Hyperspectral imaging and computer vision fused Chinese wolfberry fruit quality grading system

Abstract

The invention discloses a hyperspectral imaging and computer vision fused Chinese wolfberry fruit quality grading system, and relates to the field of Chinese wolfberry fruit quality grading, and the system comprises a data collection module which is used for setting a Chinese wolfberry fruit quality preliminary grading device and a hyperspectral imaging collection device; the feature extraction module is used for grading the wolfberry quality features and establishing a wolfberry quality grading list; the data optimization module is used for establishing a wolfberry characteristic hyperspectral optimization model and improving the wolfberry characteristic hyperspectral recognizability; the feature analysis module is used for analyzing the size, saturation, integrity, color and component features of the Chinese wolfberry fruits; the comprehensive evaluation module is used for establishing a wolfberry quality grading comprehensive evaluation model and dividing different grades of wolfberries; and the platform display module is used for setting a wolfberry quality grading display platform and monitoring, displaying and recording shunting conditions of different grades of wolfberries. The processing efficiency and reliability of quality grading of the Chinese wolfberry fruits are effectively improved, and the complexity of quality grading of the Chinese wolfberry fruits is reduced.

Description

Technical Field

[0001] The present invention relates to the field of wolfberry quality grading, and in particular to a wolfberry quality grading system integrating hyperspectral imaging and computer vision. Background Art

[0002] Through wolfberry quality grading technology, consumers can clearly understand the quality level of wolfberries and avoid selling inferior products as good ones. At the same time, it can greatly improve producers' control over wolfberry quality and improve their ability to position wolfberry prices. However, wolfberry quality grading is a complex process, which mainly requires controlling the size, saturation, integrity, color and composition of wolfberries. Detailed analysis of these characteristics can help better help wolfberry quality grading. With the rapid development of hyperspectral imaging and computer vision fusion technology, combining hyperspectral imaging and computer vision fusion technology with wolfberry quality grading can greatly improve the efficiency of wolfberry quality grading, which is of great significance.

[0003] Although existing wolfberry quality grading technology involves the fusion of hyperspectral imaging and computer vision, this type of technology is subject to the problems of wolfberry separation and uniform distribution, making it difficult to move from the laboratory to the production workshop and handle large-scale wolfberry quality grading. Secondly, if the existing technology is used in an environment with insufficient spectral resolution or uneven light source, it is difficult to distinguish subtle spectral differences, such as slight mold and normal areas, resulting in false detection or missed detection during the detection process, thereby judging normal wolfberries as abnormal or failing to identify the abnormality, which ultimately affects the reliability and accuracy of wolfberry quality grading. Summary of the Invention

[0004] In order to solve the above technical problems, a wolfberry quality grading system integrating hyperspectral imaging and computer vision is provided. This technical solution solves the problems of wolfberry separation and uniform distribution raised in the above background technology, which makes it difficult to move from the laboratory to the production workshop and handle a large number of wolfberries for quality grading. Secondly, if the existing technology is in an environment with insufficient spectral resolution or uneven light source, it is difficult to distinguish subtle spectral differences, such as slight mold and normal areas, resulting in false detection or missed detection during the detection process, thereby judging normal wolfberries as abnormal or failing to identify the abnormality, which ultimately affects the reliability and accuracy of wolfberry quality grading.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A wolfberry quality grading system that integrates hyperspectral imaging and computer vision, including:

[0007] A data acquisition module, wherein the data acquisition module is used to set up a preliminary wolfberry quality grading device by using a channel transmission screening network, and to set up a hyperspectral imaging acquisition device at the diversion point of the transmission screening network;

[0008] A feature extraction module is used to grade the quality of wolfberries according to size, saturation, integrity, color and composition characteristics based on big data, and to establish a wolfberry quality grading list;

[0009] A data optimization module is used to establish a wolfberry characteristic hyperspectral optimization model based on the wolfberry characteristic hyperspectral acquisition curve graph and the standard curve graph generated by the hyperspectral imaging acquisition device to improve the recognizability of the wolfberry characteristic hyperspectral;

