Intelligent agricultural aerial image processing method based on artificial intelligence

Abstract

The invention relates to the technical field of image processing, and particularly discloses an intelligent agricultural aerial image processing method based on artificial intelligence, and the method comprises the steps: carrying out the fine-grained image feature extraction of a to-be-processed agricultural aerial image through employing a computer vision technology based on deep learning, the image feature representation of each local area in the agricultural aerial image and the image semantic association relationship between the local areas are captured, so that the deep semantic information of the agricultural aerial image is mined, and the sharp agricultural aerial image is reconstructed. Therefore, the resolution and definition of the agricultural aerial image can be effectively improved, and a reliable data source is provided for subsequent image data analysis and processing.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and more specifically, to an artificial intelligence-based smart agricultural aerial image processing method. Background Art

[0002] With the rapid advancement of science and technology, artificial intelligence (AI) technology has gradually penetrated into all areas of people's lives. In the field of agriculture, smart agriculture, as an innovative development direction of modern agricultural science and technology, has brought unprecedented changes to agricultural production by integrating artificial intelligence technology, big data analysis and advanced image processing methods, greatly improving the efficiency and intelligence level of agricultural production.

[0003] In smart agriculture, drone aerial photography technology is widely used in agricultural monitoring, crop growth assessment, pest and disease identification and other aspects due to its high efficiency and convenience. However, in practical applications, due to the influence of various factors such as weather, lighting, shooting angle, etc., the farmland images collected by drones often have low resolution and unstable image quality, making it difficult to directly use them as source data for subsequent data analysis and processing. At present, most of the existing agricultural aerial image processing technologies focus on image preprocessing, such as simple image denoising and contrast enhancement, which often ignores the deep semantic information of the image, resulting in limited image processing effects and unable to meet the needs of smart agriculture for high-quality farmland images. Therefore, an optimized smart agricultural aerial image processing method based on artificial intelligence is expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent agricultural aerial image processing method based on artificial intelligence, which uses computer vision technology based on deep learning to perform fine-grained image feature extraction on the agricultural aerial images to be processed, and captures the image feature representation of each local area in the agricultural aerial image, as well as the image semantic association relationship between local areas, thereby mining the deep semantic information of the agricultural aerial image, so as to reconstruct the clear agricultural aerial image. In this way, the resolution and clarity of agricultural aerial images can be effectively improved, providing a reliable data source for subsequent image data analysis and processing.

[0005] Accordingly, according to one aspect of the present application, a method for processing aerial images of smart agriculture based on artificial intelligence is provided, which comprises: Obtain agricultural aerial images to be processed; Extracting image features of each local area of the agricultural aerial image to be processed to obtain a sequence of agricultural aerial image local area feature vectors; Perform local region - to - region semantic association encoding on the sequence of the local region image feature vectors of the agricultural aerial photography to obtain a sequence of context - associated local region image feature vectors of the agricultural aerial photography; Perform dimension reconstruction on the sequence of the context - associated local region image feature vectors of the agricultural aerial photography to obtain a global agricultural aerial photography feature matrix; Generate a sharpened agricultural aerial photography image based on the global agricultural aerial photography feature matrix.

[0006] In the above - mentioned artificial - intelligence - based intelligent agricultural aerial photography image - processing method, extracting the local region image features of the to - be - processed agricultural aerial photography image to obtain a sequence of local region image feature vectors of the agricultural aerial photography includes: performing image block - dividing processing on the to - be - processed agricultural aerial photography image to obtain a sequence of local agricultural aerial photography image blocks; performing image feature extraction on the sequence of the local agricultural aerial photography image blocks to obtain the sequence of the local region image feature vectors of the agricultural aerial photography.

[0007] In the above - mentioned artificial - intelligence - based intelligent agricultural aerial photography image - processing method, performing image feature extraction on the sequence of the local agricultural aerial photography image blocks to obtain the sequence of the local region image feature vectors of the agricultural aerial photography includes: respectively passing each local agricultural aerial photography image block in the sequence of the local agricultural aerial photography image blocks through an image feature extractor based on a convolutional neural network model to obtain the sequence of the local region image feature vectors of the agricultural aerial photography.

[0008] In the above - mentioned artificial - intelligence - based intelligent agricultural aerial photography image - processing method, the image feature extractor based on the convolutional neural network model includes an input layer, a two - dimensional convolutional layer, an activation layer based on the ReLU activation function, a pooling layer, and an output layer.

