Crop disease identification method, device, equipment and medium

Abstract

The application provides a crop disease identification method, device, equipment and medium, the method comprises: acquiring the RGB image and the NIR image of the target part on the crop under different shooting angles; determining the corresponding RGB weight and NIR weight according to the RGB image and the NIR image under each shooting angle; determining the disease area belonging to the target disease on the image under each shooting angle according to the RGB image and the NIR image under each shooting angle and the corresponding RGB weight and NIR weight, realizing the rapid identification of diseases at different angles of crops, and further monitoring the disease conditions of different growth parts of crops, and providing support for crop disease information acquisition and disease control.

Description

Technical Field

[0001] This invention relates to the field of crop disease prediction technology, and in particular to a method, apparatus, equipment and medium for identifying crop diseases. Background Technology

[0002] Crop diseases not only severely impact crop yields but also reduce their quality. Traditional field diagnosis of diseases is done manually, which suffers from limitations such as fatigue and significant subjective errors, making timely and accurate disease localization difficult. However, with the continuous development of artificial intelligence and deep learning technologies, the automated identification and diagnosis of crop disease images using these technologies has become a major research focus both domestically and internationally in recent years. Deep learning technology, with its powerful feature extraction and self-learning capabilities, can significantly improve the accuracy and efficiency of disease detection; its strong adaptability supports complex environmental changes in the field; and its excellent generalization ability allows for the identification and classification of new and unseen diseases. Deep learning technology has significant advantages in crop disease detection, helping to reduce agricultural losses and improve agricultural production efficiency.

[0003] While existing deep learning technologies can identify and detect crop diseases with good accuracy, they still suffer from problems such as unclear lesion characteristics at the location of collected lesions, overly simplistic detection, and low detection quality.

[0004] Invention Content

[0005] To address the problems existing in the prior art, the present invention provides a method, apparatus, equipment and medium for identifying crop diseases.

[0006] This invention provides a method for identifying crop diseases, comprising:

[0007] Acquire RGB and NIR images of target parts of crops from different shooting angles;

[0008] Based on the RGB image and the NIR image at each shooting angle, determine the corresponding RGB weights and NIR weights;

[0009] Based on the RGB and NIR images at each shooting angle, and the corresponding RGB and NIR weights, the disease area belonging to the target disease on the image at each shooting angle is determined.

[0010] According to a method for identifying crop diseases provided by the present invention, the method further includes:

[0011] The target disease is identified as corresponding to the target channel in the RGB channel.

[0012] Based on the RGB image and the NIR image at each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area;

[0013] Based on the target channel color value and the NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, and the comprehensive difference value of the disease area on the image at all shooting angles.

[0014] According to a method for identifying crop diseases provided by the present invention, determining the corresponding RGB weights and NIR weights based on the RGB image and the NIR image at each shooting angle includes:

[0015] Obtain the RGB image features of the RGB image;

[0016] Obtain the NIR image features of the NIR image;

[0017] The RGB image features and the NIR image features are fused to obtain the fused features;

[0018] Based on the fusion features, RGB weights and NIR weights are determined, wherein the sum of the RGB weights and NIR weights is 1.

[0019] According to a crop disease identification method provided by the present invention, the step of determining the disease region belonging to the target disease on the image at each shooting angle based on the RGB image and the NIR image at each shooting angle, and the corresponding RGB weights and NIR weights, includes:

[0020] Obtain the RGB image features of the RGB image;

[0021] Obtain the NIR image features of the NIR image;

[0022] The RGB image features and the NIR image features are fused to obtain the fused features;

[0023] Based on the fusion features, RGB disease regions and NIR disease regions on the image are determined, as well as the corresponding RGB disease prediction values ​​and NIR disease prediction values.

[0024] Based on the RGB and NIR disease regions in the image, the disease regions in the image are determined by combining the RGB weights and NIR weights respectively.

[0025] Based on the RGB disease prediction values ​​and NIR disease prediction values, the predicted values ​​of disease areas on the image are determined by combining the RGB weights and NIR weights respectively.

[0026] Based on the diseased areas in the image and the predicted values, the diseased areas in the image belonging to the target disease are selected.

