Rice waist explosion rate real-time detection method and system based on image recognition

Abstract

The invention discloses a rice waist explosion rate real-time detection method and system based on image recognition, relates to an image optimization technology, is used for improving the problem that detection is not timely due to waist explosion rate change caused by the correlation degree of internal and external factors of rice, and comprises the steps of obtaining a rice image sample, recording a waist explosion area of rice particles and calculating a form change rate; collecting the surface crack density and the particle volume change rate of the rice particles in the rice image samples to evaluate the damage degree of the rice particles, generating a mark detection time interval by utilizing the form change rate and the damage degree of the rice particles, screening the rice image samples according to the detection time interval, and counting the ratio of the waist explosion area to carry out waist explosion rating. And when the waist explosion rating is low, collecting ventilation data of the rice storage environment and analyzing ventilation fluctuation, evaluating internal stress distribution in combination with the ventilation fluctuation of the rice storage environment and the form change rate of the waist explosion area, and selecting different algorithm models according to the internal stress distribution of the rice particles to predict the change trend of the waist explosion rate.

Description

Technical Field

[0001] The present invention relates to the technical field of image optimization, and more specifically to a method and system for real-time detection of rice cracking rate based on image recognition. Background Art

[0002] Rice cracking rate detection technology has gradually attracted attention with the development of intelligent agriculture and the improvement of processing quality requirements. Image optimization technology applied to rice cracking rate detection can give timely alarms, thereby extending the response time for cracking.

[0003] The prior art has the following deficiencies:

[0004] Traditional methods for detecting rice cracking rate mainly rely on manual visual inspection or simple mechanical screening. Although these methods can provide certain reference data, they are characterized by low efficiency, strong subjectivity, and difficulty in achieving continuous detection. In addition, existing technologies usually only evaluate the overall appearance of rice, fail to fully focus on the precise quantification of the key indicator of cracking rate, and fail to optimize the subtle morphological changes of static rice grains, resulting in low efficiency in rice cracking rate detection. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a real-time detection method and system for rice cracking rate based on image recognition, which calculates the rate of change of the morphology of the cracking area, generates the optimal detection time interval, evaluates the influence of internal and external factors, and predicts the change trend of the cracking rate, and selects different algorithm models to predict the rice cracking rate to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for real-time detection of rice cracking rate based on image recognition comprises the following steps:

[0008] Step S1: obtaining a plurality of rice image samples, recording the cracked area of ​​the rice grains according to the edge detection algorithm and calculating the morphological change rate, collecting the surface crack density and grain volume change rate of the rice grains in each rice image sample to evaluate the damage degree of the rice grains;

[0009] Step S2: Generate a mark detection time interval using the morphological change rate and damage degree of rice grains, and screen rice image samples according to the detection time interval, and count the proportion of the rice cracking area to perform rice cracking rating; Step S3: When the rice cracking rating is low, collect ventilation data of the rice storage environment and analyze ventilation fluctuations, and evaluate the rice cracking rate caused by internal stress distribution in combination with the ventilation fluctuations of the rice storage environment and the morphological change rate of the rice cracking area;

[0010] Step S4: Analyze the correlation between internal and external factors of rice grains according to the cracking rate caused by the internal stress distribution of rice grains and select different algorithm models to predict the changing trend of the cracking rate.

[0011] In a preferred embodiment, in step S1, the step of calculating the rate of change of the morphology of the waist burst area includes: randomly setting two time points with different time intervals in each analysis sample to detect the area of ​​the waist burst area, subtracting the area of ​​the waist burst area detected at the latter time point from the area of ​​the waist burst area detected at the previous time point, dividing the calculated result by the corresponding time interval to obtain the rate of change of the morphology of the waist burst area, merging the rate of change of the morphology of the waist burst area calculated for each analysis sample into a change rate data set, and merging the time intervals into a detection time data set.

[0012] In a preferred embodiment, in step S1, the surface crack density of the rice grains is the number of cracks per unit area, the volume change rate of the grains is the ratio of the volume of the rice grains at the detection time point to the initial volume of the rice grains, the surface crack density of the rice grains and the volume change rate of the rice grains are normalized and added together as the damage degree of the rice grains, and the damage degrees of the rice grains in each analysis sample are merged into a damage degree data set.

