Method and device for classifying growth periods of crop images

Abstract

The invention relates to the technical field of image processing, in particular to a crop image growth period classification method and device. In order to solve the technical problem that the conventional means cannot give consideration to the recognition and judgment of various growth states of a seed selection nursery, a multispectral camera is adopted to collect multiband crop images, multiple vegetation index features of the crop images are extracted, sub-image region division is performed on the crop images, entropy features of each sub-image region are extracted, and the entropy features of the sub-image regions are extracted. A feature space matrix is obtained by integrating vegetation index features and entropy features, and a neural network model is trained as a sample after a corresponding real growth period type is marked. Therefore, the neural network model can continuously improve the recognition of the incidence relation between the feature space matrix obtained by combining the entropy feature and the vegetation index feature and the corresponding growth period type. Compared with a traditional machine learning method, classification and identification of the growth periods of the crops have higher accuracy and remarkable practicability.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method and device for classifying the growth period of crop images. Background Technology

[0002] The growth period is a growth rhythm formed by crops through long-term adaptation to the natural environment, and it can characterize the growth status of crops. Growth period data can provide a scientific basis for cultivation management assessment and yield prediction.

[0003] When faced with a large number of seed selection nurseries, traditional methods cannot simultaneously identify and judge the various growth states of the nurseries in a short period of time. Summary of the Invention

[0004] In view of this, the purpose of this application is to propose a method and device for classifying the growth period of crop images, so as to solve or partially solve the above-mentioned technical problems.

[0005] To achieve the above objectives, this application provides a method for classifying the growth stage of crop images, comprising: Multispectral cameras are used to acquire crop images in multiple bands of the crop planting area, and the true growth stage type of the crop planting area is determined. Each crop image in each band is divided into multiple sub-image regions. The sub-image regions are the same for each crop image in each band, and each sub-image region corresponds to multiple bands. Extract multiple vegetation index features from crop images, extract the entropy value features of each band corresponding to each sub-image region, construct a feature space matrix with the entropy value features of each band of each sub-image region and multiple vegetation index features, and label the corresponding real growth period type of the feature space matrix as a sample. Using the samples, supervised training is performed on the pre-constructed neural network model. The classification results of the reproductive period output by the training are compared with the labeled real reproductive period types. The parameters of the neural network model are adjusted according to the differences in the comparison to obtain the adjusted neural network model. Once the training termination condition is met, the final adjusted neural network model is used as the reproductive period classification model. By using multispectral images of crops to be classified in multiple bands acquired by a multispectral camera, the feature space matrix corresponding to the crop images to be classified in multiple bands is determined. The feature space matrix to be classified is input into the reproductive period classification model to obtain the reproductive period classification result output.

[0006] Based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method described above when executing the computer program.

[0007] As can be seen from the above, the crop image growth period classification method and device provided in this application can use a multispectral camera to acquire crop images of multiple bands in the crop planting area and determine the actual growth period type corresponding to the crop planting area. Since the image presentation effects of different bands are different and the growth period information they contain is different, acquiring crop images of multiple bands ensures that the information contained in the growth period is sufficient for subsequent classification and recognition. Then, the crop image of each band is divided into multiple sub-image regions. Since the crop images of each band have the same size, the division criteria for the corresponding sub-image regions are the same, so the sub-image regions after the division of the crop image of each band are the same, resulting in multiple bands in each sub-image region. Then, the multi-vegetation index features of the crop image are extracted to determine the entropy features of each band in each sub-image region, thus classifying each sub-image region... A feature space matrix is ​​constructed using entropy features of each band and multiple vegetation index features. After labeling the corresponding growth period types, samples are obtained. These samples are then used to perform supervised training on a pre-built neural network model. The parameters of the neural network model are adjusted based on the differences between the growth period classification results obtained during training and the labeled actual growth period types, improving the accuracy of the adjusted model in growth period classification. Once the training termination condition is met, a growth period classification model is obtained. Because this model incorporates entropy features and vegetation index features during training, it achieves more accurate classification of crop growth periods. Finally, for crop images in multiple bands requiring growth period classification, the corresponding feature space matrices for each band are determined. The growth period classification model is then used to classify the growth period using these feature space matrices, resulting in the output growth period classification result. This growth period classification model combines the influence of entropy features and vegetation index features on growth period classification, ensuring accurate growth period classification and guaranteeing the real-time performance and accuracy of the obtained results. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in this application or related technologies, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating the crop image growth period classification method according to an embodiment of this application; Figure 2This is a schematic diagram of the training model for the crop image growth period classification method according to an embodiment of this application; Figure 3 This is a structural block diagram of a crop image growth period classification device according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0011] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0012] Definitions: EVA-Net: Entropy and Vegetation Index Analysis Network.

[0013] NDVI: Normalized Difference Vegetation Index.

[0014] NDRE: Normalized Difference Red Edge Index.

[0015] GNDVI: Green Normalized Difference Vegetation Index.

[0016] SR: Simple Ratio Index.

[0017] EVI: Enhanced Vegetation Index.

[0018] GLI: Green Leaf Index.

[0019] RECI: Red Edge Chlorophyll Index.

[0020] MTCI: MERIS Terrestrial Chlorophyll Index.

[0021] SIPI: Structure Insensitive Pigment Index.

[0022] MSR: Modified Simple Ratio.