[0027] According to a method for identifying crop diseases provided by the present invention, determining a comprehensive index value of the diseased area on the image at each shooting angle based on the target channel color value and the NIR color value includes:

[0028] Based on the target channel color value and the NIR color value, the comprehensive index value of the disease area on the image at each shooting angle is determined using the first calculation formula;

[0029] The first calculation formula includes:

[0030]

[0031] IV = λ1·IV j1 +…+λ m IV jm +…+(1-λ1-…-λ m -…-λ n-1 IV jn

[0032] Where IV is the composite index value, IV jm It is the index value at the m-th shooting angle, NIR. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle, k m It is the adjustment parameter for the m-th shooting angle, λ m It is the angle weight under the m-th shooting angle, where m is an integer between 1 and n.

[0033] According to a method for identifying crop diseases provided by the present invention, determining the comprehensive difference value of disease areas in images under all shooting angles based on the target channel color value and the NIR color value includes:

[0034] Based on the target channel color value and the NIR color value, the comprehensive difference value of the disease area on the image under all shooting angles is determined by the second calculation formula.

[0035] The second calculation formula includes:

[0036]

[0037] Where DV is the composite difference value, and NIR is the total difference value. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle.

[0038] The present invention also provides a crop disease identification device, comprising:

[0039] The acquisition module is used to acquire RGB and NIR images of target parts of crops from different shooting angles;

[0040] The determination module is used to determine the corresponding RGB weights and NIR weights based on the RGB image and the NIR image at each shooting angle;

[0041] The identification module is used to determine the disease area belonging to the target disease on the image at each shooting angle based on the RGB image and the NIR image at each shooting angle, as well as the corresponding RGB weights and NIR weights.

[0042] According to the present invention, a crop disease identification device further includes a computing module for:

[0043] The target disease is identified as corresponding to the target channel in the RGB channel.

[0044] Based on the RGB image and the NIR image at each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area;

[0045] Based on the target channel color value and the NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, and the comprehensive difference value of the disease area on the image at all shooting angles.

[0046] The present invention 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 program to implement the crop disease identification method as described above.

[0047] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the crop disease identification method as described above.

[0048] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the crop disease identification method as described above.

[0049] This invention provides a method, apparatus, device, and medium for identifying crop diseases. By collecting RGB and NIR images of target parts of crops from different shooting angles, the corresponding RGB and NIR weights are determined. Based on the RGB and NIR images from each shooting angle, and the corresponding RGB and NIR weights, the diseased area belonging to the target disease in the image at each shooting angle is determined. This enables rapid identification of diseases in crops from different angles simultaneously, thereby monitoring the disease status of different growth parts of crops and providing support for crop disease information acquisition and disease prevention and control. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the crop disease identification method provided by the present invention;

[0052] Figure 2 This is a schematic diagram of the structure of the crop image acquisition device provided by the present invention;

[0053] Figure 3 This is a schematic diagram illustrating the use of the crop image acquisition device provided by the present invention;

[0054] Figure 4 This is a diagram of the architecture of the dual-stream deep fusion neural network model provided by the present invention;

[0055] Figure 5 This is a schematic diagram of the structure of the crop disease identification device provided by the present invention;

[0056] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] The following is combined with Figures 1-6This invention describes the crop disease identification method, apparatus, equipment, and medium.

[0059] Figure 1 This diagram illustrates a flowchart of a crop disease identification method provided by the present invention. (See attached diagram.) Figure 1 The method includes:

[0060] 11. Acquire RGB and NIR images of target parts of crops from different shooting angles;

[0061] 12. Determine the corresponding RGB weights and NIR weights based on the RGB and NIR images from each shooting angle;

[0062] 13. Based on the RGB and NIR images at each shooting angle, and the corresponding RGB and NIR weights, determine the disease area belonging to the target disease on the image at each shooting angle.

[0063] Regarding steps 11-13 above, it should be noted that in this invention, when a crop (such as wheat, corn, etc.) suffers from a certain disease, areas matching that disease will exist throughout the entire crop plant, such as the stems, leaves, flowers, and fruits. However, the growth characteristics of crops can make it difficult to observe areas exhibiting disease characteristics from certain angles. For example, on leaves growing southward, disease characteristics are more obvious on the side facing the ground. If disease analysis techniques are used on crop images, it is difficult to analyze whether the leaves of the crop have a disease when the image is taken with the leaves facing upwards. Therefore, it is necessary to collect crop images from multiple shooting angles. In addition, it should be noted that the crop images collected by the shooting device are RGB images and NIR images. Obtaining multiple image features from multiple image types increases the recognition and fusion of disease characteristics.