[0013] In a preferred embodiment, in step S2, when the morphological change rate and damage degree of the rice grains are used to generate the mark detection time interval, the minimum absolute shrinkage and selection operator is used for processing, and the specific steps are as follows:

[0014] The three input features of each rice image sample are integrated into a feature matrix, and the sparsity constraint can be set as the L1 norm penalty term: , the loss function is: , is the true value of the input feature, is the model output value, n is the number of feature matrices, p is the number of input features in the feature matrix, is the regression coefficient of the model, and λ is the regularization parameter used to control the degree of sparsity.

[0015] In a preferred embodiment, in step S2, after generating the mark detection time interval, the waist burst area ratio in the rice image sample within the mark detection time interval is compared with a preset waist burst ratio threshold to classify the rice image sample as having a high waist burst rating or a low waist burst rating.

[0016] In a preferred embodiment, in step S3, the ventilation data of the rice storage environment is the ventilation rate of the rice storage environment, a plurality of equal time periods are set to perform statistics on the ventilation rate of the rice storage environment, and the standard deviation of the ventilation rate in each time period is used as the ventilation fluctuation of the rice storage environment.

[0017] In a preferred embodiment, in step S3, the average value of ventilation fluctuations of the rice storage environment at different periods is selected as the ventilation fluctuation benchmark of the rice storage environment;

[0018] Mark the rice storage environment collection time when the ventilation fluctuation is 0, record the rice grain burst rate at the marked collection time, calculate the external damage burst rate in combination with the rice grain burst rate under the ventilation fluctuation benchmark, calculate the external damage ratio by combining the ventilation data when the rice image sample is collected with the ventilation fluctuation benchmark, and finally, take the product of the external damage burst rate and the external damage ratio as the external damage burst rate of the grains in the rice image sample;

[0019] The rice grain burst rate caused by the internal stress distribution of the rice grains is obtained by subtracting the grain damage rate from the external damage burst rate.

[0020] In a preferred embodiment, in step S4, when the difference between the external damage cracking rate of the rice grains and the cracking rate caused by the internal stress distribution is within a preset correlation threshold, it is judged that the correlation between the internal and external factors of the rice grains is strong; otherwise, it is judged that the correlation between the internal and external factors of the rice grains is weak.

[0021] In a preferred embodiment, in step S4, when the correlation between the internal and external factors of the rice grains in the rice image sample is weak, the random forest algorithm in the decision tree model is used to predict the changing trend of the rice grain burst rate in the rice image sample; otherwise, the convolutional neural network model under the deep learning framework is used for prediction.

[0022] A rice cracking rate real-time detection system based on image recognition, used to implement the above-mentioned rice cracking rate real-time detection method based on image recognition, including an image acquisition module, a sample analysis module, an environmental assessment module and a prediction optimization module;

[0023] The image acquisition module is used to collect rice image samples on the sample plane to be tested;

[0024] The sample analysis module is used to extract the rice grain profile information and remove noise interference, and calculate the area of ​​the rice grain cracking region;

[0025] The environmental assessment module is used to collect ventilation data of the rice storage environment;

[0026] The prediction optimization module is used to select the appropriate algorithm model to output the waist burst rate prediction results according to the correlation between internal and external factors.

[0027] The technical effects and advantages of the method and system for real-time detection of rice cracking rate based on image recognition of the present invention are as follows:

[0028] The present invention obtains a plurality of rice image samples, records the cracking area of ​​rice grains and calculates the morphological change rate, collects the surface crack density and the grain volume change rate of the rice grains in each rice image sample to evaluate the damage degree of the rice grains, generates a marking detection time interval by using the morphological change rate of the rice grains and the damage degree, screens the rice image samples corresponding to the detection time interval according to the damage of the rice grains, improves the analysis reliability of the image data, facilitates the subsequent prediction and selection processing of the algorithm model, screens the rice image samples according to the detection time interval, and performs cracking rating by counting the proportion of cracking areas. When the cracking rating is low, the ventilation data of the rice storage environment is collected and the ventilation fluctuation is analyzed. The lower the cracking rating of the rice is, the more attention needs to be paid to the rice grains in the corresponding image samples. The internal stress distribution is evaluated in combination with the ventilation fluctuation of the rice storage environment and the morphological change rate of the cracking area. Different algorithm models are selected according to the internal stress distribution of the rice grains to predict the cracking rate change trend, thereby improving the accuracy of rice cracking recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The present invention is a schematic diagram of a method for real-time detection of rice cracking rate based on image recognition.