[0023] The growth period is a crucial monitoring indicator for crops in agriculture. Growth period data can provide a scientific basis for cultivation management assessment and yield prediction. Accurate identification of the growth period and matching it with scientific management measures can effectively address the production challenges brought about by climate change and extreme weather conditions, thereby assisting in the breeding of stress-resistant and widely adaptable crop varieties and ensuring stable and high crop yields.

[0024] Traditional methods cannot make selections that take into account various breeding traits in a short period of time, and crops with similar morphological characteristics at adjacent growth stages are difficult to distinguish accurately with the naked eye. Therefore, satellite remote sensing technology is now used for crop growth stage monitoring. It has shown high monitoring accuracy and reliability in large-scale applications, but it is limited by weather, resolution and acquisition frequency, making it difficult to meet the needs of fine-grained monitoring at the micro-level.

[0025] Thresholding, maximum slope and curve fitting methods based on optical remote sensing are the most widely used in crop growth period monitoring research. However, they are prone to suppressing important phenological characteristics due to overfitting and can only be analyzed after the growth period, making real-time monitoring impossible.

[0026] The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0027] The crop image growth period classification method proposed in this application includes crops such as wheat, rice, and millet.

[0028] like Figure 1 and 2 As shown, the method includes: Step 101: Use a multispectral camera to acquire crop images of multiple bands in the crop planting area and determine the true growth period type of the crop planting area.

[0029] In practice, the multispectral camera can be mounted on a drone, allowing the drone to carry the camera and move around the crop planting area (e.g., a test planting area) to capture crop images in multiple wavelengths. These multiple wavelengths include blue light, green light, red light, red-edge light, and near-infrared light.

[0030] The true growth period type can be determined by conducting surveys of crop planting areas, and then the crop planting areas can be regionalized and divided according to the true growth period type.

[0031] The growth period includes: the booting stage, the heading stage, the flowering stage, and the maturity stage.

[0032] Since images from different bands present different effects and contain different information about the growth period, acquiring crop images from multiple bands can ensure that they contain enough information about the growth period, which is convenient for subsequent classification and recognition.

[0033] Step 102: Divide the crop image of each band into multiple sub-image regions. The sub-image regions are the same for each band, and each sub-image region corresponds to multiple bands.

[0034] In practice, the method of dividing multiple sub-image regions can be to divide them according to a predetermined size, which is determined proportionally based on the small area of ​​the crop planting region.

[0035] Step 103: Extract multiple vegetation index features from the crop image, extract the entropy features of each band corresponding to each sub-image region, and combine the entropy features of each band of each sub-image region (such as...) Figure 2 As shown, near-infrared-entropy maps, red-entropy maps, blue-entropy maps, green-entropy maps, and red-edge-entropy maps are used to form a matrix H), and a feature space matrix (H, VI) is constructed with various vegetation index features (VI). The feature space matrix is ​​used to label the corresponding real growth period type as a sample.

[0036] In practice, since entropy features can better characterize the pixel characteristics of the corresponding region, it is easier to identify and judge the growth stage type; in addition, vegetation index features can also characterize the growth status of crops, which is also very helpful for identifying and judging the growth stage type.

[0037] Step 104: Using the samples, supervised training is performed on the pre-constructed neural network model. The classification results of the reproductive period output by the training are compared with the labeled real reproductive period types. The parameters of the neural network model are adjusted according to the differences in the comparison to obtain the adjusted neural network model. The training termination condition is determined to be met, and the final adjusted neural network model is used as the reproductive period classification model.

[0038] In practice, the entropy features of each band are combined with the corresponding vegetation index features to construct a feature space matrix to form samples. The samples obtained in this way can be used to train the neural network model by combining the entropy features and vegetation index features. This allows the neural network model to continuously improve the correlation between the feature space matrix obtained by combining the entropy features and vegetation index features and the corresponding growth period type, thereby ensuring that the growth period classification model obtained after training can classify and identify crop growth periods more accurately.

[0039] Step 105: Using multispectral images of the crop to be classified acquired by a multispectral camera, determine the feature space matrix corresponding to the crop images to be classified in multiple bands.

[0040] In practice, a drone equipped with a multispectral camera can be used to collect images of the crop to be classified in multiple bands in real time. After removing the labels of the true growth period type in steps 102 and 103 above, the images of the crop to be classified in multiple bands are processed to obtain the feature space matrix to be classified.

[0041] Step 106: Input the feature space matrix to be classified into the reproductive period classification model to obtain the reproductive period classification result output.

[0042] The above scheme utilizes a multispectral camera to acquire crop images across multiple bands of the crop planting area and determine the actual growth stage type corresponding to the crop planting area. Since different bands present different image effects and contain different growth stage information, acquiring crop images across multiple bands ensures sufficient information about the growth stage, facilitating subsequent classification and identification. Then, each band of crop image is divided into multiple sub-image regions. Because the crop images of each band have the same size, the division criteria for the corresponding sub-image regions are the same, resulting in identical sub-image regions for each band, thus each sub-image region contains multiple bands. Next, the multi-vegetation index features of the crop images are extracted to determine the entropy features of each band in each sub-image region. This entropy feature of each band in each sub-image region is then compared with the multi-vegetation index features. A feature space matrix is ​​constructed using indicative features, and samples are obtained by labeling the corresponding growth period types. The pre-built neural network model is then trained in a supervised manner using these samples. The parameters of the neural network model are adjusted based on the differences between the growth period classification results obtained during training and the labeled actual growth period types, improving the accuracy of the adjusted model in growth period classification. Once the training termination condition is met, a growth period classification model is obtained. Because this model incorporates entropy and vegetation index features during training, it achieves more accurate classification of crop growth periods. Finally, for crop images in multiple bands requiring growth period classification, the corresponding feature space matrices for each band are determined. The growth period classification model is then used to classify the growth period using these feature space matrices, outputting the growth period classification result. This growth period classification model combines the influence of entropy and vegetation index features on growth period classification, ensuring accurate and timely classification results.