[0010] A feature analysis module is used to extract the spectral optimization data values corresponding to different wolfberry characteristics based on the wolfberry characteristic hyperspectral optimization data values, and analyze the size, saturation, integrity, color and composition characteristics of the wolfberry;

[0011] A comprehensive evaluation module is used to establish a comprehensive evaluation model for wolfberry quality grading, and to sort wolfberries of different grades based on the analysis results of wolfberry size, saturation, integrity, color, and composition characteristics;

[0012] The platform display module is used to set up a wolfberry quality grading display platform, to monitor and display the diversion of wolfberries of different grades, and to record the processing data of the diversion of wolfberries of different grades.

[0013] Preferably, the method of using a channel transmission screening net to set up a preliminary wolfberry quality grading device and setting up a hyperspectral imaging acquisition device at the diversion point of the transmission screening net specifically includes:

[0014] A multi-layer channel transmission and screening net is set up to divert and preliminarily screen the piles of wolfberries through multi-layer, vibration and limited hole methods.

[0015] The multi-level, vibration and pore-limiting methods for diversion and preliminary screening specifically include:

[0016] Set up multiple filters with holes at fixed distances, arrange them vertically from large to small according to the size of the filter holes, and layer them;

[0017] According to the layered filter screens, filters with holes at fixed intervals are installed on the conveyor belt at each layer to build a multi-layer channel transmission and screening network. Each layer of the channel transmission and screening network is composed of multiple individual filter screens arranged in parallel. The area of the filter screen is determined by the width of the channel transmission and screening network and the acquisition range of the hyperspectral imaging acquisition device.

[0018] A vibration device is installed under or at both ends of the conveyor belt, which can make the multi-layer channel conveyor screening net vibrate up and down and left and right at a certain frequency;

[0019] At the entrance of the multi-layer channel conveying and screening net, a limited channel for the entry of wolfberries is set to ensure that the wolfberries can be evenly and thinly covered on the multi-layer channel conveying and screening net;

[0020] At the exit of the multi-layer channel transmission and screening network, a hyperspectral imaging acquisition device and a hyperspectral imaging acquisition stationary platform are set up to collect spectral data of the filtered wolfberries transmitted by the multi-layer channel transmission and screening network;

[0021] According to the size of wolfberries and the limitation of the filter holes, the wolfberries are separated and preliminarily screened;

[0022] The specific steps of collecting the spectral data of the filtered wolfberry transmitted by the multi-layer channel transmission and screening network include:

[0023] The individual filter screens are conveyed through a multi-layer channel conveying screening net to a stationary stage for hyperspectral imaging acquisition;

[0024] When all the filters conveyed by the multi-layer channel transmission screening network reach the hyperspectral imaging acquisition stationary platform, the conveyor belt stops conveying and waits for a fixed time before continuing to convey, wherein the fixed time is the time for wolfberry spectrum acquisition and data processing;

[0025] The collection, diversion and preliminary screening of wolfberry spectral data are completed in a cycle.

[0026] Preferably, the method of grading wolfberry quality based on big data according to size, saturation, integrity, color and composition characteristics, and establishing a wolfberry quality grading list specifically includes:

[0027] Based on big data, the size, saturation, integrity, color and composition characteristics of wolfberry are extracted;

[0028] According to the spectral values corresponding to the size, saturation, integrity, color and composition characteristics of wolfberry, the quality grades corresponding to different spectral value ranges are determined;

[0029] A quality grading list for wolfberries was established, which detailed the correspondence between the size, saturation, integrity, color and composition characteristics, spectral values and the quality grading of wolfberries.

