Crop maturity identification method based on deep learning

Abstract

The invention provides a crop maturity identification method based on deep learning, and the method achieves the precise identification of crop maturity through the cooperative work of a multi-source data obtaining device, a data processing module, a data analysis module and a computer terminal. The method comprises the following specific steps: acquiring appearance characteristic data through primary data acquisition, preprocessing the data, analyzing the crop maturity probability based on the preprocessed data, performing secondary data acquisition, analyzing the crop maturity according to the maturity characteristic data, generating and executing a crop harvesting scheme, and finally optimizing an identification model through a training model. The method combines color features, texture features, morphological features, time correction factors and other multi-dimensional data of crops to calculate the crop maturation probability, and when the maturation probability reaches a preset value, subsequent steps are performed, so that the risk of misjudgment caused by subjective deviation of artificial experience and invalid data redundancy is reduced; the accuracy and efficiency of crop maturity identification are improved.

Description

Technical Field

[0001] The present invention belongs to the field of smart agriculture technology, and specifically discloses a crop maturity identification method based on deep learning. Background Art

[0002] Accurately identifying crop maturity is a core component of agricultural production. Accurate maturity identification can prevent premature or late harvesting and reduce yield losses due to misjudgment. Combined with maturity distribution data, targeted adjustments to irrigation and fertilization strategies can be made to avoid wasted resources. Traditional crop maturity identification methods rely primarily on manual judgment or single sensor data, which suffers from the following drawbacks:

[0003] Highly subjective: This method relies on manual judgment. Farmers or experts make subjective assessments by visually observing characteristics such as color and shape. This method is easily affected by lighting conditions and individual experience, resulting in high error rates and impractical for large-scale application.

[0004] Single data: Existing technologies often rely on single features, such as color, and lack the fusion of multi-source features. Key indicators are not fully explored, and maturity cannot be fully reflected.

[0005] Poor timeliness: Static models are difficult to adapt to dynamic changes in crop growth and environmental disturbances, resulting in low prediction accuracy;

[0006] Lack of feedback mechanism: Existing technologies are mostly one-way predictions, without using actual harvest results for model iterative optimization. This leads to a decrease in accuracy after long-term application.

[0007] Therefore, it is necessary to invent a crop maturity recognition method based on deep learning to solve the above problems. Summary of the Invention

[0008] In order to overcome the above-mentioned defects of the prior art, the present invention provides a crop maturity recognition method based on deep learning, which realizes accurate recognition of crop maturity through the collaborative work of multi-source data acquisition equipment, data processing modules, data analysis modules and computer terminals; the specific steps include: primary data collection to obtain appearance feature data, data preprocessing, analyzing the crop maturity probability based on the preprocessed data and performing secondary data collection, analyzing the crop maturity according to the maturity feature data, generating and executing the crop harvesting plan, and finally optimizing the recognition model through training model, which effectively solves the problems mentioned in the background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a crop maturity recognition method based on deep learning, comprising a multi-source data acquisition device, a data processing module, a data analysis module, and a computer terminal, specifically comprising the following steps:

[0010] S1. The multi-source data acquisition device performs a data acquisition to obtain the appearance feature data of the crops in the area to be detected;

[0011] S2. The data processing module processes the acquired appearance feature data to obtain pre-processed data;

[0012] S3. The data analysis module calculates the probability of crop maturity in the area to be detected based on the preprocessed data. When the maturity probability reaches a preset value, the multi-source data acquisition device performs secondary data acquisition to obtain the maturity characteristic data of the crops in the area to be detected;

[0013] S4, the data analysis module calculates the maturity of the crops in the area to be detected based on the maturity characteristic data;

[0014] S5. The computer terminal generates a crop harvesting plan based on maturity and weather data, issues the plan, and provides feedback on the harvesting results.

[0015] S6. Using crop maturity as a label and chemical composition, physical properties, and harvest results as features, the training model is used to automatically learn and optimize the weight coefficients in the maturity analysis model calculation formula.

[0016] Based on the above embodiment, the appearance feature data is a crop picture and radar cloud point data of the area to be detected;

[0017] The maturity characteristic data include: the compliance rate of key chemical components that can reflect the maturity of crops, and the compliance rate of key physical properties that can reflect the maturity of crops;

[0018] The harvest results include: crop quality grade and percentage of mature crops.

[0019] Based on the above embodiment, the maturity probability is calculated as follows: , in the formula F c is the color feature value, F t is the texture feature value, F m is the morphological feature value, ε(t) is the time correction factor, and ω is the weight coefficient, which is dynamically adjusted based on the deep learning model.

