Jujube classificator, method and system based on machine vision

Abstract

The embodiment of the invention provides a Chinese date classificator, method and system based on machine vision. The jujube classificator comprises a mixed light source device which comprises a visible light source module, an ultraviolet light source module and an infrared light source module and is used for irradiating target jujubes in turn based on the visible light source module, the ultraviolet light source module and the infrared light source module; the image acquisition device is used for acquiring a to-be-processed image of the target jujube based on the high-definition camera, and the to-be-processed image comprises an image of the target jujube irradiated by the visible light source module, the ultraviolet light source module and the infrared light source module in turn; the recognition device is used for performing spectral data analysis on the to-be-processed image and determining the grade of the target jujube based on an analysis result; and the mechanical control execution device is used for conveying the target jujubes to different storage areas based on the grades. Through the jujube screening device, the problem that the intelligent degree is not high when jujubes are screened through related technologies is solved.

Description

Technical Field

[0001] The invention relates to the technical field of grading Chinese medicinal materials, and in particular to a jujube selection machine, a selection method and a selection system based on machine vision. Background Art

[0002] Jujube, as a traditional Chinese medicinal material, is widely used in diet therapy and medicinal diet. It contains rich nutrients and a variety of pharmacologically active ingredients, such as polysaccharides, vitamins, amino acids, etc., and has a variety of health benefits such as enhancing immunity, anti-oxidation, and anti-fatigue. Due to the differences in the growth environment, variety, and maturity of jujube, its pharmacological properties also vary to a certain extent. Therefore, quality screening of jujube to ensure its efficacy and safety has important practical significance and market demand.

[0003] At present, the relevant technologies for screening jujubes mainly include manual selection, mechanical screening and traditional photoelectric sorting technology. Manual selection relies on the workers' experience and visual inspection, which is inefficient and easily affected by subjective factors, making it difficult to ensure the consistency and accuracy of screening. Although mechanical screening technology can increase the screening speed, it mainly relies on the physical size differences of jujubes and cannot evaluate the intrinsic quality of jujubes. Traditional photoelectric sorting technology identifies the surface characteristics of jujubes through spectral analysis, which makes it difficult to evaluate the intrinsic quality of jujubes.

[0004] Although these technologies have improved the efficiency and accuracy of jujube screening to a certain extent, they generally have the problem of low intelligence. Summary of the invention

[0005] The embodiments of the present invention provide a jujube selection machine, selection method and selection system based on machine vision, so as to at least solve the problem that the intelligent degree of jujube screening using related technologies is not high.

[0006] According to one embodiment of the present invention, a jujube selection machine based on machine vision is provided, comprising: a mixed light source device, comprising a visible light source module, an ultraviolet light source module, and an infrared light source module, for irradiating target jujubes based on the visible light source module, the ultraviolet light source module, and the infrared light source module in turn; an image acquisition device, for acquiring a to-be-processed image of the target jujube based on a high-definition camera, the to-be-processed image comprising an image of the target jujube irradiated by the visible light source module, the ultraviolet light source module, and the infrared light source module in turn; an identification device, for performing spectral data analysis on the to-be-processed image, and determining the grade of the target jujube based on the analysis result; and a mechanical control execution device, for transporting the target jujube to different storage areas based on the grade.

[0007] According to another embodiment of the present invention, a method for selecting jujubes based on machine vision is also provided, comprising: using visible light, ultraviolet light, and infrared light to alternately irradiate target jujubes; collecting images to be processed of the target jujubes based on a high-definition camera, the images to be processed comprising images of the target jujubes under alternate irradiation of visible light, ultraviolet light, and infrared light, respectively; performing spectral data analysis on the images to be processed, and determining the grade of the target jujubes based on the analysis results; and transporting the target jujubes to different storage areas based on the grade.

[0008] According to another embodiment of the present invention, a jujube selection system based on machine vision is also provided, comprising: a jujube selection machine as described in any one of the above items; a fine material storage device for storing target jujubes that meet preset grades, wherein the target jujubes are moved into the fine material storage device by the mechanical control execution device.

[0009] According to yet another embodiment of the present invention, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any one of the above method embodiments when run.

[0010] According to yet another embodiment of the present invention, there is provided an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0011] According to yet another embodiment of the present invention, a computer program product is provided, comprising computer instructions, which implement the steps in any of the above method embodiments when executed by a processor.

[0012] Through one of the embodiments of the present invention, due to the use of a mixed light source device, visible light, ultraviolet light, and infrared light can be emitted in turn to illuminate the jujube in turn. This multi-spectral imaging technology can capture the characteristics of the jujube from different angles and spectral ranges, providing richer information than a single light source. This multi-light source setting helps to more comprehensively identify the surface and internal quality of the jujube, including color, shape, maturity, etc., thereby improving the intelligent level of jujube screening.

[0013] Since the images of jujube are collected by high-definition cameras, these images can reveal different characteristics of jujube under different light sources. High-definition imaging technology allows the subtle features of jujube to be clearly displayed, providing a basis for subsequent precise analysis.

[0014] Since the recognition system is used to analyze the spectral data of the collected images, this analysis can determine the grade of jujube based on the image data. Through machine learning and deep learning technology, the model can be trained to identify the grade characteristics of jujube and realize automatic grade determination.

[0015] According to the grade of the dates determined by the identification system, the mechanical control execution system transports the dates to different storage areas. This automated sorting process reduces manual intervention and improves sorting efficiency and accuracy.