[0030] Preferably, the data optimization module specifically includes:

[0031] An acquisition data processing unit, configured to generate multiple sets of wolfberry characteristic hyperspectral acquisition data cubes based on multiple sets of wolfberry characteristic spectral data acquired by the hyperspectral imaging acquisition device;

[0032] A standard data processing unit, the standard data processing unit is used to obtain standard data of the wolfberry characteristic spectral data based on a control experiment, and to establish a wolfberry characteristic hyperspectral standard data cube;

[0033] A data segmentation unit, which is used to segment the data cube by controlling a single variable, extract the spectral curve at a specific position, and establish a wolfberry characteristic hyperspectral acquisition curve graph and a standard curve graph;

[0034] A data optimization unit, the data optimization unit is used to filter the wolfberry characteristic hyperspectral acquisition curve and the wolfberry characteristic hyperspectral standard curve to reduce the influence of noise and peaks and improve the smoothing quality of the image;

[0035] A reference data unit, which is used to obtain reference data of dark current and background noise of the acquisition device, as well as reference data of light source non-uniformity and response difference of the acquisition device based on big data or test experiments;

[0036] The optimization model unit is used to establish a wolfberry characteristic hyperspectral optimization model, optimize wolfberry characteristic spectral data, and improve the recognizability of wolfberry characteristic hyperspectral data.

[0037] Preferably, the wolfberry characteristic hyperspectral optimization model expression is:

[0038]

[0039] Where R is the optimized data value of wolfberry characteristic hyperspectral, g z is the standard data value of wolfberry characteristic hyperspectral, L is the brightness compensation value, g is the collected data value of wolfberry characteristic hyperspectral, δ is the balance weight, I d is the reference data of dark current and background noise of the acquisition device, I w It is the reference data for the non-uniformity of light source and the response difference of acquisition device.

[0040] Preferably, the feature analysis module specifically includes:

[0041] A data acquisition unit, configured to acquire all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral acquisition data cube and optimized data values of the wolfberry characteristic hyperspectral according to the whole-body scanning result of the wolfberry by the hyperspectral imaging acquisition device;

[0042] A two-dimensional image acquisition unit, configured to acquire a two-dimensional image of the entire wolfberry according to all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral data cube;

[0043] a size and integrity judgment unit, configured to judge the size and integrity of the wolfberries based on the area and boundary curvature of the wolfberry's overall two-dimensional image;

[0044] The analysis model unit is used to optimize the data value according to the wolfberry characteristic hyperspectral, establish a wolfberry characteristic hyperspectral analysis model, and analyze the saturation, color and composition characteristics of wolfberry.

[0045] Preferably, the method for determining the size and integrity of wolfberries specifically includes:

[0046] Based on big data, we obtain the area value range of the two-dimensional image of wolfberries at different wolfberry grades, and set the size classification thresholds of wolfberries at different grades respectively.

[0047] Determine whether the area of the entire two-dimensional image of the wolfberry is larger than the wolfberry size classification threshold of the current grade. If so, it means that the size of the wolfberry meets the current grade. If not, it means that the size of the wolfberry does not meet the current grade and needs to flow to the next layer for judgment;

[0048] Based on big data, a threshold value for the change rate of the boundary curve of the overall two-dimensional image of wolfberries is set. The actual change rate of the boundary curve of the overall two-dimensional image of wolfberries is judged to be greater than the threshold. If so, it means that the integrity of the wolfberries is unqualified and needs to be eliminated. If not, it means that the integrity of the wolfberries meets the standard.

[0049] The wolfberry characteristic hyperspectral analysis model expression is:

[0050]

[0051] Where, γ i is the weight of the i-th feature, is the optimized data value of the wolfberry feature hyperspectral corresponding to the jth position of the i-th feature, N is the number of all positions corresponding to the i-th feature, that is, the number of all coordinates of the wolfberry feature hyperspectral acquisition data cube spatial dimension, α i is the standard judgment threshold, a i is the minimum value of the hyperspectral optimization data value corresponding to the wolfberry feature i, b i The maximum value of the hyperspectral optimized data value corresponding to the wolfberry feature i.