[0020] Based on the above embodiment, the calculation formula of the color feature value is: , in the formula A t is the total number of pixels in the crop image, S g is the number of pixels in the image with the typical color of the crop at the mature stage in the crop image, ‌ΔH is the angular difference between the mean value of the H channel of the current image and the standard H value of the mature sample;

[0021] Based on the above embodiment, the calculation formula of the texture feature value is: , in the formula is the multi-scale texture information entropy extracted by local binary patterns, F tmax is the maximum value of texture feature value in historical data, Gabor max is the maximum value of Gabor filter energy, k represents different detection levels or neighborhood scales, n is the total number of different detection levels or neighborhood scales, Gabor en is the Gabor filter energy, γ is the weight coefficient;

[0022] Based on the above embodiment, the calculation formula of the morphological feature value is: , N in the formula g is the number of mature fruits detected in a unit canopy area, N max A is the theoretical maximum number of mature fruits per unit area, c is the effective canopy area, D c is the current fertility days, D m is the theoretical maturity days, σ is the standard deviation of the Gaussian function that adjusts the weights of morphological characteristics;

[0023] Based on the above embodiment, the calculation formula of the time correction factor is: , β is the environmental response rate, which is fitted by historical data.

[0024] Based on the above embodiment, the calculation formula of ΔH is: , in the formula H m is the standard H value of mature samples, H c is the mean value of the H channel of the current image.

[0025] Based on the above embodiment, the maturity calculation formula of the area to be detected is: In the formula, n is the number of chemical components that can reflect the maturity of crops, m is the number of physical characteristics that can reflect the maturity of crops, and C a P is the chemical composition compliance rate that can reflect the maturity of crops. b In order to reflect the physical characteristics of crop maturity, is the weight coefficient corresponding to the chemical composition compliance rate that can reflect the maturity of crops, is the weight coefficient corresponding to the physical property compliance rate of the crop maturity, and The sum is 1.

[0026] The technical effects and advantages of the present invention are as follows:

[0027] 1. By combining multi-source data such as crop color, texture, morphology, and time correction factors, and using a deep learning model to dynamically adjust the weights of each feature, this method can comprehensively reflect the complex changes in crop maturity. Compared with traditional methods that rely on a single feature, it significantly reduces the risk of misjudgment due to one-sided features and improves the accuracy and reliability of maturity identification.

[0028] 2. A phased data collection strategy is adopted, with a primary collection of appearance features and a secondary collection of maturity features. Secondary data acquisition is triggered only when the maturity probability reaches a preset threshold. This mechanism effectively avoids redundant collection and processing of invalid data, reduces computing resource waste, and further improves analysis efficiency and system response speed by focusing on key maturity feature data.

[0029] 3. By combining actual harvest results with data on chemical composition, physical properties, and other data as model training features and combining them with dynamic weight adjustment, a closed-loop system of "data collection-analysis-execution-feedback" is formed. This mechanism enables the model to continuously adapt to dynamic changes in crop growth and environmental disturbances, maintaining high prediction accuracy over the long term, and addressing the issues of poor timeliness and long-term performance degradation associated with traditional static models. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0031] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0032] Figure 2 It is an overall step diagram of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] The present invention provides Figure 1 The crop maturity identification method based on deep learning is characterized by comprising a multi-source data acquisition device, a data processing module, a data analysis module, and a computer terminal;

[0035] In a more specific application of the present invention, the multi-source data acquisition device is used to acquire appearance feature data and maturity feature data required for primary and secondary data acquisition. The appearance feature data is a crop image of the area to be inspected. The maturity feature data includes: the compliance rate of key chemical components that can reflect the maturity of the crop, and the compliance rate of key physical properties that can reflect the maturity of the crop. For example, taking corn as the inspection target, the compliance rates of the key chemical components are specifically: the compliance rate of starch and moisture in corn kernels, and the compliance rates of the key physical properties are specifically: the compliance rate of corn kernel hardness, the compliance rate of 100-kernel weight, and the compliance rate of color change. The multi-source data acquisition device can specifically be a combination of a high-resolution RGB camera, a near-infrared spectrometer, a lidar, an environmental sensor, and a dielectric sensor.

[0036] The data processing module is responsible for image denoising, normalization, color space conversion, such as RGB to HSV, and spectral data filtering; Gaussian filtering denoising is performed using the OpenCV library, redundant background is cropped, and plant ROI area is extracted; the image is converted to HSV space and the H channel mean is calculated; the image pixel values ​​are normalized to [0, 1], and the spectral data is standardized according to the historical maximum value;

[0037] The data analysis module is used to calculate the crop maturity probability and crop maturity of the area to be detected based on the preprocessed data preprocessed by the data processing module;

[0038] The computer terminal is used to generate a crop harvesting plan based on maturity and weather data, issue the harvesting plan and provide feedback on the harvesting results.