[0043] In some embodiments, step 102 includes: Step 1021: Perform radiometric correction on the crop images for each band to obtain radiometrically corrected crop images for each band.

[0044] In practice, radiation correction refers to the process of correcting and adjusting the radiation caused by differences in illumination. It can eliminate the impact of differences in illumination on crop images of various wavelengths and improve the stability of the reflectance of crop images of various wavelengths.

[0045] Step 1022: Perform image fusion on the radiometrically corrected crop images for each band to obtain the fused crop image for each band.

[0046] In practice, image fusion aims to match and adjust spatial resolution and spectral information to improve the matching degree of spatial resolution and spectral information.

[0047] Step 1023: Perform image stitching processing on the fused crop images of each band to obtain the processed images of each band.

[0048] In practice, since there are multiple frames of crop images acquired in each band, there are also multiple frames of fused crop images in each band. Therefore, image stitching is to stitch together the fused crop images of a certain band to form a single frame of processed image in that band, thus ensuring the integrity of the processed image of the crop planting area and facilitating subsequent area division.

[0049] Step 1024: For each band-processed image, divide the band-processed image into multiple sub-image regions according to a predetermined size.

[0050] In practice, the processed image of each band is divided according to a predetermined size. Since the processed images of each band have the same size and position, the sub-image regions of each band, after being divided according to the same predetermined size, also have the same size and position. This ensures that the same sub-image region in the resulting multiple sub-image regions corresponds to multiple bands.

[0051] This process involves using masking software to divide the processed image for each band into mask regions of predetermined size, thus creating multiple mask regions, each serving as a sub-image region. The positions and sizes of the corresponding mask regions for each band of the processed image are identical.

[0052] Through the above scheme, after radiometric correction, image fusion and image stitching, it is ensured that the complete processed image of each band is not affected by the difference in illumination, and the matching degree of the corresponding spatial resolution and spectral information is improved, which facilitates the division of regions to obtain multiple sub-image regions.

[0053] Radiometric correction, image fusion, and image stitching can be specifically processed and performed using preprocessing software (such as Pix4Dmapper software).

[0054] In some embodiments, step 1021 includes: Step 10211: Within a predetermined time period, determine the calibration plate image of the standard grayscale calibration plate placed in the crop planting area.

[0055] In practice, radiometric correction is performed using preprocessing software. Within a predetermined timeframe (e.g., one hour) before and after the drone-borne multispectral camera takes images, a standard grayscale calibration board with reflectance grayscale is horizontally placed in the crop planting area. This ensures the calibration board is completely centered within the field of view of the drone-borne multispectral camera, and the shooting angle aligns with the solar altitude angle. This allows for the capture of the calibration board image. After the drone-borne multispectral camera captures crop images in multiple bands, the calibration board image and the crop images in each band are input together into the preprocessing software for further processing.

[0056] Step 10212: Determine the standard reflectance of the standard grayscale calibration plate in each band, and determine the average DN value of the calibration plate image. Divide the standard reflectance of each band by the average DN value to obtain the radiometric calibration coefficient corresponding to each band.

[0057] Step 10213: Perform radiometric correction on the crop image of the corresponding band according to the radiometric calibration coefficient of each band to obtain the radiometrically corrected crop image for each band.

[0058] The above scheme allows for radiometric correction of crop images in each band using a defined radiometric calibration coefficient, eliminating the impact of illumination differences on crop images in each band and improving the stability of reflectivity in crop images in each band.

[0059] In some embodiments, step 1022 includes: Step 10221: Spatial registration is performed on the radiometrically corrected crop images of each band to obtain spatially registered crop images of each band. Step 10222: Determine the resolution of the spatially registered crop image for each band, and combine and match the low-resolution brightness information with a resolution less than or equal to the resolution threshold and the high-resolution brightness information with a resolution greater than the resolution threshold to obtain the fused crop image for each band.

[0060] The above scheme incorporates a multispectral fusion algorithm into the preprocessing software. This algorithm is used to spatially register crop images after radiometric correction in different bands. Then, the low-resolution and high-resolution brightness information are combined to ensure the spectral accuracy of the fused crop images in each band and to improve the spatial resolution.

[0061] In some embodiments, step 1023 includes: For each band: determine that there are multiple fused crop images in the band, align the fused crop images in time and space, complete the image stitching process, and obtain the processed image for that band.

[0062] In practice, preprocessing software is used to perform temporal and spatial metadata alignment on the multi-frame fused crop images of each band. Specifically, a motion structure algorithm is used to match corresponding points in different frames of the fused crop images under that band, ensuring that each frame of the fused crop image has a predetermined number of corresponding points for matching. This is used to calculate relative pose parameters, and the fused crop images of each frame under that band are then unified into a unified coordinate system to complete the stitching process. This results in a processed image (i.e., orthophoto) corresponding to each band that can completely cover the entire crop planting area.

[0063] The above method ensures the uniformity and integrity of the processed images of the crop planting area, which facilitates subsequent region division.