[0052] Preferably, the establishment of a comprehensive evaluation model for wolfberry quality grading, combining the analysis results of wolfberry size, saturation, integrity, color and component characteristics, and sorting wolfberries of different grades specifically includes:

[0053] Based on the analysis results of wolfberry's size, saturation, integrity, color and composition characteristics, a sample training set for the wolfberry quality grading and comprehensive evaluation model was established;

[0054] According to the evaluation results of wolfberry characteristic hyperspectral standard data, the target training set of wolfberry quality grading comprehensive evaluation model was established;

[0055] Based on the Pearson correlation coefficient, the effects of wolfberry size, saturation, integrity, color and composition characteristics on wolfberry quality were determined.

[0056] The correlation between wolfberry size, saturation, integrity, color and composition characteristics and wolfberry quality was used as the weight value of each characteristic in the comprehensive evaluation model of wolfberry quality grading;

[0057] According to the sample training set, target training set, weight value of each feature and wolfberry quality grading list of the wolfberry quality grading comprehensive evaluation model, a wolfberry quality grading comprehensive evaluation model was established based on the machine learning algorithm. The quality grading of wolfberries was comprehensively evaluated, the quality grade of each wolfberry was determined, and wolfberries of different grades were diverted.

[0058] Preferably, the wolfberry quality grading display platform is provided to monitor and display the diversion of wolfberries of different grades, and record the processing data of the diversion of wolfberries of different grades, specifically including:

[0059] Set up a wolfberry quality grading display platform to build the operating environment of the wolfberry quality grading comprehensive evaluation model and ensure the normal operation of the system;

[0060] Based on the wolfberry quality grading display platform, it is used to monitor and display the diversion of wolfberries of different grades, and provide feedback and early warning of abnormal situations during wolfberry quality grading;

[0061] Based on the wolfberry quality grading display platform, the IoT technology is used to receive and store the processing data of wolfberries of different grades, and mark the data of wolfberries that do not meet the standards;

[0062] Based on the wolfberry quality grading display platform, it is used to update and optimize the evaluation results of the wolfberry quality grading comprehensive evaluation model to improve system reliability and system applicability.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] By utilizing the channel transmission screening net, a preliminary wolfberry quality grading device is set up, and the piles of wolfberries are diverted and preliminarily screened through multi-level, vibration and limited hole methods, and a hyperspectral imaging acquisition device is set at the diversion point of the transmission screening net to collect the spectral data of the filtered wolfberries transmitted by the multi-layer channel transmission screening net. Secondly, through the data optimization module, according to the wolfberry characteristic hyperspectral acquisition curve and standard curve generated by the hyperspectral imaging acquisition device, a wolfberry characteristic hyperspectral optimization model is established to optimize the wolfberry characteristic spectral data and improve the recognizability of the wolfberry characteristic hyperspectral. Furthermore, through the feature analysis module, according to the wolfberry characteristic hyperspectral optimization data value, the spectral optimization data values corresponding to different wolfberry characteristics are extracted respectively, and by setting the threshold and establishing the wolfberry characteristic hyperspectral analysis model , the size, saturation, integrity, color and composition characteristics of wolfberry were analyzed, wolfberry that met the single condition standard was screened out, and wolfberry that did not meet the standard was eliminated. Finally, based on the Pearson correlation coefficient, the influence of the size, saturation, integrity, color and composition characteristics of wolfberry on the quality of wolfberry was determined, and the correlation between the size, saturation, integrity, color and composition characteristics of wolfberry and the quality of wolfberry was used as the weight value of each feature in the comprehensive evaluation model of wolfberry quality grading. Based on the machine learning algorithm, a comprehensive evaluation model of wolfberry quality grading was established, which comprehensively evaluated the quality grading of wolfberry, determined the quality grade of each wolfberry, and diverted wolfberries of different grades, thereby effectively improving the processing efficiency and reliability of wolfberry quality grading and reducing the complexity of wolfberry quality grading. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a structural block diagram of a wolfberry quality grading system that integrates hyperspectral imaging and computer vision according to the present invention;

[0066] Figure 2 The present invention uses a channel transmission screening network to set up a preliminary wolfberry quality grading device, and sets a hyperspectral imaging acquisition device at the diversion point of the transmission screening network.