[0016] By using machine vision technology to detect and study the phenotypic characteristics, surface defects, and internal defects of jujubes, the grading quality of jujubes is guaranteed, the competitiveness of the jujube commodity market is enhanced, and the economic benefits of the jujube industry are improved.

[0017] To sum up, the embodiments of the present invention effectively solve the problem of low intelligence level of traditional jujube screening technology through the above-mentioned technical scheme, realize the intelligent selection of jujubes, improve the efficiency and accuracy of jujube screening, and provide a new solution for the automation and intelligence of the jujube industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic block diagram of the structure of a jujube selection machine based on machine vision according to an embodiment of the present invention; Figure 2 is a schematic block diagram of the structure of a hybrid light source device according to an embodiment of the present invention; Figure 3 is a schematic block diagram of a structure of an identification device according to an embodiment of the present invention; Figure 4 is another schematic block diagram of the structure of an identification device according to an embodiment of the present invention; Figure 5 is a flow chart of a method for selecting jujubes based on machine vision according to an embodiment of the present invention; Figure 6 is a flow chart of a method for determining the grade of a target jujube according to an embodiment of the present invention; Figure 7 is a flow chart of a method for obtaining a fuzzy dynamic function curve according to an embodiment of the present invention; Figure 8 4 is a flow chart of a method for determining the comprehensive membership of a target jujube according to an embodiment of the present invention.

[0019] Explanation of the accompanying drawings: 1. Mixed light source device; 11. Visible light source module; 12. Ultraviolet light source module; 13. Infrared light source module; 2. Image acquisition device; 3. Recognition device; 31. Functionalization module; 32. Feature extraction module; 33. Index extraction module; 34. Bayesian reasoning network; 35. Classification module; 36. Fuzzy set module; 4. Mechanical control execution device. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0021] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and in combination with the embodiments.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0023] In this embodiment, a jujube selection machine based on machine vision is provided. Figure 1 is a schematic block diagram of a jujube selection machine based on machine vision according to an embodiment of the present invention. Figure 1 As shown, the jujube selection machine includes: The hybrid light source device 1 includes a visible light source module 11, an ultraviolet light source module 12, and an infrared light source module 13, and is used to irradiate the target jujube in turn based on the visible light source module 11, the ultraviolet light source module 12, and the infrared light source module 13; In an exemplary embodiment, Figure 2 is a schematic block diagram of the structure of a hybrid light source device according to an embodiment of the present invention. Figure 2 As shown, the hybrid light source device 1 integrates a visible light source module 11, an ultraviolet light source module 12, and an infrared light source module 13. Specifically, a group of LED lamps or xenon lamps are used as visible light sources to provide standard white light illumination for capturing intuitive features such as the color and shape of jujube. Ultraviolet LED lamps are used as light sources to detect tiny flaws on the surface of jujube or fluorescent reactions of specific chemicals, which are invisible under visible light. Infrared LED lamps or lasers are used as light sources to penetrate the epidermis of jujube to detect internal qualities such as the sugar content and maturity of the jujube. Through programming control, the three light source modules are used to irradiate the jujube in turn according to a certain time sequence or conditions to obtain comprehensive spectral information.

[0024] The image acquisition device 2 is used to acquire the to-be-processed image of the target jujube based on a high-definition camera, wherein the to-be-processed image includes the image of the target jujube irradiated by the visible light source module 11, the ultraviolet light source module 12, and the infrared light source module 13 in turn; In an exemplary embodiment, a high-definition industrial camera is used with an appropriate lens to ensure that clear images of jujubes can be obtained under different light sources. The camera is fixed at a specific position of the machine, and through automatic control, the images of jujubes under different light sources are synchronously collected and the image data is transmitted to the processing system.

[0025] The recognition device 3 is used to perform spectral data analysis on the image to be processed and determine the grade of the target jujube based on the analysis result; In an exemplary embodiment, a high-performance image processing and analysis software is provided to process the image data obtained from the image acquisition device 2. Machine learning and deep learning algorithms, such as convolutional neural networks (CNNs), are used to extract and learn features of the jujube images to identify the grade of the jujube. The software can analyze the color, shape, texture and other features of the jujube, and determine the quality grade of the jujube in combination with spectral data.

[0026] The mechanical control execution device 4 is used to transport the target dates to different storage areas based on their grades.

[0027] In an exemplary embodiment, the mechanical control execution device 4 may include a conveyor belt, a sorting mechanical arm or a pneumatic valve, etc., for sorting the jujubes into different storage areas according to their grade. According to the jujube grade information provided by the identification device 3, the action of the mechanical arm or the pneumatic valve is controlled by programming logic to sort the jujubes into designated areas. For example, high-quality jujubes are transported to one area, while low-quality or abnormal jujubes are transported to another area.

[0028] By adopting the above technical solution, since the mixed light source device 1 is used, visible light, ultraviolet light, and infrared light can be emitted in turn to illuminate the jujube in turn. This multi-spectral imaging technology can capture the characteristics of the jujube from different angles and spectral ranges, providing richer information than a single light source. This multi-light source setting helps to more comprehensively identify the surface and internal quality of the jujube, including color, shape, maturity, etc., thereby improving the intelligent level of jujube screening.