[0064] In some embodiments, step 103 extracts the entropy features of each band corresponding to each sub-image region, including: Step 1031: For each sub-image region, the entropy value of the sub-image region is determined using the information entropy algorithm.

[0065] For example, after processing each band, the image is reconstructed to 150×400 pixels, and then divided into non-overlapping sub-image regions of 10×10 pixels, resulting in 600 sub-image regions. This is based on the formula corresponding to the information entropy algorithm. Where H is the entropy value, The probability of grayscale value x appearing in the sub-image region (e.g., obtained by dividing the number of times the corresponding grayscale value x appears, n, by the total number of pixels in the non-overlapping sub-image region).

[0066] Step 1032: For each band, perform matrix processing on the entropy values ​​of each sub-image region of the band to obtain the entropy value features of the band.

[0067] For example, if there are 5 bands, and each band has 600 entropy values, then arranging these 600 entropy values ​​in a matrix will give the entropy features corresponding to the 5 bands.

[0068] The above scheme enables entropy features to better characterize the pixel features of the corresponding region, facilitating the identification and judgment of the reproductive period type.

[0069] In some embodiments, step 103 extracts multiple vegetation index features from the crop image, including: Step 1033: Determine the spectral reflectance corresponding to the crop image for each band, and determine multiple vegetation index features based on the spectral reflectance using a vegetation index feature algorithm.

[0070] In practice, vegetation index characteristics include at least one of the following: NDVI, NDRE, GNDVI, SR, EVI, GLI, RECI, MTCI, SIPI, and MSR.

[0071] The spectral reflectance of crop images in different spectral bands is determined, based on the spectral reflectance determined after the above radiometric correction.

[0072] For example, determining the reflectance of the blue light band in this single sub-image region. Reflectivity in the green light band Reflectivity in the red light band Reflectivity of the red-edge band Reflectivity in the near-infrared band .

[0073] Thus, the formulas for calculating the various vegetation index features of this single sub-image region are as follows: ; ; ; ; ; ; ; ; ; .

[0074] The above method yields vegetation index features for each sub-image region. These vegetation index features can characterize crop growth status and are of great help in identifying and judging growth stage types.

[0075] In some embodiments, step 104 includes: Step 1041: Divide a predetermined number of samples into a training set and a test set according to a predetermined ratio.

[0076] Step 1042: Perform linear compression using the feature space matrix of the samples in the training set to obtain spatial data of a predetermined dimension.

[0077] Step 1043: Pre-construct a neural network model including an input layer, a hidden layer, and an output layer, wherein the number of nodes in the input layer matches the predetermined dimension.

[0078] Step 1044: The spatial data of the predetermined dimension is input into the input layer through the input layer node, and the input layer sends the spatial data of the predetermined dimension to the hidden layer.

[0079] Step 1045: Use the hidden layer to extract features from the spatial data of the predetermined dimension, identify the reproductive period using the extracted features, and send the identification results of each reproductive period to the output layer.

[0080] Step 1046: The output layer integrates the identification results of each reproductive stage to determine the reproductive stage classification result corresponding to the sample in the training set and outputs it.

[0081] Step 1047: Compare the fertility period classification result with the actual fertility period type corresponding to the sample, determine the difference value, adjust the parameters of each layer of the neural network model according to the difference value, obtain the adjusted neural network model, and complete the training process of a sample in the training set.

[0082] Step 1048: Process the samples in the test set using the adjusted neural network model to obtain the verification output results. Determine the accuracy of the verification output results based on the verification output results and the actual fertility type of the samples labeled in the corresponding test set.

[0083] Step 1049: In response to the accuracy being less than the accuracy threshold, the adjusted neural network model is further trained using the next sample in the training set until the determined accuracy is greater than or equal to the accuracy threshold, and the final adjusted neural network model is used as the reproductive period classification model.

[0084] Alternatively, in step 10410, in response to the accuracy being greater than or equal to the accuracy threshold, the final adjusted neural network model is used as the reproductive period classification model.

[0085] In practice, during the testing of the adjusted neural network model using the test set, a classification accuracy threshold is set using the overall classification accuracy model, and / or an F1 score threshold is set using the F1 score evaluation model. Specifically, the overall classification accuracy obtained by testing the adjusted neural network model on the test set samples is determined using the overall classification accuracy model, and / or the F1 score obtained by testing the adjusted neural network model on the test set samples is determined using the F1 score evaluation model. When the overall classification accuracy is greater than or equal to the classification accuracy threshold, and / or the F1 score is greater than or equal to the F1 score threshold, the model construction is considered complete, and the final adjusted neural network model is used as the reproductive period classification model.

[0086] Accuracy includes: overall classification precision and / or F1 score. Accuracy thresholds include: classification precision threshold and / or F1 score threshold.

[0087] In addition, the EVA-Net neural network model uses the Adam optimizer and mean squared error loss function, and an early stopping strategy is set to prevent overfitting.

[0088] By using the above scheme, training and testing samples combining entropy features and vegetation index features to train and test the neural network model, it is possible to ensure that the neural network model continuously improves its correlation between the feature space matrix obtained by combining entropy features and vegetation index features and the corresponding growth period type. In this way, the growth period classification model obtained after training is more accurate in classifying and identifying crop growth periods.

[0089] In some embodiments, step 105 includes: Step 1051: Collect images of the crop to be classified in multiple bands using a multispectral camera.

[0090] In practice, drones carrying multispectral cameras can be used to capture images of crops at different growth stages in multiple wavelengths. These multiple wavelengths include blue, green, red, red-edge, and near-infrared bands.