[0067] Figure 3 A flowchart for establishing a wolfberry characteristic hyperspectral optimization model based on the wolfberry characteristic hyperspectral acquisition curve graph and standard curve graph generated by the hyperspectral imaging acquisition device of the present invention to improve the recognizability of the wolfberry characteristic hyperspectral;

[0068] Figure 4 This is a flowchart of the invention that optimizes the data values of the wolfberry characteristics hyperspectrally, extracts the spectral optimization data values corresponding to different wolfberry characteristics, and analyzes the size, saturation, integrity, color and composition characteristics of the wolfberry. DETAILED DESCRIPTION

[0069] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0070] Reference Figure 1 As shown in the figure, a wolfberry quality grading system integrating hyperspectral imaging and computer vision includes:

[0071] A data acquisition module, wherein the data acquisition module is used to set up a preliminary wolfberry quality grading device by using a channel transmission screening network, and to set up a hyperspectral imaging acquisition device at the diversion point of the transmission screening network;

[0072] A feature extraction module is used to grade the quality of wolfberries according to size, saturation, integrity, color and composition characteristics based on big data, and to establish a wolfberry quality grading list;

[0073] A data optimization module is used to establish a wolfberry characteristic hyperspectral optimization model based on the wolfberry characteristic hyperspectral acquisition curve graph and the standard curve graph generated by the hyperspectral imaging acquisition device to improve the recognizability of the wolfberry characteristic hyperspectral;

[0074] A feature analysis module is used to extract the spectral optimization data values corresponding to different wolfberry characteristics based on the wolfberry characteristic hyperspectral optimization data values, and analyze the size, saturation, integrity, color and composition characteristics of the wolfberry;

[0075] A comprehensive evaluation module is used to establish a comprehensive evaluation model for wolfberry quality grading, and to sort wolfberries of different grades based on the analysis results of wolfberry size, saturation, integrity, color, and composition characteristics;

[0076] The platform display module is used to set up a wolfberry quality grading display platform, to monitor and display the diversion of wolfberries of different grades, and to record the processing data of the diversion of wolfberries of different grades.

[0077] Reference Figure 1 As shown, the data optimization module specifically includes:

[0078] An acquisition data processing unit, configured to generate multiple sets of wolfberry characteristic hyperspectral acquisition data cubes based on multiple sets of wolfberry characteristic spectral data acquired by the hyperspectral imaging acquisition device;

[0079] A standard data processing unit, the standard data processing unit is used to obtain standard data of the wolfberry characteristic spectral data based on a control experiment, and to establish a wolfberry characteristic hyperspectral standard data cube;

[0080] A data segmentation unit, which is used to segment the data cube by controlling a single variable, extract the spectral curve at a specific position, and establish a wolfberry characteristic hyperspectral acquisition curve graph and a standard curve graph;

[0081] A data optimization unit, the data optimization unit is used to filter the wolfberry characteristic hyperspectral acquisition curve and the wolfberry characteristic hyperspectral standard curve to reduce the influence of noise and peaks and improve the smoothing quality of the image;

[0082] A reference data unit, which is used to obtain reference data of dark current and background noise of the acquisition device, as well as reference data of light source non-uniformity and response difference of the acquisition device based on big data or test experiments;

[0083] The optimization model unit is used to establish a wolfberry characteristic hyperspectral optimization model, optimize wolfberry characteristic spectral data, and improve the recognizability of wolfberry characteristic hyperspectral data.