[0029] Since the images of jujube are collected by high-definition cameras, these images can reveal different characteristics of jujube under different light sources. High-definition imaging technology allows the subtle features of jujube to be clearly displayed, providing a basis for subsequent precise analysis.

[0030] Since the recognition system is used to analyze the spectral data of the collected images, this analysis can determine the grade of jujube based on the image data. Through machine learning and deep learning technology, the model can be trained to identify the grade characteristics of jujube and realize automatic grade determination.

[0031] According to the grade of the dates determined by the identification system, the mechanical control execution system transports the dates to different storage areas. This automated sorting process reduces manual intervention and improves sorting efficiency and accuracy.

[0032] By using machine vision technology to detect and study the phenotypic characteristics, surface defects, and internal defects of jujubes, the grading quality of jujubes is guaranteed, the competitiveness of the jujube commodity market is enhanced, and the economic benefits of the jujube industry are improved.

[0033] To sum up, the embodiments of the present invention effectively solve the problem of low intelligence level of traditional jujube screening technology through the above-mentioned technical scheme, realize the intelligent selection of jujubes, improve the efficiency and accuracy of jujube screening, and provide a new solution for the automation and intelligence of the jujube industry.

[0034] In one embodiment, Figure 3 is a schematic block diagram of a structure of an identification device according to an embodiment of the present invention. Figure 3 As shown, the identification device 3 includes: Functionalization module 31, used for transforming the spectrum data of each target jujube into continuous functional data to obtain a dynamic function curve; In an exemplary embodiment, the module is responsible for transforming the spectrum data of jujube into continuous functional data to obtain a dynamic function curve. This can be achieved by processing discrete observed data through data smoothing technology and using a function curve feature segmentation extraction algorithm of a statistical depth method. For example, by using mathematical tools such as spline interpolation or Fourier transform, discrete spectral data points are fitted into a continuous function curve, thereby capturing the dynamic change characteristics of jujube spectrum data.

[0035] The feature extraction module 32 is used to extract key features based on the dynamic function curve to obtain the internal structure and change trend of the target jujube; wherein the key features include at least one of the following: statistical moment, shape feature, spectral feature; In an exemplary embodiment, the feature extraction module 32 extracts key features based on the dynamic function curve, such as statistical moments (mean, variance), shape features (curvature, concavity) and spectral features (peak, valley). This can be achieved by extracting functional data features in segments, which can provide solutions to the problems of limited expression of global statistical features and susceptibility of significant point features to noise interference in functional data classification algorithms. Specifically, machine learning algorithms, such as support vector machines (SVM) or partial least squares regression (PLSR) models, can be used to extract key features that are helpful for classification from continuous functional data.

[0036] For example, a support vector machine (SVM) can map data to a high-dimensional space through a kernel function to find the optimal segmentation hyperplane for the data. When processing continuous function data, SVM can identify features that are critical to the classification task.

[0037] 1. Feature mapping: SVM maps the original data to a high-dimensional space through kernel techniques, such as radial basis function (RBF) kernel or polynomial kernel. In this way, even data that is not linearly separable in the original space may be linearly separable in the high-dimensional space.

[0038] 2. Key feature extraction: During the training process, SVM looks for data points closest to the decision boundary, namely support vectors. The features corresponding to these support vectors are actually the key features that are most helpful for classification.

[0039] 3. Classification decision: SVM improves the robustness of classification by maximizing the margin between two classes. In the context of functional data, this means that SVM emphasizes those features that vary the most between different classes.

[0040] In the feature extraction of continuous function data, PLSR extracts key features by establishing the correlation between independent variables and dependent variables.

[0041] 1. Establish correlation: PLSR establishes a prediction model by finding new orthogonal projection directions (principal components) so that the projected dependent variable and independent variable have the largest covariance.

[0042] 2. Feature extraction: The PLSR model reduces dimensionality by extracting principal components, which are linear combinations of the original high-dimensional data and can explain the variability in the data to the greatest extent.

[0043] 3. Prediction and classification: Through the PLSR model, the attribute values ​​of the test set samples can be predicted and classified accordingly. The coefficients (loadings) and scores (scores) in the PLSR model can be regarded as key features because they reflect the contribution of the variables to the prediction target.

[0044] In summary, both SVM and PLSR can extract key features that are helpful for classification from continuous functional data, but their methods and focuses are different. SVM focuses on finding the optimal segmentation hyperplane and support vector, while PLSR focuses on establishing the correlation between variables and extracting principal components. You can choose one of these two methods or use them in combination according to the actual situation.

[0045] An index extraction module 33 is used to determine the spectral abnormality index of the target jujube based on the key features; In an exemplary embodiment, the module determines the spectral abnormality index of the jujube based on the key features. This can be done by building a mathematical model, such as a linear regression model or a logistic regression model, to predict continuous variables or classification problems, thereby determining the abnormality index of the jujube. For example, the extracted peak and valley features can be combined with the physical and chemical indicators of the jujube to establish an abnormality index for subsequent abnormality detection.

[0046] A Bayesian inference network 34 is used to model the spectral abnormality index of the target jujube to determine the abnormal probability of the spectral data of the target jujube; In an exemplary embodiment, the Bayesian reasoning network 34 is used to model the spectral abnormality index of the jujube to determine the abnormal probability of the spectral data of the jujube. This can be done by using the structure of the Bayesian network, combining the spectral data of the jujube and the abnormality index, to perform probabilistic reasoning and prediction. The Bayesian network can process uncertainty data and provide a posterior probability of the abnormal state of the jujube, thereby achieving a quantitative evaluation of the abnormality of the jujube.