[0091] Step 1052: After performing radiometric correction, image fusion, and image stitching on the crop images to be classified in the multiple bands, preprocessed images to be classified in the multiple bands are obtained.

[0092] In practice, the process of radiometric correction, image fusion, and image stitching of the crop images to be classified is the same as the process of radiometric correction, image fusion, and image stitching of the crop images described above, resulting in preprocessed images to be classified in multiple bands.

[0093] Step 1053: Divide the preprocessed image to be classified in each band into multiple sub-image regions to be classified, and each sub-image region to be classified corresponds to multiple bands.

[0094] In practice, to facilitate identification, the preprocessed image to be classified needs to be divided into regions to obtain multiple sub-image regions to be classified. The region division criteria for the preprocessed image to be classified in each band are the same, so that each sub-image region to be classified corresponds to multiple bands, that is, the size and position of the sub-image to be classified in multiple bands are the same.

[0095] Step 1054: Determine multiple vegetation index features corresponding to crop images to be classified in multiple bands, extract the entropy value features of each band corresponding to each sub-image region to be classified, and construct a feature space matrix to be classified by combining the entropy value features of each band of each sub-image region to be classified with the multiple vegetation index features to be classified.

[0096] In practice, the entropy value of each sub-image region to be classified is calculated according to the formula of the information entropy algorithm corresponding to step 1031 above. The entropy values ​​of each sub-image region to be classified are arranged in a matrix according to their corresponding positions to obtain the entropy value features to be classified. The above operation is repeated for each band to obtain the entropy value features to be classified for each band.

[0097] According to the vegetation index feature calculation formula in step 1033 above, calculate the various vegetation index features corresponding to the crop image to be classified.

[0098] Finally, the entropy features of each band to be classified and various vegetation index features to be classified are combined to construct a feature space matrix to be classified. This feature space matrix can then be input into a trained growth period classification model. The growth period classification model is used to analyze the entropy features and vegetation index features to be classified in the feature space matrix to determine and output an accurate growth period classification result.

[0099] In summary, multi-band crop images were acquired using a multispectral camera mounted on a drone. Radiometric correction, image fusion, and image stitching were then applied to process these images of the crop-growing area. After subdividing the image into sub-regions, the entropy features of each band and multiple vegetation index features were integrated to obtain a feature space matrix. After labeling the corresponding actual growth stage types, this matrix served as samples for training a neural network model. This process allowed the neural network model to continuously improve its ability to identify the correlation between the feature space matrix obtained from the combination of entropy features and vegetation index features and the corresponding growth stage type. Compared to traditional machine learning methods, crop type identification based on growth stage exhibits higher accuracy and stronger generalizability, demonstrating significant practical value.

[0100] The following section describes in detail the crop growth period classification method of the crop images in this application, using wheat as an example.

[0101] Includes the following steps: S1. Data Acquisition: High-resolution, multi-band crop images of the crop-growing area (e.g., the experimental area) were acquired using a drone equipped with a multispectral camera. Ground-based growth stage surveys were conducted simultaneously with drone sampling. When acquiring high-resolution, multi-band crop images using drones, the sampling was conducted between 12:00 PM and 2:00 PM on a clear, cloudless day to obtain stable and clear crop images. The multiple bands included five channels: blue light, green light, red light, red-edge, and near-infrared. The actual growth stage types sampled on the ground included: booting stage, heading stage, flowering stage, and maturity stage.

[0102] S2. Preprocessing and Correction: Pix4D mapper 4.8.6 software is used to process crop images across multiple bands for preprocessing and correction. The specific workflow is as follows: Radiometric Correction: Before the drone flight, a standard grayscale calibration board is placed in the center of the crop planting area. The drone, equipped with a multispectral camera, hovers 1 meter vertically to take pictures. The calibration board image and the crop images of each band are input into Pix4D mapper 4.8.6 preprocessing software. In the "Processing Options" module of the preprocessing software, "Radiometric Correction" is selected, and the standard reflectance of the standard grayscale calibration board is input. The preprocessing software automatically calculates the radiometric calibration coefficients for each band. Based on the radiometric calibration coefficients of each band, radiometric correction is performed on the crop images of the corresponding bands, and then the DN values ​​are converted into absolute reflectance to obtain the radiometrically corrected crop images for each band.

[0103] Image fusion: Spatially register the radiometrically corrected crop images of each band to obtain spatially registered crop images of each band. Determine the resolution of the spatially registered crop images of each band. Combine and match the low-resolution brightness information with a resolution less than or equal to the resolution threshold and the high-resolution brightness information with a resolution greater than the resolution threshold to obtain the fused crop images of each band.

[0104] Image stitching: The preprocessing software imports GPS and IMU data from the UAV and uses a motion structure algorithm to match corresponding points in the fused crop images of different frames within the same band. This ensures that each frame of the fused crop image has a predetermined number (e.g., 68) of corresponding points for matching, calculating relative attitude parameters, and unifying the fused crop images of each frame within the same band into a unified coordinate system to complete the stitching process. This results in a processed image (i.e., an orthophoto) corresponding to each band that can completely cover the entire crop planting area.

[0105] S3. Manually divide the sub-image regions into corresponding cell masks. Open the processed image in QGIS and draw a predetermined number (640) cell masks (e.g., each cell mask is 1.2m × 5m in size) according to the actual cell boundaries of the crop planting area. Extract each cell mask from the multi-band processed image of 5 channels and label it with the corresponding real growth period type. Divide the processed image into multiple sub-image regions, extract the sub-image regions corresponding to the multi-band of 5 channels, and label the real growth period type of the sub-image region on the sampling date as a sample to establish a sample set. Divide the sample set into a training set and a test set in an 8:2 ratio.