[0064] In this invention, the height and density of crop plants vary depending on their growth stage. The completeness of the photographic coverage also varies. Sometimes, different parts of the crop need to be photographed to determine the disease. For example, one part of the photographic equipment may photograph the upper part of the crop, while another part may photograph the lower part.

[0065] In this invention, the device for acquiring crop images is as follows: Figure 2As shown, the device consists of a rotating shaft, a rotating arm, and an intermediate connecting component. An RGB+NIR camera is mounted on both the rotating shaft and the rotating arm to collect paired disease image data, enabling the acquisition of RGB and NIR images from different shooting angles. The rotating arm and the intermediate connecting component are driven by a motor, allowing adjustment of the angle between them (90 degrees to 180 degrees) to collect disease data from different parts of the crop. The rotating shaft and the intermediate connecting component are also driven by a motor, supporting 360-degree rotation for disease data collection from different locations. The rotating shaft and rotating arm can be easily operated with a simple button press.

[0066] A schematic diagram of the device for acquiring crop images is shown below. Figure 3 As shown, this device can quickly acquire images of crop diseases within this small cluster area. During acquisition, the position of the fixed intermediate component remains unchanged, and the initial setting sets the angle r between the intermediate component and the rotation axis to be... low The angle r between the intermediate component and the rotating arm is 0 degrees. up It is 120 degrees. Adjust different angles r low This allows for the collection of data from the lower leaves of crops, adjusting different angles (r). up The method enables the collection of data from the upper leaves of crops, that is, to obtain disease image data from different angles along the vertical growth direction within the small cluster area, which can be used for subsequent disease stress analysis.

[0067] In this invention, after obtaining RGB and NIR images of the target part of the crop from different shooting angles, a trained two-stream deep fusion neural network model can be used to analyze the images. During the analysis, based on the RGB and NIR images from each shooting angle, the corresponding RGB and NIR weights are determined. Based on the RGB and NIR images from each shooting angle, and the corresponding RGB and NIR weights, the disease area belonging to the target disease on the image at each shooting angle is determined.

[0068] Specifically, this invention designs a dual-stream deep fusion neural network as a disease detection model based on the confidence-based fusion model. It compensates for the arbitrariness of subjectively setting fusion weights by adaptively generating fusion weights using a fully connected neural network, thereby improving the overall performance of disease detection.

[0069] like Figure 4The diagram shows the architecture of a two-stream deep fusion neural network model, primarily composed of two integrated processing modules: a two-stream weighted fully connected neural network (DWCNN) and a two-stream deep convolutional neural network (DDCNN). The DWCNN calculates the weights for both the RGB and NIR images. The DDCNN uses two separate classification subnetworks to extract features from the two image data streams respectively, outputting the classification score (CLS) and bounding box (Bbox) for each subnetwork. These scores are then recombined and fused using the calculated weights to produce the final detection result. The bounding box represents the diseased area of ​​the crop, and the classification score represents the probability that the diseased area within the bounding box represents a specific disease.

[0070] The crop disease identification method provided by this invention determines the corresponding RGB weights and NIR weights by collecting RGB and NIR images of target parts of crops at different shooting angles. Based on the RGB and NIR images at each shooting angle, as well as the corresponding RGB and NIR weights, the disease area belonging to the target disease on the image at each shooting angle is determined. This enables rapid identification of diseases on crops from different angles simultaneously, thereby monitoring the disease status of different growth parts of crops and providing support for crop disease information acquisition and disease prevention and control.

[0071] A further method described above mainly considers the disease stress at different growth sites and compares the differences in disease stress at these different sites to comprehensively characterize the disease stress experienced by crops in the region.

[0072] The target disease is identified as corresponding to the target channel in the RGB channel.

[0073] Based on the RGB and NIR images from each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area;

[0074] Based on the target channel color value and NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, as well as the comprehensive difference value of the disease area on the image at all shooting angles.

[0075] It should be noted that in this invention, different crop diseases have different sensitive wavelengths that are reflected in the image. For example, the sensitive wavelengths of wheat powdery mildew are mostly located in the red (R) and near-infrared (NIR) regions. Therefore, for the detection of a specific crop disease, the sensitive wavelengths can be determined based on the disease, and these channels are considered as target channels.