[0047] The grading module 35 is used to determine the grade of the target jujube based on the abnormal probability.

[0048] In an exemplary embodiment, the grading module 35 determines the grade of the jujube based on the abnormal probability. This can be done by defuzzifying the abnormal probability determined by the Bayesian inference network 34 to obtain the defuzzified abnormal probability, and then grading the jujube according to these probabilities. For example, a threshold value can be set to categorize the jujube into different grades such as "high quality", "medium quality" and "low quality", and the jujube is transported to different storage areas.

[0049] By adopting the above technical solution, the spectrum data of jujube is converted into continuous function-type data by using the functionalization module 31, which provides a richer and continuous data basis for the feature extraction module 32. The feature extraction module 32 extracts key features such as statistical moments, shape features and spectral features on this basis, which can more accurately reflect the internal structure and change trend of jujube. The synergistic effect of the two modules makes the feature analysis of jujube more accurate and comprehensive.

[0050] The key features provided by the feature extraction module 32 are used by the index extraction module 33 to determine the spectral abnormality index of the jujube. These indexes are calculated based on the key features of the jujube and can quantify the degree of abnormality of the jujube. The synergy of the two modules makes the abnormality detection of the jujube more scientific and reliable.

[0051] The spectral abnormality index determined by the index extraction module 33 is used by the Bayesian reasoning network 34 to model the abnormal probability of the spectral data of the jujube. The Bayesian reasoning network 34 uses these indicators to perform probabilistic reasoning and prediction, and provides the posterior probability of the abnormal state of the jujube. The synergy of the two modules makes the abnormal probability assessment of the jujube more accurate and reliable.

[0052] The abnormal probabilities determined by the Bayesian inference network 34 are used by the grading module 35 to grade the jujubes. The grading module 35 classifies the jujubes according to these probabilities, divides the jujubes into different grades, and transports them to different storage areas. The synergy of the two modules realizes the automated and intelligent grading of jujubes.

[0053] The various modules of the entire recognition device 3 work together to form a complete process from data collection to final classification. The functionalization module 31 converts the spectral data into continuous functional data, the feature extraction module 32 extracts key features, the index extraction module 33 determines the abnormality index, the Bayesian reasoning network 34 calculates the abnormality probability, and finally the classification module 35 performs classification according to the probability. This process not only improves the intelligent level of jujube screening, but also improves the efficiency and accuracy of screening, and provides a new solution for the automation and intelligence of the jujube industry.

[0054] Therefore, through this modular synergy, the jujube sorting machine can achieve efficient and accurate sorting of jujubes, ensuring the consistency and reliability of jujube quality, while also improving the efficiency of jujube processing and sales.

[0055] In one embodiment, Figure 4 is another schematic block diagram of the structure of the identification device according to an embodiment of the present invention. Figure 4 As shown, the identification device 3 also includes: A fuzzy set module 36, for converting the spectral data of the target jujube into a first fuzzy set and a second fuzzy set, and assigning a membership function to each spectral data point in the first fuzzy set and the second fuzzy set, so as to map the spectral data to the first fuzzy set and the second fuzzy set based on the membership function, so as to convert the spectral data into fuzzy spectral data; And, it is used to enable the functionalization module 31 to obtain a fuzzy dynamic function curve based on the fuzzy spectral data.

[0056] In an exemplary embodiment, the fuzzy set module 36 first needs to convert the spectral data of the target jujube into a fuzzy set. Based on the "S-type membership function", "trapezoidal membership function" and "triangular membership function", corresponding membership functions can be defined for the spectral data of the jujube. For example, the sigmoid membership function is used to describe the membership of a certain feature (such as sugar content) in the spectral data of the jujube. This function can map the sugar content value of the jujube to the interval [0,1], indicating the membership degree of the high and low sugar content of the jujube.

[0057] For each spectral data point, the fuzzy set module 36 needs to assign a membership function. This function defines the degree to which the data point belongs to a certain fuzzy set. For example, if a certain spectral feature value of jujube falls within a certain interval, then it belongs to the "high sugar content" set to a high degree; otherwise, it belongs to the "low sugar content" set to a high degree.

[0058] Based on the membership function, the fuzzy set module 36 maps the spectral data to the first fuzzy set and the second fuzzy set. This means that each spectral data point of the jujube will obtain a membership degree belonging to different sets according to its characteristic value. For example, if the spectral data of the jujube indicates that its sugar content is very high, its membership degree belonging to the "high sugar content" set will be very high, while its membership degree belonging to the "low sugar content" set will be very low.

[0059] The functionalization module 31 uses these fuzzy spectral data to obtain a fuzzy dynamic function curve, which means that the dynamic function curve is no longer a single deterministic curve, but a curve composed of multiple fuzzy sets, which can more accurately reflect the uncertainty and fuzziness of the jujube spectral data.

[0060] Due to the combination of fuzzy set theory and machine vision, abnormal conditions of jujube can be identified more accurately. Fuzzy logic can help deal with uncertainty and incomplete information in the data mining process and improve the effect of data mining. This intelligent processing method helps to improve the intelligence level of jujube screening, especially in the detection of surface defects and internal quality assessment of jujube.