[0030] Figure 2 The present invention is a flow chart of a real-time detection system for rice cracking rate based on image recognition. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0032] The present invention obtains multiple rice image samples, records the cracking areas of rice grains and calculates the morphological change rate, collects the surface crack density and the grain volume change rate of the rice grains in each rice image sample to evaluate the damage degree of the rice grains, generates a marking detection time interval by using the morphological change rate and the damage degree of the rice grains, screens the rice image samples according to the detection time interval, and performs cracking rating by counting the proportion of cracking areas. When the cracking rating is low, the ventilation data of the rice storage environment is collected and the ventilation fluctuation is analyzed, and the internal stress distribution is evaluated in combination with the ventilation fluctuation of the rice storage environment and the morphological change rate of the cracking area, and different algorithm models are selected according to the internal stress distribution of the rice grains to predict the cracking rate change trend, thereby improving the accuracy of rice cracking recognition.

[0033] Example 1, a real-time detection method for rice cracking rate based on image recognition, such as Figure 1As shown, the following steps are included:

[0034] Step S1: obtaining a plurality of rice image samples, recording the cracked area of ​​the rice grains according to the edge detection algorithm and calculating the morphological change rate, collecting the surface crack density and grain volume change rate of the rice grains in each rice image sample to evaluate the damage degree of the rice grains;

[0035] Step S2: Generate a mark detection time interval using the morphological change rate and damage degree of rice grains, and screen rice image samples according to the detection time interval, and count the proportion of the rice cracking area to perform rice cracking rating; Step S3: When the rice cracking rating is low, collect ventilation data of the rice storage environment and analyze ventilation fluctuations, and evaluate the rice cracking rate caused by internal stress distribution in combination with the ventilation fluctuations of the rice storage environment and the morphological change rate of the rice cracking area;

[0036] Step S4: Analyze the correlation between internal and external factors of rice grains according to the cracking rate caused by the internal stress distribution of rice grains and select different algorithm models to predict the changing trend of the cracking rate.

[0037] The specific implementation is as follows:

[0038] In step S1, when obtaining rice image samples, an equal amount of rice in the same processing batch is selected as an analysis sample, two time points are randomly set within different time intervals to detect the morphological changes of the waist-burst area, and rice grain images on multiple sample planes to be tested are obtained by an industrial camera. The collected images are processed using an edge detection algorithm and contour information of the rice grains is extracted, and the contour information of the rice grains is denoised. The area of ​​the waist-burst area of ​​the rice grain sample is calculated according to the contour information after denoising, and the area values ​​of the waist-burst area at two time points are recorded. The morphological change rate of the waist-burst area is obtained by subtracting the area of ​​the waist-burst area detected at a later time point from that at a time point, and then dividing the difference by the corresponding time interval.

[0039] The morphological change rates of the cracked areas calculated for each analysis sample were merged into a change rate data set, and the time intervals were merged into a detection time data set. The midpoint of the time interval of each rice image sample was used as the detection time point for the surface crack density and particle volume change rate of the rice grains and collected.

[0040] Furthermore, the image sequences of various rice grain image samples at different detection time points are used to extract the waist-burst area through image segmentation algorithm. At the same time, the midpoint of the time interval between every two adjacent image time points is used as the detection time point to construct a detection time data set.

[0041] Among them, the image segmentation algorithm may include but is not limited to a threshold segmentation-based method, an edge detection-based method, a region growing-based method, and a deep learning-based image semantic segmentation algorithm; the above algorithms can be flexibly selected or used in combination according to the actual image quality and particle edge features to ensure the accuracy and stability of rice contour extraction.

[0042] The surface crack density of rice grains is obtained by counting the number of cracks per unit area. The volume change rate of the rice grains is the ratio of the volume of the rice grains at the detection time point to the initial volume of the rice grains. The surface crack density of the rice grains and the volume change rate of the rice grains are normalized and added together to obtain the damage degree of the rice grains. The damage degrees of the rice grains in each analysis sample are merged into a damage degree data set.

[0043] Among them, the surface crack density of rice grains refers to the number of cracks on the surface of rice grains per unit area. It is specifically used to quantitatively characterize the distribution of cracks on the surface of rice grains. If the value is larger, it means that there are more structural damages on the surface of the grains, which may affect its integrity during processing, transportation or cooking.