[0076] Subsequently, based on the RGB and NIR images at each shooting angle, the target channel color value and NIR color value corresponding to the disease area are determined. Finally, based on the target channel color value and NIR color value, the comprehensive index value of the disease area on the image at each shooting angle, as well as the comprehensive difference value of the disease area on the image at all shooting angles, are calculated.

[0077] Furthermore, based on the target channel color value and NIR color value, the comprehensive index value of the disease area on the image at each shooting angle is determined using the first calculation formula;

[0078] The first calculation formula includes:

[0079]

[0080] IV = λ1·IV j1 +…+λ m IV jm +…+(1-λ1-…-λ m -…-λ n-1 IV jn

[0081] Where IV is the composite index value, IV jm It is the index value at the m-th shooting angle, NIR. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle, k m It is the adjustment parameter for the m-th shooting angle, λ m It is the angle weight under the m-th shooting angle, where m is an integer between 1 and n.

[0082] Based on the target channel color value and NIR color value, determine the comprehensive difference value of the disease area on the image under all shooting angles, including:

[0083] Based on the target channel color value and NIR color value, the second calculation formula is used to determine the comprehensive difference value of the disease area on the image under all shooting angles;

[0084] The second calculation formula includes:

[0085]

[0086] Where DV is the composite difference value, and NIR is the total difference value. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle.

[0087] The following explanation uses wheat powdery mildew as an example to illustrate the acquisition of the aforementioned comprehensive index value and comprehensive difference value. Existing studies have shown that the sensitive bands of powdery mildew are mostly located in the red light (R) and near-infrared (NIR) regions. The wheat disease stress index NRPMI (NIR-RGB, powdery mildew disease index) is also calculated using these two sensitive bands. The disease stress index consists of data pairs, i.e., NRPMI = [IV, DV], where IV is the calculated comprehensive index value, and DV is the comprehensive difference value between the two characteristic bands at different angles.

[0088] This new disease stress index provides both the comprehensive disease index value for a region and the differences between the two sensitive bands of the disease, thus comprehensively characterizing disease stress. The index first calculates the comprehensive index value (IV) of wheat in the collected area by combining the angle information from the R and NIR bands. Then, due to the inherent differences between different parts and angles, the difference value (DV) between the two characteristic bands of the collected area is calculated. This combination of data yields the NRPMI. To obtain this index, the aforementioned acquisition device is used, setting the initial and changing angles between the rotating arm and the rotating axis. The red (R) and near-infrared (NIR) information from the disease images acquired at these different angles is then obtained as data support for the new disease stress index.

[0089] like Figure 3 As shown, the angle between the rotating arm and the intermediate part is set to r. up Its adjustable angle range is 90 degrees to 180 degrees. The rotating arm and intermediate component are driven by a motor and can automatically set the angle to achieve the collection of disease samples from the upper parts of wheat. The rotating shaft and intermediate component are also driven by a motor, and the included angle between the rotating arm and the intermediate component is set to r. low Its adjustable angle range is 0 degrees to 360 degrees, enabling the collection of disease data from the lower parts of wheat.

[0090] The data acquisition device, relying on automatic control, can easily drive the rotating shaft and arm to acquire image data of the upper and lower parts of wheat from different angles. During acquisition, the position of the intermediate component needs to be kept constant to ensure that the relative position of the device remains fixed during each acquisition. Secondly, r up and r low To ensure consistency in shooting angles, the initial angle needs to be set, and finally r up and r low Several adjustment angles were selected to obtain information on the R and NIR bands at different angles. Finally, a new disease stress index, NRPMI, was calculated to comprehensively characterize wheat disease stress and stress differences, and the overall disease detection results along the wheat growth direction were analyzed. The specific steps are as follows:

[0091] A. Two-band information of the upper part of the wheat obtained by adjusting the angle of the rotating arm, NIR up1 and R up1 and NIR up2 and R up2

[0092] The new disease stress index (NRPMI) for the upper region was calculated. up =[IV up DV up The calculation formula is as follows:

[0093] First, calculate the IV values ​​at different angles:

[0094]

[0095] Then, by setting weights for different angle VI values, the IV representing the upper region is calculated. up .

[0096] IV up =λ·IV up1 +(1-λ)·IV up2 , where λ is the weight value of the angle (usually set to 0.5, but can be set according to the actual data collected).