[0061] In one embodiment, the fuzzy set module 36 is further configured to enable the feature extraction module 32 to extract fuzzy key features from the fuzzy dynamic function curve.

[0062] In an exemplary embodiment, the feature extraction module 32 extracts key features such as statistical moments, shape features and spectral features from these fuzzified dynamic function curves. These features will contain the membership information of the fuzzy set, so as to more comprehensively describe the spectral characteristics of the jujube.

[0063] In one embodiment, the fuzzy set module 36 is further configured to enable the indicator extraction module 33 to determine a fuzzified spectral abnormality indicator based on the fuzzified key features.

[0064] In an exemplary embodiment, the index extraction module 33 will determine the spectral abnormality index of the jujube based on these fuzzy key features. These indexes will contain abnormal probability information of the jujube spectral data, which is helpful to identify the abnormal situation of the jujube.

[0065] In one embodiment, the fuzzy set module 36 is further used to enable the Bayesian reasoning network 34 to determine the fuzzy abnormality probability based on the fuzzy spectral abnormality index, so that the grading module 35 determines the grade of the target jujube based on the fuzzy abnormality probability.

[0066] In an exemplary embodiment, the Bayesian inference network 34 will determine the abnormal probability of the spectral data of the jujube based on these fuzzy spectral abnormality indicators. This probability will be fuzzy and can more accurately reflect the uncertainty of the spectral data of the jujube.

[0067] In one embodiment, the fuzzy set module 36 is further configured to: The abnormal probability determined by the Bayesian inference network 34 is defuzzified to obtain a defuzzified abnormal probability, so that the grading module 35 determines the grade of the target jujube based on the defuzzified abnormal probability.

[0068] In an exemplary embodiment, the fuzzy set module 36 defuzzifies the abnormal probability determined by the Bayesian inference network 34 to obtain a defuzzified abnormal probability. This step can be implemented using a method such as the Center of Gravity Defuzzification method, so that the grading module 35 can determine the grade of the jujube based on these defuzzified probabilities.

[0069] In one embodiment, the fuzzy set module 36 is further configured to: Determine an evaluation index set and an evaluation grade set of the fuzzy spectrum data of the target jujube; wherein the evaluation index set includes at least one evaluation index, and the evaluation grade set includes at least one evaluation grade; In an exemplary embodiment, the evaluation index set of jujube spectrum data is determined , these indicators may include features such as color, shape, texture, etc. At the same time, define the evaluation level set , such as "normal", "abnormal", etc.

[0070] Performing fuzzy evaluation on each evaluation index one by one to determine the membership degree of the fuzzy spectrum data for each evaluation level to obtain a plurality of first membership degrees; In an exemplary embodiment, for each evaluation indicator Carry out fuzzy evaluation one by one to determine the jujube spectrum data for each evaluation level Membership These memberships can be obtained through expert experience or learning. Where m and n are positive integers. , .

[0071] Fuzzy logic operation is performed on multiple first memberships to determine the comprehensive membership of the target jujube.

[0072] In an exemplary embodiment, the uncertainty and fuzziness of the jujube spectrum data are processed using fuzzy set operation rules, such as fuzzy intersection, union, complement, etc. For example, if a spectral feature of jujube belongs to both "high sugar content" and "low sugar content", its final membership can be determined by fuzzy logic operation.

[0073] By adopting the above technical solution, the spectral data obtained by the mixed light source device 1 is converted into fuzzy spectral data by using the fuzzy set module 36, which enables the system to handle the uncertainty and ambiguity of the surface and internal quality of jujube under different light sources. Through the fuzzy set theory, the spectral data obtained by visible light, ultraviolet light and infrared light can be better understood and interpreted, thereby improving the accuracy of jujube quality assessment.

[0074] The high-definition image data provided by the image acquisition device 2 is converted into fuzzy spectral data through the fuzzy set module 36, which helps to more accurately express the subtle features and potential quality of the jujube in the subsequent feature extraction and analysis. This processing method of the fuzzy set module 36 enhances the expression ability of the image data and provides richer information for the accurate grading of the jujube.

[0075] The functionalization module 31, feature extraction module 32, index extraction module 33 and Bayesian reasoning network 34 in the recognition device 3 all rely on the fuzzy spectrum data provided by the fuzzy set module 36. These modules process data using fuzzy set theory, which can more effectively identify the abnormal conditions and intrinsic quality of jujubes and achieve more accurate jujube grade determination.

[0076] The data processed by the fuzzy set module 36 directly affects the sorting decision of the mechanical control execution device 4. Based on the jujube grade information obtained by the fuzzy set theory, the mechanical control execution device 4 can sort the jujubes into different storage areas more accurately, thereby improving the automation level and efficiency of sorting.

[0077] The use of fuzzy set module 36 plays the role of a bridge and link in the entire jujube selection machine system. It not only improves the flexibility and robustness of data processing, but also enhances the system's ability to comprehensively evaluate the quality of jujubes. Through the processing of fuzzy set module 36, the jujube selection machine can more accurately identify the surface defects and internal quality of jujubes, achieve efficient and intelligent selection, improve the efficiency and accuracy of jujube screening, and provide a new solution for the automation and intelligence of the jujube industry.

[0078] In summary, the fuzzy set module 36 works together with other devices and modules of the jujube selection machine to form an efficient and accurate jujube selection system, making the jujube selection process more intelligent and automated.