[0044] It should be noted that there is no unique method for normalizing the surface crack density of rice grains and the volume change rate of rice grains. For example, when the Min-Max normalization method is used to process the surface crack density of rice grains and the volume change rate of rice grains, the greater the surface crack density or volume change rate of the rice grains, the higher the degree of damage to the rice grains.

[0045] The particle volume change rate refers to the ratio of the volume of the rice particle at the detection time point to the initial volume. The particle volume change rate is calculated by taking the ratio of the estimated volume of the particle at the current moment to the volume of the particle at the initial moment.

[0046] Furthermore, the particle volume change rate is obtained by using a high-resolution image acquisition system or three-dimensional reconstruction technology to obtain the morphological data of rice grains at different time points, and using an image segmentation algorithm to extract the contour information of the particles; then, the three-dimensional volume of the particles is calculated based on the particle contour through a volume estimation method, such as based on a rotational projection method or a multi-view stereo reconstruction technology, to obtain the particle volume at different time points; then, the particle volume at the initial moment is used as a reference to calculate the particle volume change rate.

[0047] Specifically, the setting of the time point can be determined by the experimenter based on actual factors such as the number of samples, sampling frequency, equipment response speed, and analysis requirements, which will not be elaborated here.

[0048] In step S2, the minimum absolute shrinkage and selection operator are used to screen the rice image samples and generate a mark detection time interval, and the proportion of the rice image sample with cracked rice in the mark detection time interval is counted. The specific steps are as follows:

[0049] The morphological change rate dataset of the waist explosion area, the detection time dataset and the damage degree dataset are used as input features. The sparsity constraint is added to the loss function to set the influence weight coefficient. The coordinate descent method is used to solve the weight coefficient, and the output detection time interval is used as the marking time interval.

[0050] In the least absolute shrinkage and selection operator, the three input features of each rice image sample are integrated into a feature matrix, and the sparsity constraint can be set as the L1 norm penalty term: , the loss function is: , is the true value of the input feature, is the model output value, n is the number of feature matrices, p is the number of input features in the feature matrix, is the regression coefficient of the model, and λ is the regularization parameter used to control the degree of sparsity.

[0051] When rating the cracked rice grains, the proportion of cracked areas in the rice image sample at the marked detection time interval is compared with a preset cracked proportion threshold. When the proportion of cracked areas in the rice image sample at the detection time interval is lower than the cracked proportion threshold, the cracked rice image sample is judged to have a high cracked rating; when the proportion of cracked areas in the rice image sample at the detection time interval exceeds the cracked proportion threshold, the cracked rice image sample is judged to have a low cracked rating.

[0052] It should be noted that the weight coefficient is used to integrate the three input features in this example into a feature matrix. The coordinate descent method is an optimization algorithm that finds the minimum or maximum value of the function by gradually optimizing each coordinate. In this example, it is used to calculate the optimal weight coefficient of the three input features. The above-mentioned marker detection time interval screening requires the use of a central processing unit to complete data processing and model training, and the prediction results are used to guide subsequent detection operations.

[0053] In step S3, when the rice image sample has a low waist explosion rating, ventilation data of the rice storage environment is collected, the ventilation data of the rice storage environment is the ventilation rate of the rice storage environment, a plurality of equal time periods are set to perform statistics on the ventilation rate of the rice storage environment, and a standard deviation of the ventilation rate is calculated as the ventilation fluctuation of the rice storage environment;

[0054] The influence of the external storage environment of rice on the rice burst rate is evaluated according to the ventilation fluctuation of the rice storage environment, the average value of the ventilation fluctuation of the rice storage environment at different periods is selected as the ventilation fluctuation benchmark of the rice storage environment, and the rice burst rate under the ventilation fluctuation benchmark is detected;

[0055] Among them, the rice cracking rate detection method under the ventilation fluctuation benchmark includes storage environment parameter collection, sample image processing and cracking identification statistics, which can effectively quantify the impact of the external ventilation environment on the structural stability of rice, and is suitable for application scenarios such as rice storage stability assessment, storage optimization control and quality prediction.