[0009] In the above - mentioned artificial - intelligence - based intelligent agricultural aerial photography image - processing method, performing local region - to - region semantic association encoding on the sequence of the local region image feature vectors of the agricultural aerial photography to obtain a sequence of context - associated local region image feature vectors of the agricultural aerial photography includes: passing the sequence of the local region image feature vectors of the agricultural aerial photography through a local region - to - region semantic association feature extractor based on a one - dimensional extended convolutional neural network model to obtain the sequence of the context - associated local region image feature vectors of the agricultural aerial photography.

[0010] In the above-mentioned method for processing agricultural aerial photography images based on artificial intelligence, passing the sequence of the agricultural aerial photography local region image feature vectors through a local region inter-semantic association feature extractor based on a one-dimensional dilated convolutional neural network model to obtain the sequence of the context-associated agricultural aerial photography local region image feature vectors includes: processing the sequence of the agricultural aerial photography local region image feature vectors with the following one-dimensional dilated convolution formula to obtain the sequence of the context-associated agricultural aerial photography local region image feature vectors; wherein, the one-dimensional dilated convolution formula is: ; wherein, is the -th agricultural aerial photography local region image feature vector in the sequence of the agricultural aerial photography local region image feature vectors, represents concatenation processing, represents the time window formed by concatenating the -th agricultural aerial photography local region image feature vectors in the sequence of the agricultural aerial photography local region image feature vectors, is the length of the original convolutional kernel, is the dilation rate, represents a one-dimensional dilated convolutional kernel of size , represents performing a convolution operation on the time window with the one-dimensional dilated convolutional kernel of size , is the bias term, and , is a non-linear activation function, is the -th context-associated agricultural aerial photography local region image feature vector in the sequence of the context-associated agricultural aerial photography local region image feature vectors.

[0011] In the above-mentioned method for processing agricultural aerial photography images based on artificial intelligence, generating a sharpened agricultural aerial photography image based on the global agricultural aerial photography feature matrix includes: optimizing the feature distribution of the global agricultural aerial photography feature matrix to obtain an optimized global agricultural aerial photography feature matrix; performing sharpened image reconstruction on the optimized global agricultural aerial photography feature matrix to obtain the sharpened agricultural aerial photography image.

[0012] In the above-mentioned method for processing agricultural aerial photography images based on artificial intelligence, optimizing the feature distribution of the global agricultural aerial photography feature matrix to obtain an optimized global agricultural aerial photography feature matrix includes: performing fusion correction on the global agricultural aerial photography feature matrix and the sequence of the context-associated agricultural aerial photography local region image feature vectors to obtain the optimized global agricultural aerial photography feature matrix.

[0013] In the above artificial intelligence-based intelligent agricultural aerial photography image processing method, reconstructing a sharpened image from the optimized global agricultural aerial photography feature matrix to obtain a sharpened agricultural aerial photography image includes: passing the optimized global agricultural aerial photography feature matrix through a resolution enhancer based on a generative adversarial network to obtain the sharpened agricultural aerial photography image.

[0014] Compared with the prior art, the artificial intelligence-based intelligent agricultural aerial photography image processing method provided by this application uses computer vision technology based on deep learning to perform fine-grained image feature extraction on the agricultural aerial photography image to be processed, capture the image feature representations of each local area in the agricultural aerial photography image, and the image semantic association relationships between local areas, so as to mine the deep semantic information of the agricultural aerial photography image, and use this to reconstruct the sharpened agricultural aerial photography image. In this way, the resolution and clarity of the agricultural aerial photography image can be effectively improved, providing a reliable data source for subsequent image data analysis and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a flowchart of an artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application.

[0017] Figure 2 It is a schematic structural diagram of an artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application.

[0018] Figure 3 It is a flowchart of extracting the image features of each local area of the agricultural aerial photography image to be processed to obtain a sequence of agricultural aerial photography local area image feature vectors in the artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application.