[0039] like Figure 2 As shown, the specific steps include:

[0040] S1. The multi-source data acquisition device performs a data acquisition to obtain the appearance feature data of the crops in the area to be detected;

[0041] In a preferred technical solution of the present application, the appearance feature data is a crop picture of the area to be detected and radar cloud point data;

[0042] Furthermore, the crop image of the area to be detected is obtained by taking pictures using an RGB camera, and the radar cloud point data is obtained by scanning the canopy structure using a lidar;

[0043] S2. The data processing module processes the acquired appearance feature data to obtain pre-processed data

[0044] In the preferred technical solution of the present application, the specific process of processing the acquired appearance feature data is as follows: Gaussian filtering and denoising are performed through the OpenCV library; redundant background is cropped and the plant ROI area is extracted; the image is converted to HSV space, the mean of the H channel is calculated, and the number of pixels S in the image of the typical color of the crop at the maturity stage in the crop image is counted. g Normalize the image pixel values ​​to [0, 1], use the statistical outlier filter of the PCL library to remove the flying point noise, generate the canopy height model based on the cloud point data, and calculate the effective canopy area A c , identify the fruit point cloud clusters by density clustering and count the number of clusters per unit area as N g estimated value of;

[0045] S3. The data analysis module calculates the probability of crop maturity in the area to be detected based on the preprocessed data. When the maturity probability reaches a preset value, the multi-source data acquisition device performs secondary data acquisition to obtain the maturity characteristic data of the crops in the area to be detected;

[0046] In a preferred technical solution of the present application, the preset value is 0.4, and the theoretical value of the maturity probability is in the interval [0, 0.5];

[0047] In the preferred technical solution of this application, the maturity probability is calculated as follows: , in the formula F c is the color feature value, F t is the texture feature value, F m is the morphological feature value, ε(t) is the time correction factor, and ω is the weight coefficient, which is dynamically adjusted based on the deep learning model.

[0048] In a preferred technical solution of this application, the calculation formula of the color feature value is: , in the formula A t is the total number of pixels in the crop image, S g is the number of pixels in the image with the typical color of the crop at the mature stage in the crop image, ‌ΔH is the angular difference between the mean value of the H channel of the current image and the standard H value of the mature sample;

[0049] In a preferred technical solution of this application, the calculation formula of the texture feature value is: , in the formula is the multi-scale texture information entropy extracted by local binary patterns, F tmax is the maximum value of texture feature value in historical data, Gabor max is the maximum historical energy of Gabor filter, k represents different detection levels or neighborhood scales, n is the total number of different detection levels or neighborhood scales, Gabor enis the Gabor filter energy, γ is the weight coefficient, and for cereal crops, γ∈[0.3,0.4] is universal;

[0050] In a preferred technical solution of this application, the calculation formula of the morphological characteristic value is: , N in the formula g is the number of mature fruits detected in a unit canopy area, N max A is the theoretical maximum number of mature fruits per unit area, c is the effective canopy area, D c is the current fertility days, D m is the theoretical maturity days, σ is the standard deviation of the Gaussian function that adjusts the weights of morphological characteristics;

[0051] In a preferred technical solution of this application, the calculation formula of the time correction factor is: , β is the environmental response rate, which is fitted by historical data.

[0052] In a preferred technical solution of the present application, the calculation formula of ΔH is: , in the formula H m is the standard H value of mature samples, H c is the mean value of the H channel of the current image.

[0053] In a preferred technical solution of the present application, the maturity characteristic data of the crops in the area to be detected specifically include: the starch compliance rate, moisture compliance rate, hardness compliance rate, 100-kernel weight compliance rate and color change compliance rate of corn kernels;

[0054] Furthermore, the starch compliance rate was calculated by matching the spectral characteristics detected by the near-infrared spectrometer with the starch content calibration model. The near-infrared spectrometer was combined with a dielectric sensor to calculate the moisture compliance rate based on the linear relationship between the dielectric constant and moisture content. The hardness compliance rate was calculated by direct sampling detection with a portable hardness meter, and the 100-grain weight compliance rate was obtained by manual sampling and weighing.