[0079] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0080] The embodiment of the present invention also provides a method for selecting jujubes based on machine vision. Figure 5 is a flow chart of a method for selecting jujubes based on machine vision according to an embodiment of the present invention. Figure 5 As shown, the method includes: Step S501, using visible light, ultraviolet light, and infrared light to irradiate the target jujube in turn; In an exemplary embodiment, visible light, ultraviolet light, and infrared light are used to illuminate the target jujube in turn, capturing the characteristics of the jujube from different angles and spectral ranges. This multispectral imaging technology provides richer information than a single light source, and helps to fully identify the surface and internal quality of the jujube, including color, shape, maturity, etc.

[0081] Step S502, collecting the to-be-processed images of the target jujube based on the high-definition camera, wherein the to-be-processed images include images of the target jujube under the irradiation of visible light, ultraviolet light, and infrared light in turn; In an exemplary embodiment, a high-definition industrial camera is used to capture images of jujube, which can reveal different characteristics of jujube under different light sources. High-definition imaging technology allows the subtle features of jujube to be clearly displayed, providing a basis for subsequent accurate analysis.

[0082] Step S503, performing spectral data analysis on the image to be processed, and determining the grade of the target jujube based on the analysis result; In an exemplary embodiment, the obtained image data can be processed using high-performance image processing and analysis software. Through machine learning and deep learning algorithms, such as convolutional neural networks (CNN), the image of the jujube is feature extracted and learned to identify the grade of the jujube. This analysis can determine the grade of the jujube based on the image data and realize automated grade determination.

[0083] Step S504: transport the target dates to different storage areas based on their grades.

[0084] In an exemplary embodiment, manual intervention is reduced through an automated sorting process, thereby improving sorting efficiency and accuracy.

[0085] Through the above steps S501 to S504, the problem of low intelligence level of traditional jujube screening technology is effectively solved, the intelligent selection of jujube is realized, the efficiency and accuracy of jujube screening are improved, and a new solution is provided for the automation and intelligence of the jujube industry.

[0086] In one embodiment, Figure 6 is a flow chart of a method for determining the grade of a target jujube according to an embodiment of the present invention. Figure 6 As shown, the method also includes: Step S601, transforming the spectrum data of each target jujube into continuous function data to obtain a dynamic function curve; In an exemplary embodiment, the spectrum data of jujube can be transformed into continuous function data to obtain a dynamic function curve. This can be achieved by processing discrete observed data through data smoothing technology and using a function curve feature segmentation extraction algorithm of a statistical depth method. For example, by using mathematical tools such as spline interpolation or Fourier transform, discrete spectral data points are fitted into a continuous function curve, thereby capturing the dynamic change characteristics of jujube spectrum data.

[0087] Step S602, extracting key features based on the dynamic function curve to obtain the internal structure and change trend of the target jujube; wherein the key features include at least one of the following: statistical moment, shape feature, spectral feature; In an exemplary embodiment, key features such as statistical moments (mean, variance), shape features (curvature, concavity) and spectral features (peak, valley) can be extracted from dynamic function curves. This can be achieved through an algorithm for segmented extraction of functional data features, which can provide solutions to the problems of limited expression of global statistical features and susceptibility of significant point features to noise interference in functional data classification algorithms. Specifically, machine learning algorithms such as support vector machines (SVM) or partial least squares regression (PLSR) models can be used to extract key features that are helpful for classification from continuous functional data.

[0088] For example, a support vector machine (SVM) can map data to a high-dimensional space through a kernel function to find the optimal segmentation hyperplane for the data. When processing continuous function data, SVM can identify features that are critical to the classification task.

[0089] 1. Feature mapping: SVM maps the original data to a high-dimensional space through kernel techniques, such as radial basis function (RBF) kernel or polynomial kernel. In this way, even data that is not linearly separable in the original space may be linearly separable in the high-dimensional space.

[0090] 2. Key feature extraction: During the training process, SVM looks for data points closest to the decision boundary, namely support vectors. The features corresponding to these support vectors are actually the key features that are most helpful for classification.

[0091] 3. Classification decision: SVM improves the robustness of classification by maximizing the margin between two classes. In the context of functional data, this means that SVM emphasizes those features that vary the most between different classes.

[0092] In the feature extraction of continuous function data, PLSR extracts key features by establishing the correlation between independent variables and dependent variables.

[0093] 1. Establish correlation: PLSR establishes a prediction model by finding new orthogonal projection directions (principal components) so that the projected dependent variable and independent variable have the largest covariance.

[0094] 2. Feature extraction: The PLSR model reduces dimensionality by extracting principal components, which are linear combinations of the original high-dimensional data and can explain the variability in the data to the greatest extent.

[0095] 3. Prediction and classification: Through the PLSR model, the attribute values ​​of the test set samples can be predicted and classified accordingly. The coefficients (loadings) and scores (scores) in the PLSR model can be regarded as key features because they reflect the contribution of the variables to the prediction target.

[0096] In summary, both SVM and PLSR can extract key features that are helpful for classification from continuous functional data, but their methods and focuses are different. SVM focuses on finding the optimal segmentation hyperplane and support vector, while PLSR focuses on establishing the correlation between variables and extracting principal components. You can choose one of these two methods or use them in combination according to the actual situation.