[0097] Then, the combined difference value DV between the NIR and R bands at different angles in the upper region was calculated. up :

[0098]

[0099] Finally, the new disease stress index (NRPMI) for the upper wheat region was obtained. up =[IV up DV up The new index can not only express the quantitative value of disease in the upper region of wheat, but also quantify the difference value of the two band information that can characterize the disease, so as to more completely and comprehensively characterize the disease stress in the region.

[0100] B. Same steps as A. The two-band information of the lower part of the wheat grain, obtained by adjusting the angle of the rotation axis, is NIR. low1 and R low1 and NIR low2 and R low2

[0101] The new disease stress index (NRPMI) for this lower region was calculated. low =[IV low DV low The calculation formula is as follows:

[0102] First, calculate the IV values ​​at different angles:

[0103]

[0104] Then, by setting weights for different angle VI values, the IV representing the lower region is calculated. low .

[0105] IV low =λ·IV low1 +(1-λ)·IV low2 , where λ is the weight value of the angle (usually set to 0.5, but can be set according to the actual data collected).

[0106] Then, the combined difference value DV between the NIR and R bands at different angles in the lower region was calculated. low :

[0107]

[0108] Finally, the new disease stress index (NRPMI) for the lower wheat region was obtained. low =[IV low DV low The new index can not only express the quantitative value of disease in the lower wheat region, but also quantify the difference value of the two band information that can characterize the disease, so as to more completely and comprehensively characterize the disease stress in the region.

[0109] Based on the above, the disease stress index (NRPMI) for the upper and lower parts of wheat can be calculated. up =[IV up DV up ] and NRPMI low =[IV low DV low The value of ] is used to characterize the disease stress of different growth parts of wheat in the small area, and at the same time, it can be used to compare the differences in disease stress in these different parts to comprehensively characterize the disease stress suffered by wheat growth in the area.

[0110] The further method described above mainly explains the process of determining the corresponding RGB weights and NIR weights based on the RGB and NIR images at each shooting angle, as detailed below:

[0111] Obtain the RGB image features of an RGB image;

[0112] Obtain NIR image features from NIR images;

[0113] The RGB image features and NIR image features are fused to obtain the fused features;

[0114] Based on the fusion features, RGB weights and NIR weights are determined, where the sum of the RGB weights and NIR weights is 1.

[0115] In this regard, it should be noted that, as Figure 4 As shown, a pair of RGB and NIR images are first fed into the first four convolutional and pooling layers of a Two-Stream Deep Convolutional Neural Network (DDCNN) to extract image features from each image data stream. Each image feature extraction layer in DDCNN uses Conv1-4 from VGG-16 as its backbone. Then, the image features from the two images are fused, generating a two-stream feature map (i.e., fused features) through a cascaded layer (CONCAT). The generated two-stream feature map is then used as input to DWCNN, and the network ultimately calculates the RGB weights w. RGB and NIR weight w NIR , as the fusion ratio of the two streams in a two-stream deep convolutional neural network (DDCNN), where w RGB +w NIR =1.

[0116] The Two-Stream Weighted Fully Connected Neural Network (DWCNN) consists of one pooling layer (DWPool), three fully connected layers (DWFC1, DWFC2, and DWFC3), and a Softmax layer. The number of channels in DWFC1, DWFC2, and DWFC3 are initially set to 512, 64, and 2, respectively. The Softmax layer is the last layer of DWCNN. The output of the Softmax layer is w. RGB and w NIR .

[0117] A further step in the above method mainly explains the process of determining the disease area belonging to the target disease in the image at each shooting angle based on the RGB and NIR images at each shooting angle, as well as the corresponding RGB and NIR weights, as follows:

[0118] Obtain the RGB image features of an RGB image;

[0119] Obtain NIR image features from NIR images;

[0120] The RGB image features and NIR image features are fused to obtain the fused features;

[0121] Based on the fusion features, the RGB disease regions and NIR disease regions on the image are determined, as well as the corresponding RGB disease prediction values ​​and NIR disease prediction values.

[0122] Based on the RGB and NIR disease regions in the image, the disease regions in the image are determined by combining the RGB weights and NIR weights respectively.

[0123] Based on the RGB disease prediction values ​​and NIR disease prediction values, the predicted values ​​of disease areas on the image are determined by combining the RGB weights and NIR weights respectively.