[0019] Figure 4 It is a flowchart of generating a sharpened agricultural aerial photography image based on the global agricultural aerial photography feature matrix in the artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Next, the embodiments of the present application will be described in more detail with reference to the accompanying drawings. The above and other objects, features, and advantages of the present application will become more apparent. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein. At the same time, the accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 It is a flowchart of an artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application. Figure 2 It is a schematic diagram of the architecture of an artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the artificial intelligence-based intelligent agricultural aerial photography image processing method according to an embodiment of the present application includes the steps of: S110, obtaining an agricultural aerial photography image to be processed; S120, extracting the image features of each local area of the agricultural aerial photography image to be processed to obtain a sequence of agricultural aerial photography local area image feature vectors; S130, performing local area inter-semantic association coding on the sequence of agricultural aerial photography local area image feature vectors to obtain a sequence of context-associated agricultural aerial photography local area image feature vectors; S140, performing dimension reconstruction on the sequence of context-associated agricultural aerial photography local area image feature vectors to obtain a global agricultural aerial photography feature matrix; S150, generating a sharpened agricultural aerial photography image based on the global agricultural aerial photography feature matrix.

[0022] As mentioned in the above background art, in intelligent agriculture, unmanned aerial vehicle (UAV) aerial photography technology provides an efficient and convenient solution for agricultural monitoring, crop growth assessment, pest and disease identification, etc. However, due to the fact that aerial photography images are easily affected by various factors such as weather, light, and shooting angle, the farmland images collected by UAVs often have low resolution and unstable image quality, and it is difficult to directly apply them as source data to subsequent data analysis and processing. Most of the existing agricultural aerial photography image processing technologies focus on the preprocessing of images, such as simple image denoising, contrast enhancement, etc., and often ignore the deep semantic information of the images, resulting in limited image optimization processing effects.

[0023] In response to the above technical problems, the technical concept of this application is to use computer vision technology based on deep learning to perform fine-grained image feature extraction on the agricultural aerial images to be processed, capture the image feature representation of each local area in the agricultural aerial images, and the image semantic association relationship between local areas, so as to mine the deep semantic information of the agricultural aerial images, so as to reconstruct the clear agricultural aerial images. In this way, the resolution and clarity of agricultural aerial images can be effectively improved, providing a reliable data source for subsequent image data analysis and processing.

[0024] In the above-mentioned smart agricultural aerial image processing method based on artificial intelligence, the step S110 obtains the agricultural aerial image to be processed. It should be understood that in smart agriculture, agricultural aerial images, as an important information carrier, play an irreplaceable role. Through agricultural aerial images, key information such as the overall situation of farmland, vegetation growth status, and pest and disease conditions can be intuitively understood, providing strong data support for intelligent agricultural management. However, considering that the image quality of agricultural aerial images has a vital impact on intelligent agricultural management. If the image quality is poor, it may cause data distortion, which in turn affects the accuracy of image data analysis and agricultural decision-making, and even brings unnecessary losses. Therefore, it is expected to further improve the image quality to improve the accuracy of agricultural data analysis. Based on this, in the technical solution of the present application, computer vision technology is used to mine image semantic features of agricultural aerial images to be processed, and clear image reconstruction is performed based on the image semantic features of the agricultural aerial images to be processed to improve the resolution and clarity of agricultural aerial images, so as to better meet the data requirements of intelligent agricultural management.

[0025] In the above-mentioned smart agricultural aerial image processing method based on artificial intelligence, the step S120 extracts the image features of each local area of the agricultural aerial image to be processed to obtain a sequence of agricultural aerial local area image feature vectors. Specifically, Figure 3 The present invention is a flowchart of extracting the image features of each local area of the agricultural aerial image to be processed in the artificial intelligence-based smart agricultural aerial image processing method according to an embodiment of the present application to obtain a sequence of agricultural aerial local area image feature vectors. Figure 3 As shown, the step S120 includes: S121, performing image block processing on the agricultural aerial image to be processed to obtain a sequence of agricultural aerial local image blocks; S122, performing image feature extraction on the sequence of agricultural aerial local image blocks to obtain a sequence of agricultural aerial local area image feature vectors.

[0026] Specifically, in step S121, the to-be-processed agricultural aerial image is subjected to image block processing to obtain a sequence of local agricultural aerial image blocks. It should be understood that considering that the agricultural aerial image usually has a large size and complex texture information, directly processing the entire image may lead to problems such as large computational load and slow processing speed, and may also cause loss of local detail information of the image. Therefore, in order to more effectively extract the detail features in the to-be-processed agricultural aerial image and reduce the computational burden, in the technical solution of this application, the to-be-processed agricultural aerial image is further subjected to image block processing to divide the to-be-processed agricultural aerial image into multiple local image blocks. In this way, through image block processing, independent image feature extraction can be performed on each local agricultural aerial image block to improve the fine-grained analysis ability of the features of different regions in the farmland, so as to more accurately mine the image semantic features of the to-be-processed agricultural aerial image, significantly reduce the computational complexity, and improve the processing speed.