[0084] Reference Figure 1 As shown, the feature analysis module specifically includes:

[0085] A data acquisition unit, configured to acquire all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral acquisition data cube and optimized data values of the wolfberry characteristic hyperspectral according to the whole-body scanning result of the wolfberry by the hyperspectral imaging acquisition device;

[0086] A two-dimensional image acquisition unit, configured to acquire a two-dimensional image of the entire wolfberry according to all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral data cube;

[0087] a size and integrity judgment unit, configured to judge the size and integrity of the wolfberries based on the area and boundary curvature of the wolfberry's overall two-dimensional image;

[0088] The analysis model unit is used to optimize the data value according to the wolfberry characteristic hyperspectral, establish a wolfberry characteristic hyperspectral analysis model, and analyze the saturation, color and composition characteristics of wolfberry.

[0089] It can be explained that this scheme sets up a preliminary grading device for wolfberry quality by utilizing a channel transmission screening net, diverts and preliminarily screens the piles of wolfberries through multi-level, vibration and hole-limiting methods, and sets up a hyperspectral imaging acquisition device at the diversion point of the transmission screening net to collect the spectral data of the filtered wolfberries transmitted by the multi-layer channel transmission screening net. Secondly, through the data optimization module, according to the wolfberry characteristic hyperspectral acquisition curve and standard curve generated by the hyperspectral imaging acquisition device, a wolfberry characteristic hyperspectral optimization model is established to optimize the wolfberry characteristic spectral data and improve the recognizability of the wolfberry characteristic hyperspectrum. Furthermore, through the feature analysis module, according to the wolfberry characteristic hyperspectral optimization data value, the spectral optimization data values corresponding to different wolfberry characteristics are extracted respectively. By setting the threshold and establishing the wolfberry characteristic hyperspectral Spectral analysis model was used to analyze the size, saturation, integrity, color and composition characteristics of wolfberry, screen out wolfberry that met the single condition standard, and eliminate wolfberry that did not meet the standard. Finally, based on the Pearson correlation coefficient, the effects of the size, saturation, integrity, color and composition characteristics of wolfberry on the quality of wolfberry were determined, and the correlation between the size, saturation, integrity, color and composition characteristics of wolfberry and the quality of wolfberry were used as the weight value of each feature in the comprehensive evaluation model of wolfberry quality grading. Based on the machine learning algorithm, a comprehensive evaluation model of wolfberry quality grading was established to comprehensively evaluate the quality grading of wolfberry, determine the quality grade of each wolfberry, and divert wolfberries of different grades, thereby effectively improving the processing efficiency and reliability of wolfberry quality grading and reducing the complexity of wolfberry quality grading.

[0090] Reference Figure 2 As shown, the method of using a channel transmission screening net to set up a preliminary wolfberry quality grading device and setting a hyperspectral imaging acquisition device at the diversion point of the transmission screening net specifically includes:

[0091] A multi-layer channel transmission and screening net is set up to divert and preliminarily screen the piles of wolfberries through multi-layer, vibration and limited hole methods.

[0092] The multi-level, vibration and pore-limiting methods for diversion and preliminary screening specifically include:

[0093] Set up multiple filters with holes at fixed distances, arrange them vertically from large to small according to the size of the filter holes, and layer them;

[0094] According to the layered filter screens, filters with holes at fixed intervals are installed on the conveyor belt at each layer to build a multi-layer channel transmission and screening network. Each layer of the channel transmission and screening network is composed of multiple individual filter screens arranged in parallel. The area of the filter screen is determined by the width of the channel transmission and screening network and the acquisition range of the hyperspectral imaging acquisition device.

[0095] A vibration device is installed under or at both ends of the conveyor belt, which can make the multi-layer channel conveyor screening net vibrate up and down and left and right at a certain frequency;

[0096] At the entrance of the multi-layer channel conveying and screening net, a limited channel for the entry of wolfberries is set to ensure that the wolfberries can be evenly and thinly covered on the multi-layer channel conveying and screening net;

[0097] At the exit of the multi-layer channel transmission and screening network, a hyperspectral imaging acquisition device and a hyperspectral imaging acquisition stationary platform are set up to collect spectral data of the filtered wolfberries transmitted by the multi-layer channel transmission and screening network;

[0098] According to the size of wolfberries and the limitation of the filter holes, the wolfberries are separated and preliminarily screened;