[0124] Based on the diseased areas in the image and the predicted values, the diseased areas in the image belonging to the target disease are selected.

[0125] In this regard, it should be noted that... (See also...) Figure 4 The front part of the Two-Stream Deep Convolutional Neural Network (DDCNN) consists of four convolutional layers and pooling layers to extract image features, i.e., lesion features, from each image. Each feature extraction layer in DDCNN uses Conv1-4 from VGG-16 as its backbone. Then, the features from the two images are fused and a two-stream feature map is generated through a cascaded layer (CONCAT). This two-stream feature map serves as both the input to the DWCNN network and the input to the rear part of the DDCNN network. The rear part of the DDCNN network mainly consists of four sub-networks (RGB-Cls, NIR-Cls, RGB-Bbox, and NIR-Bbox). RGB-Cls and NIR-Cls calculate the classification scores for RGB and NIR respectively, while RGB-Bbox and NIR-Bbox generate the bounding boxes (containing the lesion region) for RGB and NIR respectively. The outputs of these sub-networks use the two-stream weights w calculated by the DWCNN network. RGB and w NIR Two-stream fusion is performed to produce the final detection result. The detection loss term is defined as:

[0126]

[0127] Among them, L DD Defined as classification loss L cls and bounding box loss L Bbox The sum, where λ is the regularization parameter between the two losses, and T is the training set. and These are the predicted classification score and the bounding box, respectively. The ground truth for training labels is set to 1 for positive samples and 0 for negative samples. For each positive sample, its bounding box is set to... To calculate the bounding box loss.

[0128] Classification loss L cls :

[0129] Bounding box loss L Bbox :

[0130] func is used for learning. and The relationship between them; and These are the predicted classification score and the bounding box, respectively. It is based on RGB classification scores and NIR classification score The weighted calculation yields:

[0131]

[0132] in It is a bounding box predicted by RGB and NIR predicted bounding box The weighted calculation yields:

[0133]

[0134] The aforementioned dual-stream weighted fusion mechanism can achieve reliable disease detection results.

[0135] The crop disease identification device provided by the present invention is described below. The crop disease identification device described below and the crop disease identification method described above can be referred to in correspondence.

[0136] Figure 5 A schematic diagram of the structure of a crop disease identification device provided by the present invention is shown below. Figure 5 The device includes:

[0137] The acquisition module 51 is used to acquire RGB and NIR images of target parts on crops from different shooting angles;

[0138] The determination module 52 is used to determine the corresponding RGB weights and NIR weights based on the RGB image and NIR image at each shooting angle;

[0139] The identification module 53 is used to determine the disease area belonging to the target disease on the image at each shooting angle based on the RGB image and NIR image at each shooting angle, as well as the corresponding RGB weight and NIR weight.

[0140] In a further embodiment of the above-described apparatus, the apparatus further includes a computing module for:

[0141] The target disease is identified as corresponding to the target channel in the RGB channel.

[0142] Based on the RGB and NIR images from each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area;

[0143] Based on the target channel color value and NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, as well as the comprehensive difference value of the disease area on the image at all shooting angles.

[0144] In a further embodiment of the aforementioned apparatus, the determining module, during the process of determining the corresponding RGB weights and NIR weights based on the RGB and NIR images at each shooting angle, is specifically used for:

[0145] Obtain the RGB image features of an RGB image;

[0146] Obtain NIR image features from NIR images;

[0147] The RGB image features and NIR image features are fused to obtain the fused features;

[0148] Based on the fusion features, RGB weights and NIR weights are determined, where the sum of the RGB weights and NIR weights is 1.

[0149] In a further embodiment of the aforementioned device, the identification module, in the process of determining the diseased area belonging to the target disease on the image at each shooting angle based on the RGB and NIR images at each shooting angle, and the corresponding RGB and NIR weights, is specifically used for:

[0150] Obtain the RGB image features of an RGB image;

[0151] Obtain NIR image features from NIR images;

[0152] The RGB image features and NIR image features are fused to obtain the fused features;

[0153] Based on the fusion features, the RGB disease regions and NIR disease regions on the image are determined, as well as the corresponding RGB disease prediction values ​​and NIR disease prediction values.

[0154] Based on the RGB and NIR disease regions in the image, the disease regions in the image are determined by combining the RGB weights and NIR weights respectively.