[0027] Specifically, in step S122, image feature extraction is performed on the sequence of local agricultural aerial image blocks to obtain a sequence of local agricultural aerial region image feature vectors. In a specific example of this application, the encoding method for performing image feature extraction on the sequence of local agricultural aerial image blocks to obtain a sequence of local agricultural aerial region image feature vectors is to pass each local agricultural aerial image block in the sequence of local agricultural aerial image blocks through an image feature extractor based on a convolutional neural network model to obtain the sequence of local agricultural aerial region image feature vectors. It should be understood that the convolutional neural network model is a deep learning model with good feature extraction ability in image processing tasks. Based on this, in the technical solution of this application, an image feature extractor based on a convolutional neural network model is used to perform image feature extraction on each local agricultural aerial image block respectively to obtain the image feature representation of each local region. Specifically, the image feature extractor based on the convolutional neural network model performs layer-by-layer convolution and pooling operations on the input local agricultural aerial image block to learn the spatial structure information of the local agricultural aerial image block, extract features such as the edges, textures, and colors of the local agricultural aerial image block, so as to reflect the image semantic information of the local agricultural aerial image block and provide rich data support for subsequent image reconstruction. More specifically, the image feature extractor based on the convolutional neural network model includes an input layer, a two-dimensional convolutional layer, an activation layer based on the ReLU activation function, a pooling layer, and an output layer.

[0028] In the above artificial intelligence-based intelligent agricultural aerial photography image processing method, in step S130, local inter-region semantic association encoding is performed on the sequence of local region image feature vectors of the agricultural aerial photography to obtain a sequence of context-associated local region image feature vectors of the agricultural aerial photography. It should be understood that considering that there are certain semantic association relationships among the local regions in the agricultural aerial photography image to be processed, such as the distribution of vegetation and the spread of pests and diseases, these association relationships are of great significance for the overall semantic understanding of the agricultural aerial photography image to be processed. Therefore, in order to better extract the deep semantic information of the agricultural aerial photography image to be processed, in the technical solution of this application, local inter-region semantic association encoding is further performed on the sequence of local region image feature vectors of the agricultural aerial photography.

[0029] In a specific example of this application, the encoding method for performing local inter-region semantic association encoding on the sequence of local region image feature vectors of the agricultural aerial photography is to pass the sequence of local region image feature vectors of the agricultural aerial photography through a local inter-region semantic association feature extractor based on a one-dimensional extended convolutional neural network model to obtain the sequence of context-associated local region image feature vectors of the agricultural aerial photography. That is, a local inter-region semantic association feature extractor based on a one-dimensional extended convolutional neural network model is used to perform context-associated feature extraction on the sequence of local region image feature vectors of the agricultural aerial photography. Specifically, the one-dimensional extended convolutional neural network model performs one-dimensional convolutional operations on the local time window of the sequence of local region image feature vectors of the agricultural aerial photography to learn the spatial dependence relationship and context-associated information among the local regions, capture the semantic connection relationship between different regions in the image, thereby further improving the feature representation ability, enhancing the overall semantic understanding of the agricultural aerial photography image to be processed, and then improving the accuracy and effect of subsequent image reconstruction tasks. It is worth mentioning that compared with the traditional convolutional neural network model, the one-dimensional extended convolutional neural network model can expand the receptive field of the convolutional kernel without increasing the network parameters by introducing expansion parameters, thereby capturing a larger range of context-associated information and reducing the computational cost.

[0030] More specifically, step S130, passing the sequence of local region image feature vectors of the agricultural aerial photography through a local inter-region semantic association feature extractor based on a one-dimensional extended convolutional neural network model to obtain the sequence of context-associated local region image feature vectors of the agricultural aerial photography, includes: processing the sequence of local region image feature vectors of the agricultural aerial photography with the following one-dimensional extended convolutional formula to obtain the sequence of context-associated local region image feature vectors of the agricultural aerial photography; where the one-dimensional extended convolutional formula is: ; where is the An eigenvector of a partial area image of agricultural aerial photography, indicating connection processing, indicating the th time window formed by connecting eigenvectors of partial area images of agricultural aerial photography in the sequence of eigenvectors of partial area images of agricultural aerial photography, is the length of the original convolution kernel, is the expansion rate, indicating a one-dimensional extended convolution kernel of size ; indicating performing a convolution operation on the time window with the one-dimensional extended convolution kernel of size ; is the bias term, and ; is a non-linear activation function, is the th context-related eigenvector of a partial area image of agricultural aerial photography in the sequence of context-related eigenvectors of partial area images of agricultural aerial photography.