[0106] S4. Extract the entropy features of each sub-image region and calculate the 10 vegetation index features of the crop image to construct the feature space matrix.

[0107] The crop images of each of the five bands were reconstructed to 150×400 pixels, and then divided into non-overlapping sub-image regions of 10×10 pixels, resulting in 600 sub-image regions. For each sub-image region, an entropy value was determined using the information entropy algorithm and corresponding formula. 600 entropy values ​​were obtained for each band, for a total of 3000 entropy values. The entropy value matrix of each band was then arranged to obtain the entropy value features corresponding to the five bands. Additionally, 10 vegetation index features (NDVI, NDRE, GNDVI, SR, EVI, GLI, RECI, MTCI, SIPI, MSR) were calculated. The entropy value features of the five bands and the 10 vegetation index features were integrated to obtain the feature space matrix.

[0108] S5. Construct and validate a classification model for key growth stages of wheat.

[0109] Construct a neural network model (e.g., the EVA-Net neural network model) that includes an input layer, three hidden layers, and an output layer, using the Adam optimizer and mean squared error loss function, and setting an early stopping strategy to prevent overfitting.

[0110] A training set is introduced to train the neural network model. A test set is then introduced into the trained and adjusted neural network model for testing. A classification accuracy threshold is set using the overall classification accuracy model, and / or an F1 score threshold is set to evaluate the model. Specifically, the overall classification accuracy of the adjusted neural network model after testing on the test set samples is determined using the overall classification accuracy model, and / or the F1 score is used to evaluate the model and determine the F1 score of the adjusted neural network model after testing on the test set samples. When the overall classification accuracy is greater than or equal to the classification accuracy threshold, and / or the F1 score is greater than or equal to the F1 score threshold, the model construction is considered complete, and the final adjusted neural network model is used as the reproductive period classification model.

[0111] Four models were selected for validation experiments: the growth period classification model obtained from the EVA-Net neural network of this application, the SVM model, the LR model, and the DT model. Among the four models, the growth period classification model obtained from the EVA-Net neural network performed best in crop growth period classification and identification, with both accuracy and macro-average F1 score reaching 0.95. The accuracy and F1 score of the SVM model and the LR model were 0.92 and 0.90, respectively, while the DT model performed the worst, with both accuracy and F1 score being 0.81.

[0112] Therefore, the fertility period classification model obtained in this application can accurately classify and identify the fertility period of crops, with a high degree of accuracy.

[0113] It should be noted that the method in this embodiment can be executed by a single device, such as a computer or server. The method can also be applied in a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method in this embodiment, and the multiple devices will interact with each other to complete the method described.

[0114] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0115] Based on the same inventive concept, corresponding to any of the above embodiments, this application also provides a crop image growth period classification device.

[0116] refer to Figure 3 The device includes: The acquisition and determination module 201 is configured to acquire crop images of multiple bands in the crop planting area using a multispectral camera and determine the true growth period type of the crop planting area. The region division module 202 is configured to divide the crop image of each band into multiple sub-image regions, wherein the sub-image regions of the crop image of each band are the same, and each sub-image region corresponds to multiple bands; The feature determination module 203 is configured to extract multiple vegetation index features of crop images, extract the entropy value features of each band corresponding to each sub-image region, construct a feature space matrix with the entropy value features of each band of each sub-image region and multiple vegetation index features, and label the corresponding real growth period type of the feature space matrix as a sample. The model training module 204 is configured to use the samples to perform supervised training on a pre-built neural network model, compare the fertility period classification results output by the training with the labeled real fertility period types, adjust the parameters of the neural network model based on the differences in the comparison, obtain the adjusted neural network model, determine that the training termination condition is met, and use the finally adjusted neural network model as the fertility period classification model. The data to be classified determination module 205 is configured to use multi-band crop images to be classified acquired by a multispectral camera to determine the feature space matrix corresponding to the crop images to be classified in multiple bands. The reproductive period classification processing module 206 is configured to input the feature space matrix to be classified into the reproductive period classification model to obtain the reproductive period classification result output.

[0117] In some embodiments, the region division module 202 includes: The correction unit is configured to perform radiometric correction on the crop image for each band separately, so as to obtain a radiometrically corrected crop image for each band. The fusion unit is configured to perform image fusion on the radiometrically corrected crop images for each band to obtain a fused crop image for each band. The stitching unit is configured to perform image stitching processing on the fused crop images of each band separately to obtain the processed image of each band; The segmentation unit is configured to divide the processed image of each band into multiple sub-image regions according to a predetermined size.

[0118] In some embodiments, the correction unit is specifically configured as follows: Within a predetermined time period, determine the calibration board image of the standard grayscale calibration board placed in the crop planting area; Determine the standard reflectance of the standard grayscale calibration plate in each band, and determine the average DN value of the calibration plate image. Divide the standard reflectance of each band by the average DN value to obtain the radiometric calibration coefficient corresponding to each band. Radiometric correction is performed on the crop images of the corresponding bands based on the radiometric calibration coefficients for each band, resulting in radiometrically corrected crop images for each band.