[0055] Example of calculating the probability of crop maturity:

[0056] Assume that the detection data of a corn field is as follows:

[0057] Total image pixels A t =480000, number of mature color pixels S g =218400, current H channel mean H c =32°, mature sample standard H value H m =35°, using 3 detection scales, that is, n=3, when k=1, the multi-scale texture information entropy value is 4.2, when k=2, the multi-scale texture information entropy value is 5.1, when k=3, the multi-scale texture information entropy value is 3.8, the maximum value of texture feature value F in historical datatmax =10, Gabor filter energy Gabor en =38.6, Gabor filter energy history maximum Gabor max =82.4, weight γ=0.35, number of ripe fruits N g =22560, the theoretical maximum number of mature fruits per unit area N max =520, canopy area A c =48m², current fertility days D c =102, theoretical maturity days D m =115, Gaussian function standard deviation σ=15, environmental response rate β=0.08, color feature weight ω c =0.4, texture feature weight ω t =0.3, morphological feature weight ω m =0.3;

[0058] Calculation process:

[0059] ΔH=min(|32-35|, 360−|32-35|)=3

[0060] Color feature value:

[0061] F c =(218400 / 480000)×(1−3 / 180)=0.455×0.983≈0.447,

[0062] Texture feature value:

[0063] F t =(4.2 / 1 2 +5.1 / 2 2 +3.8 / 3 2 ) / 10+0.35×(38.6 / 82.4)=0.59+0.164≈0.754,

[0064] Morphological characteristic values:

[0065] F m =(22560 / (520×48))×exp[-(102-115) 2 / (2×15 2 )]≈0.904×0.687≈0.621,

[0066] Time correction factor:

[0067] ε(t)=1 / [1+exp(−0.08×(102−115))]=1 / (1+e 1.04 )≈0.326,

[0068] Probability of maturity:

[0069] P m =[(0.447×0.4)+(0.3×0.754)+(0.3×0.621)]×(1-0.326)=

[0070] (0.179+0.226+0.186)×0.674≈0.40;

[0071] Conclusion: P m =0.4, the current maturity probability meets the standard, triggering secondary data collection.

[0072] S4, the data analysis module calculates the maturity of the crops in the area to be detected based on the maturity characteristic data;

[0073] In a preferred technical solution of this application, the maturity calculation formula of the area to be detected is: In the formula, n is the number of chemical components that can reflect the maturity of crops, m is the number of physical characteristics that can reflect the maturity of crops, and C a P is the chemical composition compliance rate that can reflect the maturity of crops. b In order to reflect the physical characteristics of crop maturity, is the weight coefficient corresponding to the chemical composition compliance rate that can reflect the maturity of crops, is the weight coefficient corresponding to the physical property compliance rate of the crop maturity, and The sum is 1.

[0074] Furthermore, the chemical composition compliance rate includes: the starch compliance rate and moisture compliance rate of corn kernels; the physical property compliance rate that can reflect the maturity of crops includes: the hardness compliance rate, the 100-kernel weight compliance rate and the color change compliance rate;

[0075] Maturity calculation example:

[0076] Assume that the detection data of a corn field is as follows:

[0077] Starch compliance rate: 85%, weight coefficient corresponding to starch compliance rate: 0.3,

[0078] Moisture compliance rate: 90%, weight coefficient corresponding to moisture compliance rate: 0.2,

[0079] Hardness compliance rate: 88%, weight coefficient corresponding to hardness compliance rate: 0.15,

[0080] The compliance rate of 100-grain weight is 92%, and the corresponding weight coefficient is 0.2.

[0081] Color transformation compliance rate: 95%, weight coefficient corresponding to color transformation compliance rate: =0.15;

[0082] Calculation process:

[0083] M=0.85×0.3+0.9×0.2+0.88×0.15+0.92×0.2+0.95×0.15=0.8935;

[0084] Conclusion: The current corn maturity in the tested area is 89.35%, indicating that the crop is close to full maturity.

[0085] S5. The computer terminal generates a crop harvesting plan based on maturity and weather data, issues the plan, and provides feedback on the harvesting results.

[0086] In a preferred technical solution of this application, the crop harvesting solution includes:

[0087] When 0<M≤40%, it means that the crops in the current area are not mature, harvesting needs to be delayed, and field management needs to be strengthened;

[0088] When 40%<M≤70%, it means that some crops in the current area are mature. If the weather is stable in the future, they will continue to wait for maturity. If there is bad weather, mature crops will be selectively harvested in batches.

[0089] When 70%<M≤90%, it means that the crops in the current area are mature, and mechanical / manpower harvesting should be organized immediately;

[0090] When 90%<M≤100%, it means that the crops in the current area are over-ripe, and machinery and manual labor should work together to complete the harvest as soon as possible.