[0031] In the above-mentioned method for processing agricultural aerial photography images based on artificial intelligence, in step S140, the sequence of context-related eigenvectors of partial area images of agricultural aerial photography is dimensionally reconstructed to obtain a global agricultural aerial photography feature matrix. It should be understood that the sequence of context-related eigenvectors of partial area images of agricultural aerial photography contains the image features of each partial area of the agricultural aerial photography image to be processed and the semantic association relationships between each partial area, and subsequent image reconstruction tasks need to utilize the global semantic features of the image. Therefore, it is necessary to integrate the features of the sequence of context-related eigenvectors of partial area images of agricultural aerial photography to obtain a global semantic feature representation of the agricultural aerial photography image to be processed. At the same time, in order to further improve the image reconstruction effect, in the technical solution of this application, the sequence of context-related eigenvectors of partial area images of agricultural aerial photography is reconstructed into a global agricultural aerial photography feature matrix with the same dimension as the original agricultural aerial photography image, so as to achieve the purpose of retaining the original image spatial structure information during the integration of partial area image features, providing a more accurate global semantic feature representation of the image for subsequent image reconstruction tasks, and thus improving the image reconstruction effect.

[0032] In the above-mentioned method for processing agricultural aerial photography images based on artificial intelligence, in step S150, a sharpened agricultural aerial photography image is generated based on the global agricultural aerial photography feature matrix. Specifically, Figure 4 is a flowchart for generating a sharpened agricultural aerial photography image based on the global agricultural aerial photography feature matrix in the method for processing agricultural aerial photography images based on artificial intelligence according to an embodiment of the present application. As Figure 4As shown, the step S150 includes: S151, optimizing the feature distribution of the global agricultural aerial photography feature matrix to obtain an optimized global agricultural aerial photography feature matrix; S152, reconstructing a clear image from the optimized global agricultural aerial photography feature matrix to obtain the clear agricultural aerial photography image.

[0033] Specifically, in step S151, the feature distribution of the global agricultural aerial photography feature matrix is optimized to obtain an optimized global agricultural aerial photography feature matrix. It should be understood that in the above technical solution, the sequence of the local region image feature vectors of the agricultural aerial photography represents the image semantic features of the to-be-processed agricultural aerial photography image in the local image semantic space domain. After passing the sequence of the local region image feature vectors of the agricultural aerial photography through the local region semantic association feature extractor based on the one-dimensional extended convolutional neural network model, the obtained sequence of the context-associated local region image feature vectors of the agricultural aerial photography can further represent the image semantic association features between the local image semantic spaces in the global image semantic space domain. However, considering that when the sequence of the context-associated local region image feature vectors of the agricultural aerial photography is dimensionally reconstructed to obtain the global agricultural aerial photography feature matrix, the global agricultural aerial photography feature matrix may deviate from the local-global image semantic space domain feature representation of the sequence of the context-associated local region image feature vectors of the agricultural aerial photography. Therefore, it is desired to strengthen the global agricultural aerial photography feature matrix by fusing the global agricultural aerial photography feature matrix and the sequence of the context-associated local region image feature vectors of the agricultural aerial photography.

[0034] Moreover, considering the difference in the image semantic feature space distribution structure of the global agricultural aerial photography feature matrix relative to the sequence of the context-associated local region image feature vectors of the agricultural aerial photography caused by dimensional reconstruction, it is desired to improve the fusion effect of the global agricultural aerial photography feature matrix and the sequence of the context-associated local region image feature vectors of the agricultural aerial photography. Based on this, in the technical solution of the present application, first, the global agricultural aerial photography feature matrix is expanded into a first feature vector, the sequence of the context-associated local region image feature vectors of the agricultural aerial photography is concatenated into a second feature vector, and the first feature vector and the second feature vector after the expansion or concatenation of the global agricultural aerial photography feature matrix and the sequence of the context-associated local region image feature vectors of the agricultural aerial photography are fused and corrected to obtain an optimized first feature vector. Then, the optimized first feature vector is restored to the form of the global agricultural aerial photography feature matrix to obtain an optimized global agricultural aerial photography feature matrix.