[0119] In some embodiments, the fusion unit is specifically configured as follows: Spatial registration is performed on the radiometrically corrected crop images of each band to obtain spatially registered crop images of each band. The resolution of the spatially registered crop image for each band is determined. The low-resolution brightness information with a resolution less than or equal to the resolution threshold and the high-resolution brightness information with a resolution greater than the resolution threshold are combined and matched to obtain the fused crop image for each band.

[0120] In some embodiments, the splicing unit is specifically configured as follows: For each band: determine that there are multiple fused crop images in the band, align the fused crop images in time and space, complete the image stitching process, and obtain the processed image for that band.

[0121] In some embodiments, the feature determination module 203 is specifically configured to: For each sub-image region, the entropy value of that sub-image region is determined using the information entropy algorithm. For each band, the entropy values ​​of each sub-image region of that band are processed into a matrix to obtain the entropy features of that band.

[0122] In some embodiments, the feature determination module 203 is configured to: The spectral reflectance of crop images in each band is determined, and multiple vegetation index features are determined based on the spectral reflectance using a vegetation index feature algorithm.

[0123] In some embodiments, the model training module 204 is specifically configured as follows: The predetermined number of samples are divided into a training set and a test set according to a predetermined ratio; Linear compression is performed using the feature space matrix of the samples in the training set to obtain spatial data of a predetermined dimension. A neural network model comprising an input layer, hidden layers, and an output layer is pre-constructed, wherein the number of nodes in the input layer matches the predetermined dimension; The spatial data of the predetermined dimension is input into the input layer through the input layer node, and the input layer sends the spatial data of the predetermined dimension to the hidden layer. The hidden layer is used to extract features from the spatial data of the predetermined dimension, and the extracted features are used to identify the reproductive period. The hidden layer sends the identification results of each reproductive period to the output layer. The output layer integrates the identification results of each reproductive stage to determine the reproductive stage classification result corresponding to the sample in the training set and outputs it. The classification result of the reproductive period is compared with the actual reproductive period type corresponding to the sample to determine the difference value. The parameters of each layer of the neural network model are adjusted according to the difference value to obtain the adjusted neural network model, thus completing the training process of a sample in the training set. The samples in the test set are processed using the adjusted neural network model to obtain the verification output results. The accuracy of the verification output results is determined based on the verification output results and the actual fertility type of the samples in the corresponding test set. In response to the accuracy being less than the accuracy threshold, the adjusted neural network model is trained again using the next sample in the training set until the determined accuracy is greater than or equal to the accuracy threshold. The final adjusted neural network model is then used as the reproductive period classification model. Alternatively, in response to the accuracy being greater than or equal to the accuracy threshold, the final adjusted neural network model is used as the reproductive period classification model.

[0124] In some embodiments, the data to be classified determination module 205 is specifically configured as follows: Images of crops to be classified in multiple bands acquired using a multispectral camera; After performing radiometric correction, image fusion, and image stitching on the crop images to be classified in the multiple bands, preprocessed images to be classified in multiple bands are obtained. The preprocessed image to be classified for each band is divided into multiple sub-image regions to be classified, and each sub-image region to be classified corresponds to multiple bands; Multiple vegetation index features corresponding to crop images to be classified in multiple bands are determined. The entropy value features of each band corresponding to each sub-image region to be classified are extracted. The entropy value features of each band of each sub-image region to be classified are combined with the multiple vegetation index features to be classified to construct a feature space matrix to be classified.

[0125] For ease of description, the above devices are described in terms of function, divided into various modules. Of course, in implementing this application, the functions of each module can be implemented in one or more software and / or hardware.

[0126] The apparatus of the above embodiments is used to implement the corresponding method in any of the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0127] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the methods described in any of the above embodiments.

[0128] Figure 4This embodiment illustrates a more specific hardware structure of an electronic device. The device may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0129] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0130] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0131] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0132] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0133] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0134] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0135] The electronic devices described above are used to implement the corresponding methods in any of the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0136] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to perform the methods described in any of the above embodiments.

[0137] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital video disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0138] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to perform the methods described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0139] Based on the same concept, corresponding to any of the above embodiments, this application also provides a computer program product, including computer program instructions, which, when run on a computer, cause the computer to perform the method described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0140] It is understood that before using the technical solutions of the various embodiments in this application, users will be informed of the type, scope of use, and usage scenarios of the personal information involved in an appropriate manner, and user authorization will be obtained.

[0141] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose, based on the prompt message, whether to provide personal information to the software or hardware such as electronic devices, applications, servers, or storage media performing the operations described in this application.

[0142] As an optional but not limited implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0143] It is understood that the above notification and user authorization process is merely illustrative and does not limit the implementation of this application. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this application.

[0144] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0145] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0146] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of this application are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. A method for classifying the growth stage of crop images, characterized in that, include: Multispectral cameras are used to acquire crop images in multiple bands of the crop planting area, and the true growth stage type of the crop planting area is determined. Each crop image in each band is divided into multiple sub-image regions. The sub-image regions are the same for each crop image in each band, and each sub-image region corresponds to multiple bands. Extract multiple vegetation index features from crop images, extract the entropy value features of each band corresponding to each sub-image region, construct a feature space matrix with the entropy value features of each band of each sub-image region and multiple vegetation index features, and label the corresponding real growth period type of the feature space matrix as a sample. Using the samples, supervised training is performed on the pre-constructed neural network model. The classification results of the reproductive period output by the training are compared with the labeled real reproductive period types. The parameters of the neural network model are adjusted according to the differences in the comparison to obtain the adjusted neural network model. Once the training termination condition is met, the final adjusted neural network model is used as the reproductive period classification model. By using multispectral images of crops to be classified in multiple bands acquired by a multispectral camera, the feature space matrix corresponding to the crop images to be classified in multiple bands is determined. The feature space matrix to be classified is input into the reproductive period classification model to obtain the reproductive period classification result output.