[0097] Step S603, determining the spectral abnormality index of the target jujube based on the key features; In an exemplary embodiment, a mathematical model, such as a linear regression model or a logistic regression model, can be constructed to predict continuous variables or classification problems, thereby determining the abnormality index of jujube. For example, the extracted peak and valley features can be combined with the physical and chemical indicators of jujube to establish an abnormality index for subsequent abnormality detection.

[0098] Step S604, modeling the spectral abnormality index of the target jujube to determine the abnormal probability of the spectral data of the target jujube; In an exemplary embodiment, probabilistic reasoning and prediction can be performed by using the structure of a Bayesian inference network, combining the spectral data of jujube and the abnormality index. The Bayesian inference network can process uncertain data and provide a posterior probability of the abnormal state of jujube, thereby achieving a quantitative assessment of the abnormality of jujube.

[0099] Step S605, determining the grade of the target jujube based on the abnormal probability.

[0100] In an exemplary embodiment, the abnormal probability determined by the Bayesian inference network can be defuzzified to obtain the defuzzified abnormal probability, and then the jujubes can be graded according to the probabilities. For example, a threshold can be set to classify the jujubes into different grades such as "high quality", "medium quality" and "low quality", and the jujubes can be transported to different storage areas.

[0101] In one embodiment, Figure 7 is a flow chart of a method for obtaining a fuzzy dynamic function curve according to an embodiment of the present invention. Figure 7 As shown, the method also includes: Step S701, converting the spectral data of the target jujube into a first fuzzy set and a second fuzzy set, and assigning a membership function to each spectral data point in the first fuzzy set and the second fuzzy set, so as to map the spectral data to the first fuzzy set and the second fuzzy set based on the membership function, so as to convert the spectral data into fuzzy spectral data; Step S702: obtaining a fuzzy dynamic function curve based on the fuzzy spectrum data.

[0102] In an exemplary embodiment, the spectral data of the target jujube is first converted into a fuzzy set. Based on the "S-type membership function", "trapezoidal membership function" and "triangular membership function", corresponding membership functions can be defined for the spectral data of the jujube. For example, a sigmoid membership function is used to describe the membership of a certain feature (such as sugar content) in the spectral data of the jujube. This function can map the sugar content value of the jujube to the interval [0,1], indicating the degree of membership of the high and low sugar content of the jujube.

[0103] A membership function is assigned to each spectral data point. This function defines the degree to which the data point belongs to a fuzzy set. For example, if a spectral feature value of jujube falls within a certain interval, then its degree of membership to the "high sugar content" set is high; otherwise, it belongs to the "low sugar content" set to a high degree.

[0104] The spectral data is mapped to the first fuzzy set and the second fuzzy set based on the membership function. This means that each spectral data point of jujube will obtain a membership degree belonging to different sets according to its eigenvalue. For example, if the spectral data of jujube indicates that its sugar content is very high, its membership degree belonging to the "high sugar content" set will be high, while its membership degree belonging to the "low sugar content" set will be low.

[0105] The fuzzified spectral data are used to obtain the fuzzified dynamic function curve, which means that the dynamic function curve is no longer a single deterministic curve, but a curve composed of multiple fuzzy sets, which can more accurately reflect the uncertainty and fuzziness of the jujube spectral data.

[0106] In one embodiment, the method further includes: extracting fuzzy key features from the fuzzy dynamic function curve.

[0107] In an exemplary embodiment, key features such as statistical moments, shape features and spectral features are extracted from these fuzzified dynamic function curves. These features will contain the membership information of the fuzzy set, so as to more comprehensively describe the spectral characteristics of jujube.

[0108] In one embodiment, the method further includes: determining a fuzzy spectral abnormality index based on the fuzzy key features.

[0109] In an exemplary embodiment, the spectral abnormality index of the jujube is determined based on these fuzzy key features. These indexes will contain abnormal probability information of the jujube spectral data, which is helpful to identify the abnormal situation of the jujube.

[0110] In one embodiment, the method further includes: determining a fuzzy abnormality probability based on the fuzzy spectral abnormality index, and determining the grade of the target jujube based on the fuzzy abnormality probability.

[0111] In an exemplary embodiment, the abnormal probability of the spectral data of jujube is determined based on these fuzzy spectral abnormality indicators. This probability will be fuzzy and can more accurately reflect the uncertainty of the spectral data of jujube.

[0112] In one embodiment, the method further includes: defuzzifying the abnormal probability determined by the Bayesian inference network to obtain a defuzzified abnormal probability, and determining the grade of the target jujube based on the defuzzified abnormal probability.

[0113] In an exemplary embodiment, the abnormal probability determined by the Bayesian inference network is defuzzified to obtain a defuzzified abnormal probability. This step can be implemented using a method such as the Center of Gravity Defuzzification method, so that the grade of the jujube can be determined based on these defuzzified probabilities.

[0114] In one embodiment, Figure 8 is a flow chart of a method for determining the comprehensive membership of a target jujube according to an embodiment of the present invention. Figure 8 As shown, the method also includes: Step S801, determining an evaluation index set and an evaluation grade set of fuzzy spectrum data of a target jujube; wherein the evaluation index set includes at least one evaluation index, and the evaluation grade set includes at least one evaluation grade; In an exemplary embodiment, the evaluation index set of jujube spectrum data is determined , these indicators may include features such as color, shape, texture, etc. At the same time, define the evaluation level set , such as "normal", "abnormal", etc.