[0056] Furthermore, by randomly selecting representative rice samples from rice storage batches within the time period corresponding to the ventilation fluctuation benchmark, a high-resolution image acquisition system is used to obtain sample images, and image preprocessing and image segmentation algorithms are used to identify the cracking features in rice grains. By counting the ratio between the number of cracked grains and the total number of grains, the rice cracking rate under the current ventilation fluctuation benchmark is calculated, which is then used to evaluate the impact of the external ventilation environment on the structural stability of rice.

[0057] The specific selection of different periods can be based on the natural stage division of the storage cycle, seasonal changes, the operating status of ventilation equipment, the time point of environmental parameter mutation, etc., which will not be elaborated here;

[0058] Mark the rice storage environment collection time when the ventilation fluctuation is 0, record the rice grain burst rate at the collection time, subtract the rice grain burst rate at the collection time from the rice grain burst rate under the ventilation fluctuation benchmark to obtain the external damage burst rate, call the ventilation data of the rice storage environment when the rice image sample is collected and calculate the ventilation fluctuation, take the ratio of the ventilation fluctuation to the ventilation fluctuation benchmark as the external damage ratio, and take the product of the external damage burst rate and the external damage ratio as the external damage burst rate of the rice grains in the rice image sample;

[0059] Counting the number of rice grains whose volumes are lower than a preset volume threshold during the collection time, and using the ratio of the number of rice grains to the total number of grains as the grain damage rate of the rice grains;

[0060] The rice grain burst rate caused by the internal stress distribution of the rice grains is obtained by subtracting the grain damage rate of the rice grains from the external damage burst rate, and the rate is recorded.

[0061] It should be noted that the burst rate of rice grains is composed of the burst rate caused by external damage of the rice grains and the burst rate caused by internal stress distribution. The correlation between the two can also be quantified through other correlation analysis methods, providing a basis for the selection of subsequent prediction models.

[0062] In step S4, different algorithm models are selected according to the internal stress distribution of rice grains to predict the changing trend of the burst rate, and the external damage burst rate of rice grains is compared with the burst rate caused by the internal stress distribution. The closer the external damage burst rate of rice grains is to the burst rate caused by the internal stress distribution, the stronger the correlation between the internal and external factors of the rice grains is, because when the external damage burst rate of rice grains is close to the burst rate caused by the internal stress distribution, it is more likely to produce damaging interactions.

[0063] When the difference between the external cracking rate of the rice grain and the cracking rate caused by the internal stress distribution is within the preset correlation threshold, it is judged that the correlation between the internal and external factors of the rice grain is strong; otherwise, it is judged that the correlation between the internal and external factors of the rice grain is weak;

[0064] When the correlation between the internal and external factors of rice grains in the rice image samples is weak, the random forest algorithm in the decision tree model is used to predict the change trend of the rice grain cracking rate in the rice image samples; when the correlation between the internal and external factors of rice grains in the rice image samples is strong, the convolutional neural network model under the deep learning framework is used for prediction;

[0065] Among them, the random forest algorithm in the decision tree model is an integrated learning method, which belongs to the integrated classification and regression algorithm based on decision trees. The algorithm builds multiple decision tree models and integrates the prediction results of each tree for voting (classification task) or averaging (regression task) to improve the overall prediction performance and robustness;

[0066] The prediction results are output to the user interface for the operator's reference. The random forest algorithm constructs multiple decision trees and votes on the prediction results of each tree to obtain the final prediction value, while the convolutional neural network model extracts image features and completes the prediction task through multi-layer convolution and pooling operations.

[0067] It should be noted that the selection of the two algorithm models is based on the strength of the correlation between internal and external factors to ensure the accuracy and reliability of the prediction results.

[0068] Embodiment 2, a method for real-time detection of rice cracking rate based on image recognition, such as Figure 2 As shown, it includes the following modules:

[0069] Image acquisition module: used to collect rice image samples on the sample plane to be tested;

[0070] Sample analysis module: used to extract rice grain profile information and remove noise interference, and calculate the area of ​​the rice cracking region;

[0071] Environmental assessment module: used to collect ventilation data of rice storage environment;

[0072] Prediction and optimization module: used to select the appropriate algorithm model to output the waist burst rate prediction results according to the correlation between internal and external factors.

