vector to form the carbon storage prediction result. Based on the above, because semantic information is extracted, the potential semantic information in the data and images can be fully utilized. Therefore, compared with conventional methods, the prediction basis is more sufficient and reliable, resulting in a higher reliability of the obtained carbon storage prediction result. Furthermore, since semantic constraints are applied based on growth status semantic information after extracting the global optical semantic vector, the resulting optical growth semantic vector has higher semantic representation accuracy in the growth semantic space, thereby further improving the reliability of the prediction. Therefore, this can address the problem of relatively low reliability in forest carbon storage prediction in existing technologies.

[0111] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0112] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0113] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0114] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for predicting forest carbon storage based on forestry carbon sink monitoring, characterized in that, include: Acquire forest optical images and climate-related data of the target forest area, wherein the forest optical images include at least the red light band and the near-infrared band; An optical global semantic vector is extracted from the forest optical image, and semantic constraints are applied to the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector. Extract the global climate semantic vector from the climate-related data; The optical growth semantic vector and the climate global semantic vector are semantically fused to obtain the forest global semantic vector; Semantic reconstruction is performed based on the global semantic vector of the forest to form a carbon storage prediction result, wherein the carbon storage prediction result is used to reflect the carbon storage of the target forest area.

2. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 1, characterized in that, The step of extracting the optical global semantic vector from the forest optical image and semantically constraining the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector includes: Extract the global optical semantic vector from the forest optical image; The reflectance of the red light band and near-infrared band in the forest optical image is calculated to obtain the red light reflectance matrix and the near-infrared reflectance matrix, and the distribution of growth status parameters is calculated based on the red light reflectance matrix and the near-infrared reflectance matrix. A growth status semantic vector is extracted from the growth status parameter distribution, wherein the growth status semantic vector is used to characterize the growth status semantic information in the forest optical image; Based on the growth status semantic vector, the optical global semantic vector is semantically constrained to form an optical growth semantic vector.

3. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 2, characterized in that, The step of extracting the optical global semantic vector from the forest optical image includes: The red light band image and the near-infrared band image corresponding to the red light band and the near-infrared band are extracted from the forest optical image, respectively. Then, the red light band image and the near-infrared band image are convolved to obtain the red light band convolution vector and the near-infrared band convolution vector. The red band convolution vector and the near-infrared band convolution vector are concatenated to form a concatenated convolution vector, and the concatenated convolution vector is compressed to form a compressed vector. The concatenated convolution vector and the compressed vector are used to represent the global semantic information of the red band convolution vector and the near-infrared band convolution vector at two different scales. Self-attention mining is performed on the concatenated convolution vector to form a self-attention vector; The self-attention vector is compressed to form a self-attention compressed vector, wherein the self-attention compressed vector and the compressed vector have the same scale; Based on the self-attention compression vector, a gating mapping is performed on the compression vector to form an optical global semantic vector.

4. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 2, characterized in that, The step of semantically constraining the optical global semantic vector based on the growth status semantic vector to form an optical growth semantic vector includes: The growth status semantic vector is subjected to multiple levels of convolutional compression to form multiple levels of growth status compressed vectors, wherein the growth status compressed vector of the next level is formed by convolutional compression of the growth status compressed vector of the previous level. Each growth condition compression vector at each level is deconvolved and expanded to form multiple growth condition expansion vectors, wherein the multiple growth condition expansion vectors have the same size; Based on each of the growth status expansion vectors, cross-attention mining is performed on the optical global semantic vector to obtain multiple optical constraint semantic vectors; The multiple optical constraint semantic vectors are merged to form an optical growth semantic vector.

5. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 1, characterized in that, The step of semantically fusing the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector includes: Based on the optical growth semantic vector, the climate global semantic vector is semantically constrained to form a climate global constraint vector, wherein the climate global constraint vector focuses on characterizing the semantic information in the climate global semantic vector that has a correlation with the optical growth semantic vector; Based on the global climate semantic vector, the optical growth semantic vector is semantically constrained to form an optical growth constraint vector, wherein the optical growth constraint vector focuses on characterizing the semantic information in the optical growth semantic vector that has a correlation with the global climate semantic vector. Based on the global climate constraint vector, the optical growth constraint vector is gated and mapped to form a global forest semantic vector.

6. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 5, characterized in that, The step of semantically constraining the global climate semantic vector based on the optically grown semantic vector to form a global climate constraint vector includes: The optical growth semantic vector is subjected to multiple levels of convolutional compression to form multiple levels of optical growth compression vectors, wherein the optical growth compression vector of the next level is formed by convolutional compression of the optical growth compression vector of the previous level. Each optical growth compression vector at each level is deconvolved and expanded to form multiple optical growth expansion vectors, wherein the multiple optical growth expansion vectors have the same size; Based on each of the optical growth expansion vectors, cross-attention mining is performed on the global climate semantic vector to obtain multiple climate constraint semantic vectors; The multiple climate constraint semantic vectors are merged to form a global climate constraint vector.

7. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to claim 5, characterized in that, The step of gating and mapping the optical growth constraint vector based on the global climate constraint vector to form a global forest semantic vector includes: Self-attention mining is performed on the global climate constraint vector to form a climate self-attention vector; The climate self-attention vector is nonlinearly activated to form an activation parameter distribution; Based on the activation parameter distribution, the optical growth constraint vector is weighted and adjusted to form an optical growth adjustment vector. Then, the optical growth constraint vector and the optical growth adjustment vector are connected to form a forest global semantic vector.

8. The method for predicting forest carbon storage based on forestry carbon sink monitoring according to any one of claims 1-7, characterized in that, The step of extracting the global climate semantic vector from the climate-related data includes: For each climate time series data included in the climate-related data, a one-dimensional convolution is performed on the climate time series data to obtain a climate time series convolution vector, and a two-dimensional convolution is performed on the Fourier transform data corresponding to the climate time series data to obtain a climate Fourier convolution vector. Based on the cross-attention mechanism, the climate temporal convolution vector and the climate Fourier convolution vector are subjected to cross-domain semantic fusion to form a global climate semantic vector.

9. A forest carbon storage prediction device based on forestry carbon sink monitoring, characterized in that, include: The data acquisition module is used to acquire forest optical images and climate-related data of the target forest area, wherein the forest optical images include at least the red light band and the near-infrared band; The image semantic extraction module is used to extract the optical global semantic vector from the forest optical image, and to perform semantic constraints on the optical global semantic vector based on the growth status semantic information in the forest optical image to form an optical growth semantic vector. The climate semantic extraction module is used to extract a global climate semantic vector from the climate-related data; The semantic fusion module is used to perform semantic fusion on the optical growth semantic vector and the climate global semantic vector to obtain the forest global semantic vector. The semantic restoration module is used to perform semantic restoration based on the global semantic vector of the forest to form a carbon storage prediction result, wherein the carbon storage prediction result is used to reflect the carbon storage of the target forest area.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the forest carbon storage prediction method based on forestry carbon sink monitoring according to any one of claims 1-8.

Images