[0115] Step S802, performing fuzzy evaluation on each evaluation index one by one to determine the membership degree of the fuzzy spectrum data for each evaluation level to obtain a plurality of first membership degrees; In an exemplary embodiment, for each evaluation indicator Carry out fuzzy evaluation one by one to determine the jujube spectrum data for each evaluation level Membership These memberships can be obtained through expert experience or learning. Where m and n are positive integers. , .

[0116] Step S803, performing fuzzy logic operation on the multiple first membership degrees to determine the comprehensive membership degree of the target jujube.

[0117] In an exemplary embodiment, the uncertainty and fuzziness of the jujube spectrum data are processed using fuzzy set operation rules, such as fuzzy intersection, union, complement, etc. For example, if a spectral feature of jujube belongs to both "high sugar content" and "low sugar content", its final membership can be determined by fuzzy logic operation.

[0118] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by adding the necessary general hardware platform with the help of software, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0119] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0120] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0121] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0122] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0123] An embodiment of the present invention further provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method described in each embodiment of the present invention.

[0124] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0125] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A jujube selection machine based on machine vision, characterized in that: include: A mixed light source device (1), comprising a visible light source module (11), an ultraviolet light source module (12), and an infrared light source module (13), and used for irradiating target jujubes in turn based on the visible light source module (11), the ultraviolet light source module (12), and the infrared light source module (13); An image acquisition device (2) is used to acquire the image to be processed of the target jujube based on a high-definition camera, wherein the image to be processed comprises an image of the target jujube irradiated by the visible light source module (11), the ultraviolet light source module (12), and the infrared light source module (13) in turn; An identification device (3) is used to perform spectral data analysis on the image to be processed, and determine the grade of the target jujube based on the analysis result; A mechanical control execution device (4) is used to transport the target dates to different storage areas based on the grades.

2. The jujube selection machine according to claim 1, characterized in that: The identification device (3) comprises: A functionalization module (31) is used to transform the spectral data of each target jujube into continuous functional data to obtain a dynamic function curve; A feature extraction module (32) is used to extract key features based on the dynamic function curve to obtain the internal structure and change trend of the target jujube; wherein the key features include at least one of the following: statistical moment, shape feature, spectral feature; An index extraction module (33) is used to determine the spectral abnormality index of the target jujube based on the key features; A Bayesian inference network (34) is used to model the spectral abnormality index of the target jujube to determine the abnormal probability of the spectral data of the target jujube; A grading module (35) is used to determine the grade of the target jujube based on the abnormal probability.

3. The jujube selection machine according to claim 2, characterized in that: The identification device (3) further comprises: A fuzzy set module (36) is used to convert the spectral data of the target jujube into a first fuzzy set and a second fuzzy set, and assign a membership function to each spectral data point in the first fuzzy set and the second fuzzy set, so as to map the spectral data to the first fuzzy set and the second fuzzy set based on the membership function, so as to convert the spectral data into fuzzy spectral data; And, used to enable the functionalization module (31) to obtain a fuzzy dynamic function curve based on the fuzzy spectral data.

4. The jujube selection machine according to claim 3, characterized in that: The fuzzy set module (36) is also used to enable the feature extraction module (32) to extract fuzzy key features from the fuzzy dynamic function curve.

5. The jujube selection machine according to claim 4, characterized in that: The fuzzy set module (36) is also used to enable the indicator extraction module (33) to determine a fuzzified spectral abnormality indicator based on the fuzzified key feature.

6. The jujube selection machine according to claim 5, characterized in that: The fuzzy set module (36) is also used to enable the Bayesian inference network (34) to determine the fuzzy abnormality probability based on the fuzzy spectral abnormality index, so that the grading module (35) determines the grade of the target jujube based on the fuzzy abnormality probability.

7. The jujube selection machine according to claim 6, characterized in that: The fuzzy set module (36) is also used to: The abnormal probability determined by the Bayesian inference network (34) is defuzzified to obtain a defuzzified abnormal probability, so that the grading module (35) determines the grade of the target jujube based on the defuzzified abnormal probability.

8. The jujube selection machine according to claim 7, characterized in that: The fuzzy set module (36) is also used to: Determine an evaluation index set and an evaluation grade set of the fuzzy spectrum data of the target jujube; wherein the evaluation index set includes at least one evaluation index, and the evaluation grade set includes at least one evaluation grade; Performing fuzzy evaluation on each of the evaluation indicators one by one to determine the membership of the fuzzy spectrum data to each of the evaluation levels to obtain a plurality of first memberships; A fuzzy logic operation is performed on the multiple first memberships to determine the comprehensive membership of the target jujube.

9. A method for selecting jujubes based on machine vision, characterized in that: include: Use visible light, ultraviolet light, and infrared light to irradiate the target jujube in turn; Collecting the target jujube images to be processed based on a high-definition camera, wherein the target jujube images to be processed include images of the target jujube under the irradiation of visible light, ultraviolet light, and infrared light in turn; Performing spectral data analysis on the image to be processed, and determining the grade of the target jujube based on the analysis result; The target dates are transported to different storage areas based on the grades.

10. A jujube selection system based on machine vision, characterized in that: include: The jujube selection machine according to any one of claims 1 to 8; The fine material storage device is used to store target jujubes that meet a preset grade, wherein the target jujubes are moved into the fine material storage device by the mechanical control execution device (4).

Images