[0155] Based on the RGB disease prediction values ​​and NIR disease prediction values, the predicted values ​​of disease areas on the image are determined by combining the RGB weights and NIR weights respectively.

[0156] Based on the diseased areas in the image and the predicted values, the diseased areas in the image belonging to the target disease are selected.

[0157] In a further embodiment of the aforementioned apparatus, the calculation module, in the process of determining the comprehensive index value of the diseased area on the image at each shooting angle based on the target channel color value and the NIR color value, is specifically used for:

[0158] Based on the target channel color value and NIR color value, the comprehensive index value of the disease area on the image at each shooting angle is determined using the first calculation formula.

[0159] The first calculation formula includes:

[0160]

[0161] IV = λ1·IV j1 +…+λ m IV jm +…+(1-λ1-…-λ m -…-λ n-1 IV jn

[0162] Where IV is the composite index value, IV jm It is the index value at the m-th shooting angle, NIR. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle, k m It is the adjustment parameter for the m-th shooting angle, λ m It is the angle weight under the m-th shooting angle, where m is an integer between 1 and n.

[0163] In a further embodiment of the aforementioned device, the calculation module, in the process of determining the comprehensive difference value of the diseased area on the image under all shooting angles based on the target channel color value and the NIR color value, is specifically used for:

[0164] Based on the target channel color value and NIR color value, the second calculation formula is used to determine the comprehensive difference value of the disease area on the image under all shooting angles;

[0165] The second calculation formula includes:

[0166]

[0167] Where DV is the composite difference value, and NIR is the total difference value. jm It is the NIR color value at the m-th shooting angle, TD jm It is the target channel color value selected from the RGB channels under the m-th shooting angle.

[0168] This invention provides a crop disease identification device. By collecting RGB and NIR images of target parts of crops from different shooting angles, the device determines the corresponding RGB and NIR weights. Based on the RGB and NIR images from each shooting angle, as well as the corresponding RGB and NIR weights, the device identifies the diseased area of ​​the target disease in the image at each shooting angle. This enables rapid identification of diseases in crops from different angles simultaneously, thereby monitoring the disease status of different growth parts of the crop and providing support for crop disease information acquisition and disease prevention and control.

[0169] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 61, a communications interface 62, a memory 63, and a communication bus 64. The processor 61, communications interface 62, and memory 63 communicate with each other via the communication bus 64. The processor 61 can call logical instructions in the memory 63 to execute a crop disease identification method. This method includes: acquiring RGB and NIR images of target parts of the crop at different shooting angles; determining corresponding RGB and NIR weights based on the RGB and NIR images at each shooting angle; and determining the diseased area belonging to the target disease on the image at each shooting angle based on the RGB and NIR images at each shooting angle and the corresponding RGB and NIR weights.

[0170] Furthermore, the logical instructions in the aforementioned memory 63 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0171] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the crop disease identification method provided by the above methods. The method includes: acquiring RGB images and NIR images of target parts on crops at different shooting angles; determining corresponding RGB weights and NIR weights based on the RGB images and NIR images at each shooting angle; and determining the disease area belonging to the target disease on the image at each shooting angle based on the RGB images and NIR images at each shooting angle, as well as the corresponding RGB weights and NIR weights.

[0172] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the crop disease identification method provided by the above methods. The method includes: acquiring RGB images and NIR images of target parts on a crop at different shooting angles; determining corresponding RGB weights and NIR weights based on the RGB images and NIR images at each shooting angle; and determining the disease area belonging to the target disease on the image at each shooting angle based on the RGB images and NIR images at each shooting angle, as well as the corresponding RGB weights and NIR weights.

[0173] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0174] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0175] Finally, it should be noted that 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 foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying crop diseases, characterized in that, include: Acquire RGB and NIR images of target parts of crops from different shooting angles; Based on the RGB image and the NIR image at each shooting angle, determine the corresponding RGB weights and NIR weights; Based on the RGB image and NIR image at each shooting angle, as well as the corresponding RGB weights and NIR weights, determine the disease area belonging to the target disease on the image at each shooting angle; The step of determining the disease region belonging to the target disease on the image at each shooting angle based on the RGB image and the NIR image at each shooting angle, and the corresponding RGB weights and NIR weights, includes: Obtain the RGB image features of the RGB image; Obtain the NIR image features of the NIR image; The RGB image features and the NIR image features are fused to obtain the fused features; Based on the fusion features, RGB disease regions and NIR disease regions on the image are determined, as well as the corresponding RGB disease prediction values ​​and NIR disease prediction values. Based on the RGB and NIR disease regions in the image, the disease regions in the image are determined by combining the RGB weights and NIR weights respectively. Based on the RGB disease prediction values ​​and NIR disease prediction values, the predicted values ​​of disease areas on the image are determined by combining the RGB weights and NIR weights respectively. Based on the diseased areas in the image and the predicted values, the diseased areas in the image belonging to the target disease are selected.