[0091] S6. Using crop maturity as a label, maturity characteristic data, and harvest result content as features, the training model is used to automatically learn and optimize the weight coefficients in the maturity analysis model calculation formula.

[0092] In a preferred technical solution of the present application, the harvest results specifically include the crop quality grade and the proportion of mature crops;

[0093] Furthermore, the process of automatically learning and optimizing the weight coefficients in the maturity analysis model calculation formula by training the model is as follows:

[0094] Through model training, the influence of each feature on maturity is automatically learned, and the attention mechanism is introduced to dynamically allocate weights. The quality grade of crops harvested each time and the proportion of mature crops are added to the training set as new features; if the deviation between the predicted maturity and the measured quality is greater than 10%, the model is triggered to be retrained.

[0095] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A crop maturity recognition method based on deep learning, characterized by: The system includes a multi-source data acquisition device, a data processing module, a data analysis module, and a computer terminal, and specifically includes the following steps: S1. The multi-source data acquisition device performs a data acquisition to obtain the appearance feature data of the crops in the area to be detected; S2. The data processing module processes the acquired appearance feature data to obtain pre-processed data; S3. The data analysis module calculates the probability of crop maturity in the area to be detected based on the preprocessed data. When the maturity probability reaches a preset value, the multi-source data acquisition device performs secondary data acquisition to obtain the maturity characteristic data of the crops in the area to be detected; S4, the data analysis module calculates the maturity of the crops in the area to be detected based on the maturity characteristic data; S5. The computer terminal generates a crop harvesting plan based on maturity and weather data, issues the plan, and provides feedback on the harvesting results. S6. Using crop maturity as a label and chemical composition, physical properties, and harvest results as features, the training model is used to automatically learn and optimize the weight coefficients in the maturity analysis model calculation formula.

2. The crop maturity recognition method based on deep learning according to claim 1, characterized in that: The appearance feature data is a crop picture and radar cloud point data of the area to be detected; The maturity characteristic data include: the compliance rate of key chemical components that can reflect the maturity of crops, and the compliance rate of key physical properties that can reflect the maturity of crops; The harvest results include: crop quality grade and percentage of mature crops.

3. The crop maturity recognition method based on deep learning according to claim 1, characterized in that: The maturity probability is calculated as follows: , in the formula F c is the color feature value, F t is the texture feature value, F m is the morphological feature value, ε(t) is the time correction factor, and ω is the weight coefficient, which is dynamically adjusted based on the deep learning model.

4. The crop maturity recognition method based on deep learning according to claim 3, characterized in that: The calculation formula of the color characteristic value is: , in the formula A t is the total number of pixels in the crop image, S g is the number of pixels in the image with the typical color of the crop at the mature stage in the crop image, and ‌ΔH is the angular difference between the mean value of the H channel of the current image and the standard H value of the mature sample.

5. The crop maturity recognition method based on deep learning according to claim 4, characterized in that: The calculation formula of the texture feature value is: , in the formula is the multi-scale texture information entropy extracted by local binary patterns, F tmax is the maximum value of texture feature value in historical data, Gabor max is the maximum value of Gabor filter energy, k represents different detection levels or neighborhood scales, n is the total number of different detection levels or neighborhood scales, Gabor en is the Gabor filter energy, and γ is the weight coefficient.

6. The crop maturity recognition method based on deep learning according to claim 4, characterized in that: The calculation formula of the morphological characteristic value is: , N in the formula g is the number of mature fruits detected in a unit canopy area, N max A is the theoretical maximum number of mature fruits per unit area, c is the effective canopy area, D c is the current fertility days, D m is the theoretical number of days to maturity, and σ is the standard deviation of the Gaussian function that adjusts the weights of morphological characteristics.

7. The crop maturity recognition method based on deep learning according to claim 4, characterized in that: The calculation formula of the time correction factor is: , β is the environmental response rate, which is fitted by historical data.

8. The crop maturity recognition method based on deep learning according to claim 4, characterized in that: The calculation formula of ΔH is: , in the formula H m is the standard H value of mature samples, H c is the mean value of the H channel of the current image.

9. The crop maturity recognition method based on deep learning according to claim 1, characterized in that: The maturity calculation formula of the area to be detected is: In the formula, n is the number of chemical components that can reflect the maturity of crops, m is the number of physical characteristics that can reflect the maturity of crops, and C a P is the chemical composition compliance rate that can reflect the maturity of crops. b In order to reflect the physical characteristics of crop maturity, is the weight coefficient corresponding to the chemical composition compliance rate that can reflect the maturity of crops, is the weight coefficient corresponding to the physical property compliance rate of the crop maturity, and The sum is 1.

Images