[0073] In practical applications, this example provides a method of use, which is specifically implemented as follows:

[0074] First, evenly spread the rice sample to be tested on the sample plane to be tested, ensuring that the sample surface is flat and has no obvious accumulation;

[0075] The industrial camera is installed on a fixed bracket and set at a 45-degree angle to the plane of the sample to be tested to avoid the influence of mirror reflection on the image acquisition quality;

[0076] The uniform light source is located directly above the plane of the sample to be tested. It adopts an LED lamp array design, which can evenly cover the surface of the sample, thereby reducing image noise interference caused by uneven lighting;

[0077] After starting the system, the industrial camera acquires high-definition images of rice samples and transmits the image data to the sample analysis module for processing;

[0078] The edge detection algorithm in the sample analysis module specifically uses the Canny algorithm to extract the contour information of rice grains. Since the image acquisition process may be disturbed by ambient light or tiny particles on the sample surface, the noise points are removed through opening and closing operations to improve the accuracy of contour extraction.

[0079] Subsequently, the waist burst area calculation module calculates the area of ​​the waist burst area using the pixel counting method based on the de-noised contour information, and records the waist burst area values ​​at two time points; the waist burst area morphological change rate is obtained by subtracting the waist burst area at the latter time point from the area at the previous time point and dividing the result by the corresponding time interval;

[0080] The key to this step is to ensure that the calculation accuracy of the waist explosion area reaches the single pixel level through high-resolution industrial cameras and precise algorithm design.

[0081] The central processing unit receives a data set of the rate of change of the morphology of the waist explosion area, a data set of the detection time, and a data set of the degree of damage, and selects the best detection time interval through the minimum absolute shrinkage and selection operator;

[0082] Specifically, sparsity constraints are added to the loss function, influencing weight coefficients are set to eliminate redundant features, and the coordinate descent method is used to solve the optimal weight coefficient. The core principle of this process is to reduce the complexity of the model through sparsity constraints, thereby improving prediction efficiency and accuracy. Finally, the system determines the optimal detection time interval based on the screening results and applies it to subsequent sample analysis.

[0083] The system compares the actual proportion of the rice cracking area with the preset threshold to obtain the rice cracking rating of the rice image sample; if the proportion of the rice cracking area is low, the ventilation data of the rice storage environment is collected. The ventilation data of the rice storage environment is the ventilation rate of the rice storage environment, and the influence of the external environment on the grain damage is evaluated, that is, the external damage cracking rate;

[0084] Finally, the correlation between internal and external factors is analyzed based on the burst rate caused by external damage and the burst rate caused by the internal stress of rice grains. Then, different algorithm models are selected to predict the changing trend of the burst rate based on the correlation between internal and external factors of rice grains.

[0085] The technical solution combining image recognition and multi-factor analysis is realized through the division of labor and cooperation among the above modules, which not only meets the needs of modern agricultural production for intelligent quality inspection, but also provides a new technical path for rice processing quality control.

[0086] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0087] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0088] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0089] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0090] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time detection method for rice cracking rate based on image recognition, characterized in that: The following steps are included: Step S1: obtaining a plurality of rice image samples, recording the cracked area of ​​the rice grains according to the edge detection algorithm and calculating the morphological change rate, collecting the surface crack density and grain volume change rate of the rice grains in each rice image sample to evaluate the damage degree of the rice grains; Step S2: Generate a mark detection time interval using the morphological change rate and damage degree of the rice grains, and screen the rice image samples according to the detection time interval, and calculate the proportion of the rice cracking area to perform rice cracking rating; Step S3: When the rice cracking rating is low, the ventilation data of the rice storage environment is collected and the ventilation fluctuation is analyzed, and the cracking rate caused by the internal stress distribution is evaluated in combination with the ventilation fluctuation of the rice storage environment and the rate of change of the cracking area morphology; Step S4: Analyze the correlation between internal and external factors of rice grains according to the cracking rate caused by the internal stress distribution of rice grains and select different algorithm models to predict the changing trend of the cracking rate.

2. A method for real-time detection of rice cracking rate based on image recognition according to claim 1, characterized in that: In step S1, the step of calculating the rate of change of the morphology of the waist burst area includes: randomly setting two time points with different time intervals in each analysis sample to detect the area of ​​the waist burst area, subtracting the area of ​​the waist burst area detected at the latter time point from the area of ​​the waist burst area detected at the previous time point, dividing the calculated result by the corresponding time interval to obtain the rate of change of the morphology of the waist burst area, merging the rate of change of the morphology of the waist burst area calculated for each analysis sample into a change rate data set, and merging the time intervals into a detection time data set.