2. The method according to claim 1, characterized in that, The process of dividing the crop image in each band into multiple sub-image regions includes: Radiometric correction was performed on the crop images for each band to obtain radiometrically corrected crop images for each band. Image fusion was performed on the radiometrically corrected crop images for each band to obtain a fused crop image for each band. The fused crop images for each band are then stitched together to obtain the processed images for each band. For each band of processed image, the processed image of that band is divided into multiple sub-image regions according to a predetermined size.

3. The method according to claim 2, characterized in that, The process of performing radiometric correction on crop images for each band to obtain radiometrically corrected crop images for each band includes: Within a predetermined time period, determine the calibration board image of the standard grayscale calibration board placed in the crop planting area; Determine the standard reflectance of the standard grayscale calibration plate in each band, and determine the average DN value of the calibration plate image. Divide the standard reflectance of each band by the average DN value to obtain the radiometric calibration coefficient corresponding to each band. Radiometric correction is performed on the crop images of the corresponding bands based on the radiometric calibration coefficients for each band, resulting in radiometrically corrected crop images for each band.

4. The method according to claim 2, characterized in that, The radiometrically corrected crop images for each band are then fused to obtain a fused crop image for each band, including: Spatial registration is performed on the radiometrically corrected crop images of each band to obtain spatially registered crop images of each band. The resolution of the spatially registered crop image for each band is determined. The low-resolution brightness information with a resolution less than or equal to the resolution threshold and the high-resolution brightness information with a resolution greater than the resolution threshold are combined and matched to obtain the fused crop image for each band.

5. The method according to claim 2, characterized in that, The fused crop images for each band are then stitched together to obtain processed images for each band, including: For each band: Once it is determined that there are multiple fused crop images in this band, the fused crop images are aligned in time and space to complete the image stitching process, and the processed image of this band is obtained.

6. The method according to claim 1, characterized in that, The extraction of entropy features for each band corresponding to each sub-image region includes: For each sub-image region, the entropy value of that sub-image region is determined using the information entropy algorithm. For each band, the entropy values ​​of each sub-image region of that band are processed into a matrix to obtain the entropy features of that band.

7. The method according to claim 1, characterized in that, The extraction of multiple vegetation index features from crop images includes: The spectral reflectance of crop images in each band is determined, and multiple vegetation index features are determined based on the spectral reflectance using a vegetation index feature algorithm.

8. The method according to claim 1, characterized in that, The process involves using the samples to perform supervised training on a pre-constructed neural network model, comparing the training output's fertility period classification results with the labeled actual fertility period types, adjusting the parameters of the neural network model based on the differences in the comparison, obtaining an adjusted neural network model, determining that the training termination condition is met, and using the final adjusted neural network model as the fertility period classification model, including: The predetermined number of samples are divided into a training set and a test set according to a predetermined ratio; Linear compression is performed using the feature space matrix of the samples in the training set to obtain spatial data of a predetermined dimension. A neural network model comprising an input layer, hidden layers, and an output layer is pre-constructed, wherein the number of nodes in the input layer matches the predetermined dimension; The spatial data of the predetermined dimension is input into the input layer through the input layer node, and the input layer sends the spatial data of the predetermined dimension to the hidden layer. The hidden layer is used to extract features from the spatial data of the predetermined dimension, and the extracted features are used to identify the reproductive period. The hidden layer sends the identification results of each reproductive period to the output layer. The output layer integrates the identification results of each reproductive stage to determine the reproductive stage classification result corresponding to the sample in the training set and outputs it. The classification result of the reproductive period is compared with the actual reproductive period type corresponding to the sample to determine the difference value. The parameters of each layer of the neural network model are adjusted according to the difference value to obtain the adjusted neural network model, thus completing the training process of a sample in the training set. The samples in the test set are processed using the adjusted neural network model to obtain the verification output results. The accuracy of the verification output results is determined based on the verification output results and the actual fertility type of the samples in the corresponding test set. In response to an accuracy rate less than an accuracy threshold, the adjusted neural network model is trained again using the next sample in the training set until the determined accuracy rate is greater than or equal to the accuracy threshold. The final adjusted neural network model is then used as the reproductive age classification model; or... In response to the accuracy being greater than or equal to the accuracy threshold, the final adjusted neural network model is used as the reproductive period classification model.

9. The method according to claim 1, characterized in that, The process of determining the classification feature space matrix corresponding to the crop images to be classified in multiple bands acquired by a multispectral camera includes: Images of crops to be classified in multiple bands acquired using a multispectral camera; After performing radiometric correction, image fusion, and image stitching on the crop images to be classified in the multiple bands, preprocessed images to be classified in multiple bands are obtained. The preprocessed image to be classified for each band is divided into multiple sub-image regions to be classified, and each sub-image region to be classified corresponds to multiple bands; Multiple vegetation index features corresponding to crop images to be classified in multiple bands are determined. The entropy value features of each band corresponding to each sub-image region to be classified are extracted. The entropy value features of each band of each sub-image region to be classified are combined with the multiple vegetation index features to be classified to construct a feature space matrix to be classified.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 9.

Images