2. The method for identifying crop diseases according to claim 1, characterized in that, The method further includes: The target disease is identified as corresponding to the target channel in the RGB channel. Based on the RGB image and the NIR image at each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area; Based on the target channel color value and the NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, and the comprehensive difference value of the disease area on the image at all shooting angles.

3. The method for identifying crop diseases according to claim 1, characterized in that, The step of determining the corresponding RGB weights and NIR weights based on the RGB image and the NIR image at each shooting angle includes: Obtain the RGB image features of the RGB image; Obtain the NIR image features of the NIR image; The RGB image features and the NIR image features are fused to obtain the fused features; Based on the fusion features, RGB weights and NIR weights are determined, wherein the sum of the RGB weights and NIR weights is 1.

4. The crop disease identification method according to claim 2, characterized in that, The step of determining the comprehensive index value of the diseased area on the image at each shooting angle based on the target channel color value and the NIR color value includes: Based on the target channel color value and the NIR color value, the comprehensive index value of the disease area on the image at each shooting angle is determined using the first calculation formula; The first calculation formula includes: in, It is a composite index value. It is the first Index values ​​at various shooting angles It is the first NIR color values ​​at various shooting angles It is the first The target channel color value selected from the RGB channels under each shooting angle. It is the first Adjustment parameters for each shooting angle It is the first The angle weights for each shooting angle, where m is an integer between 1 and n.

5. The method for identifying crop diseases according to claim 2, characterized in that, The step of determining the comprehensive difference value of the disease area on the image under all shooting angles based on the target channel color value and the NIR color value includes: Based on the target channel color value and the NIR color value, the comprehensive difference value of the disease area on the image under all shooting angles is determined by the second calculation formula. The second calculation formula includes: Wherein, DV is the composite difference value. It is the first NIR color values ​​at various shooting angles It is the first The target channel color value selected from the RGB channels under each shooting angle.

6. A crop disease identification device, characterized in that, include: The acquisition module is used to acquire RGB and NIR images of target parts of crops from different shooting angles; The determination module is used to determine the corresponding RGB weights and NIR weights based on the RGB image and the NIR image at each shooting angle; The identification module is used to determine the disease area belonging to the target disease on the image at each shooting angle based on the RGB image and the NIR image at each shooting angle, as well as the corresponding RGB weights and NIR weights. The step of determining the disease region belonging to the target disease on the image at each shooting angle based on the RGB image and the NIR image at each shooting angle, and the corresponding RGB weights and NIR weights, includes: Obtain the RGB image features of the RGB image; Obtain the NIR image features of the NIR image; The RGB image features and the NIR image features are fused to obtain the fused features; Based on the fusion features, RGB disease regions and NIR disease regions on the image are determined, as well as the corresponding RGB disease prediction values ​​and NIR disease prediction values. Based on the RGB and NIR disease regions in the image, the disease regions in the image are determined by combining the RGB weights and NIR weights respectively. Based on the RGB disease prediction values ​​and NIR disease prediction values, the predicted values ​​of disease areas on the image are determined by combining the RGB weights and NIR weights respectively. Based on the diseased areas in the image and the predicted values, the diseased areas in the image belonging to the target disease are selected.

7. The crop disease identification device according to claim 6, characterized in that, The device further includes a computing module for: The target disease is identified as corresponding to the target channel in the RGB channel. Based on the RGB image and the NIR image at each shooting angle, determine the target channel color value and NIR color value corresponding to the diseased area; Based on the target channel color value and the NIR color value, determine the comprehensive index value of the disease area on the image at each shooting angle, and the comprehensive difference value of the disease area on the image at all shooting angles.

8. 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 program, it implements the crop disease identification method as described in any one of claims 1 to 5.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the crop disease identification method as described in any one of claims 1 to 5.

Images