3. A method for real-time detection of rice cracking rate based on image recognition according to claim 1, characterized in that: In step S1, the surface crack density of the rice grains is the number of cracks per unit area, and the volume change rate of the rice grains is the ratio of the volume of the rice grains at the detection time point to the initial volume of the rice grains. The surface crack density of the rice grains and the volume change rate of the rice grains are normalized and added together as the damage degree of the rice grains. The damage degrees of the rice grains in each analysis sample are merged into a damage degree data set.

4. A method for real-time detection of rice cracking rate based on image recognition according to claim 1, characterized in that: In step S2, when the morphological change rate and damage degree of the rice grains are used to generate the mark detection time interval, the minimum absolute shrinkage and selection operator is used for processing. The specific steps are as follows: The three input features of each rice image sample are integrated into a feature matrix, and the sparsity constraint can be set as the L1 norm penalty term: , the loss function is: , is the true value of the input feature, is the model output value, n is the number of feature matrices, p is the number of input features in the feature matrix, is the regression coefficient of the model, and λ is the regularization parameter used to control the degree of sparsity.

5. A method for real-time detection of rice cracking rate based on image recognition according to claim 4, characterized in that: In step S2, after generating the mark detection time interval, the waist burst area ratio in the rice image sample within the mark detection time interval is compared with a preset waist burst ratio threshold to classify the rice image sample as having a high waist burst rating or a low waist burst rating.

6. The method for real-time detection of rice cracking rate based on image recognition according to claim 1, characterized in that: In step S3, the ventilation data of the rice storage environment is the ventilation rate of the rice storage environment. A plurality of equal time periods are set to perform statistics on the ventilation rate of the rice storage environment, and the standard deviation of the ventilation rate in each time period is used as the ventilation fluctuation of the rice storage environment.

7. The method for real-time detection of rice cracking rate based on image recognition according to claim 1, characterized in that: In step S3, the average value of ventilation fluctuations of the rice storage environment at different periods is selected as the ventilation fluctuation benchmark of the rice storage environment; Mark the rice storage environment collection time when the ventilation fluctuation is 0, record the rice grain burst rate at the marked collection time, calculate the external damage burst rate in combination with the rice grain burst rate under the ventilation fluctuation benchmark, calculate the external damage ratio by combining the ventilation data when the rice image sample is collected with the ventilation fluctuation benchmark, and finally, take the product of the external damage burst rate and the external damage ratio as the external damage burst rate of the grains in the rice image sample; The rice grain damage rate is subtracted from the external damage cracking rate to obtain the cracking rate caused by the internal stress distribution of the rice grain. The rice grain burst rate is the ratio of the number of grains that burst during the collection period to the total number of grains. When calculating the grain damage rate of rice grains, the number of rice grains whose volume is lower than a preset volume threshold is counted within the collection time, and the ratio of the number of rice grains to the total number of grains is taken as the grain damage rate of the rice grains.

8. The method for real-time detection of rice cracking rate based on image recognition according to claim 7, characterized in that: In step S4, when the difference between the external cracking rate of the rice grains and the cracking rate caused by the internal stress distribution is within a preset correlation threshold, it is judged that the correlation between the internal and external factors of the rice grains is strong; otherwise, it is judged that the correlation between the internal and external factors of the rice grains is weak.

9. The method for real-time detection of rice cracking rate based on image recognition according to claim 8, characterized in that: In step S4, when the correlation between the internal and external factors of the rice grains in the rice image sample is weak, the random forest algorithm in the decision tree model is used to predict the changing trend of the rice grain burst rate in the rice image sample; otherwise, the convolutional neural network model under the deep learning framework is used for prediction.

10. A real-time detection system for rice cracking rate based on image recognition, based on a real-time detection method for rice cracking rate based on image recognition according to any one of claims 1 to 9, characterized in that: It includes image acquisition module, sample analysis module, environmental assessment module and prediction and optimization module; The image acquisition module is used to acquire rice image samples on the sample plane to be tested; The sample analysis module is used to extract the rice grain profile information and remove noise interference, and calculate the area of ​​the rice grain cracking region; The environmental assessment module is used to collect ventilation data of the rice storage environment; The prediction optimization module is used to select the appropriate algorithm model to output the waist burst rate prediction results according to the correlation between internal and external factors.

Images