[0035] Specifically, in the technical solution of the present application, the following fusion correction formula is used to fuse and correct the global agricultural aerial photography feature matrix and the sequence of context-associated agricultural aerial photography local area image feature vectors to obtain an optimized first feature vector, where the fusion correction formula is: ; where is the first feature vector after the global agricultural aerial photography feature matrix is expanded, is the second feature vector after the sequence of context-associated agricultural aerial photography local area image feature vectors is concatenated, , , feature vectors and have the same length , and are scale hyperparameters, represents the vector multiplication operation, represents the transpose of the feature vector, and is the distance difference hyperparameter, are the eigenvalues at each position of the optimized first feature vector.

[0036] Here, for the problem of difficult interaction reconstruction of the manifold network structure caused by the manifold distribution distance between the global agricultural aerial photography feature matrix and the sequence of context-associated agricultural aerial photography local area image feature vectors in the feature fusion scenario, through the effective approximation of the low-order mesoscale Hilbert space basis element structure to the network substructure in the complex manifold network structure, a low-order mesoscale submanifold interaction behavior construction based on distance representation is carried out, so as to understand the abnormal submanifold interaction in the network at the network interaction level and improve the feature fusion effect of the global agricultural aerial photography feature matrix and the sequence of context-associated agricultural aerial photography local area image feature vectors. In this way, by restoring the optimized first feature vector composed of to the global agricultural aerial photography feature matrix, the expression effect of the global agricultural aerial photography feature matrix can be improved, thereby improving the image quality of the clarified agricultural aerial photography image obtained by the resolution enhancer based on the generative adversarial network.

[0037] Specifically, in step S152, the optimized global agricultural aerial photography feature matrix is subjected to clear image reconstruction to obtain the clear agricultural aerial photography image. In a specific example of the present application, the implementation manner of performing clear image reconstruction on the optimized global agricultural aerial photography feature matrix to obtain the clear agricultural aerial photography image is to pass the optimized global agricultural aerial photography feature matrix through a resolution enhancer based on a generative adversarial network to obtain the clear agricultural aerial photography image. It should be understood that the resolution enhancer based on the generative adversarial network (GAN) is a generative model with powerful image generation and restoration capabilities. Specifically, the resolution enhancer based on the generative adversarial network consists of two parts: a generator and a discriminator. The generator gradually increases the spatial resolution and detail information of the image by performing multiple upsampling and convolution operations on the optimized global agricultural aerial photography feature matrix, and generates a high-resolution and clear agricultural aerial photography image. The discriminator is responsible for distinguishing between real agricultural aerial photography images and the images generated by the generator, and guides the generator to generate more realistic agricultural aerial photography images through a binary classification task. In this way, through alternating optimization and adversarial learning between the generator and the discriminator, the generator can generate more realistic clear agricultural aerial photography images, thereby realizing effective reconstruction and clarification processing of the to-be-processed agricultural aerial photography image.

[0038] In summary, the artificial intelligence-based intelligent agricultural aerial photography image processing method according to the embodiments of the present application is elucidated. It uses computer vision technology based on deep learning to perform fine-grained image feature extraction on the to-be-processed agricultural aerial photography image, captures the image feature representations of each local area in the agricultural aerial photography image, and the image semantic association relationships between local areas, thereby mining the deep semantic information of the agricultural aerial photography image, and performing clear agricultural aerial photography image reconstruction based on this. In this way, the resolution and clarity of the agricultural aerial photography image can be effectively improved, providing a reliable data source for subsequent image data analysis and processing.

[0039] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, advantages, effects, etc. mentioned in the present invention are only examples and not limitations. It cannot be considered that these advantages, advantages, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and easy understanding, rather than limitations. The above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0040] In the above embodiments, the descriptions of the various embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only illustrative. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0041] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced by the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0042] In addition, it is obvious that the word "including" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0043] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent agricultural aerial photography image processing method based on artificial intelligence, characterized in that, Including: Obtain the agricultural aerial image to be processed; Extract the image features of each local area of the agricultural aerial image to be processed to obtain a sequence of agricultural aerial local area image feature vectors; Perform semantic association encoding between local areas on the sequence of agricultural aerial local area image feature vectors to obtain a sequence of context-associated agricultural aerial local area image feature vectors; Perform dimension reconstruction on the sequence of context-associated agricultural aerial local area image feature vectors to obtain a global agricultural aerial feature matrix; Generate a sharpened agricultural aerial image based on the global agricultural aerial feature matrix.

2. The method for processing aerial images of smart agriculture based on artificial intelligence according to claim 1, wherein Extracting the image features of each local area of the agricultural aerial image to be processed to obtain a sequence of agricultural aerial local area image feature vectors includes: Perform image block processing on the agricultural aerial image to be processed to obtain a sequence of agricultural aerial local image blocks; Perform image feature extraction on the sequence of agricultural aerial local image blocks to obtain the sequence of agricultural aerial local area image feature vectors.

3. The method for processing aerial images of intelligent agriculture based on artificial intelligence according to claim 2, characterized in that, Performing image feature extraction on the sequence of agricultural aerial local image blocks to obtain the sequence of agricultural aerial local area image feature vectors includes: Pass each agricultural aerial local image block in the sequence of agricultural aerial local image blocks through an image feature extractor based on a convolutional neural network model to obtain the sequence of agricultural aerial local area image feature vectors.

4. The method for processing aerial images of intelligent agriculture based on artificial intelligence according to claim 3, characterized in that, The image feature extractor based on the convolutional neural network model includes an input layer, a two-dimensional convolutional layer, an activation layer based on the ReLU activation function, a pooling layer, and an output layer.

5. The method for processing aerial image of intelligent agriculture based on artificial intelligence according to claim 4, wherein, Performing semantic association encoding between local areas on the sequence of agricultural aerial local area image feature vectors to obtain a sequence of context-associated agricultural aerial local area image feature vectors includes: Pass the sequence of agricultural aerial local area image feature vectors through a local area inter-semantic association feature extractor based on a one-dimensional extended convolutional neural network model to obtain the sequence of context-associated agricultural aerial local area image feature vectors.

6. The method for processing aerial image of intelligent agriculture based on artificial intelligence according to claim 5, wherein, Passing the sequence of agricultural aerial local area image feature vectors through a local area inter-semantic association feature extractor based on a one-dimensional extended convolutional neural network model to obtain the sequence of context-associated agricultural aerial local area image feature vectors includes: Process the sequence of the local region image feature vectors of the agricultural aerial photography with the following one-dimensional dilated convolution formula to obtain the sequence of the context-associated local region image feature vectors of the agricultural aerial photography; wherein, the one-dimensional dilated convolution formula is: ; wherein, is the -th local region image feature vector of the agricultural aerial photography in the sequence of the local region image feature vectors of the agricultural aerial photography, represents concatenation processing, represents the time window concatenated by the -th local region image feature vector of the agricultural aerial photography in the sequence of the local region image feature vectors of the agricultural aerial photography, is the length of the original convolution kernel, is the dilation rate, represents a one-dimensional dilated convolution kernel with a size of , represents performing a convolution operation on the time window with the one-dimensional dilated convolution kernel with a size of , is the bias term, and , is a non-linear activation function, is the -th context-associated local region image feature vector of the agricultural aerial photography in the sequence of the context-associated local region image feature vectors of the agricultural aerial photography.

7. The method for processing aerial images of intelligent agriculture based on artificial intelligence according to claim 6, characterized in that, Generating a sharpened agricultural aerial image based on the global agricultural aerial feature matrix includes: Optimize the feature distribution of the global agricultural aerial feature matrix to obtain an optimized global agricultural aerial feature matrix; Perform sharpened image reconstruction on the optimized global agricultural aerial feature matrix to obtain the sharpened agricultural aerial image.

8. The method for processing aerial images of intelligent agriculture based on artificial intelligence according to claim 7, characterized in that, Optimizing the feature distribution of the global agricultural aerial feature matrix to obtain an optimized global agricultural aerial feature matrix includes: Fuse and correct the global agricultural aerial feature matrix and the sequence of context-associated agricultural aerial local area image feature vectors to obtain the optimized global agricultural aerial feature matrix.

9. The method for processing aerial images of intelligent agriculture based on artificial intelligence according to claim 8, wherein, Performing sharpened image reconstruction on the optimized global agricultural aerial feature matrix to obtain a sharpened agricultural aerial image includes: The optimized global agricultural aerial photography feature matrix is passed through a resolution enhancer based on a generative adversarial network to obtain the sharpened agricultural aerial photography image.

Images