[0099] The specific steps of collecting the spectral data of the filtered wolfberry transmitted by the multi-layer channel transmission and screening network include:

[0100] The individual filter screens are conveyed through a multi-layer channel conveying screening net to a stationary stage for hyperspectral imaging acquisition;

[0101] When all the filters conveyed by the multi-layer channel transmission screening network reach the hyperspectral imaging acquisition stationary platform, the conveyor belt stops conveying and waits for a fixed time before continuing to convey, wherein the fixed time is the time for wolfberry spectrum acquisition and data processing;

[0102] The collection, diversion and preliminary screening of wolfberry spectral data are completed in a cycle.

[0103] It can be explained that the traditional wolfberry diversion or screening is too complicated, or requires a lot of manpower, resulting in low efficiency and difficulty in maintaining uniform distribution of wolfberries when separating them. Therefore, it is easy to be affected by the accumulation of wolfberries during hyperspectral imaging acquisition, resulting in incomplete wolfberry detection, and some wolfberries cannot be detected, resulting in deviations in the final evaluation results of wolfberry quality grading. Therefore, this solution provides a wolfberry diversion solution, which uses a multi-layer channel transmission screening network to collect, divert and preliminarily screen the wolfberry spectral data. At the same time, it can also ensure that the hyperspectral imaging acquisition device ensures the stability and light sensitivity of wolfberries when collecting wolfberry spectral data, thereby greatly ensuring the accuracy and reliability of wolfberry spectral data acquisition.

[0104] Reference Figure 3 As shown, the wolfberry characteristic hyperspectral acquisition curve and standard curve generated by the hyperspectral imaging acquisition device are used to establish a wolfberry characteristic hyperspectral optimization model to improve the recognizability of the wolfberry characteristic hyperspectral. Specifically, the following steps are performed:

[0105] Generate multiple sets of wolfberry characteristic hyperspectral data cubes based on multiple sets of wolfberry characteristic spectral data collected by the hyperspectral imaging acquisition device;

[0106] Based on the control experiment, the standard data of these wolfberry characteristic spectral data are obtained, and the wolfberry characteristic hyperspectral standard data cube is established;

[0107] By controlling a single variable, the data cube is segmented, the spectral curve at a specific location is extracted, and the characteristic hyperspectral acquisition curve and standard curve of wolfberry are established.

[0108] Filter the wolfberry characteristic hyperspectral acquisition curve and wolfberry characteristic hyperspectral standard curve to reduce the influence of noise and peaks and improve the smoothing quality of the image;

[0109] Based on big data or test experiments, obtain reference data on the dark current and background noise of the acquisition device, as well as reference data on light source non-uniformity and acquisition device response differences;

[0110] Establish a wolfberry characteristic hyperspectral optimization model, optimize wolfberry characteristic spectral data, and improve the recognizability of wolfberry characteristic hyperspectral data;

[0111] The wolfberry characteristic hyperspectral optimization model expression is:

[0112]

[0113] Where R is the optimized data value of wolfberry characteristic hyperspectral, g z is the standard data value of wolfberry characteristic hyperspectral, L is the brightness compensation value, g is the collected data value of wolfberry characteristic hyperspectral, δ is the balance weight, I d is the reference data of dark current and background noise of the acquisition device, I w It is the reference data for the non-uniformity of light source and the response difference of acquisition device.

[0114] It can be explained that wolfberry is easily affected by the dark current and background noise of the acquisition device, as well as the unevenness of the light source and the response difference of the acquisition device during hyperspectral imaging acquisition. Especially in an environment with insufficient spectral resolution or uneven light source, it is difficult to distinguish subtle spectral differences, such as slight mold and normal areas, resulting in false detection or missed detection during the detection process, thereby judging normal wolfberry as abnormal or failing to identify the abnormality, which ultimately affects the reliability and accuracy of wolfberry quality grading. Therefore, this scheme obtains wolfberry characteristic hyperspectral acquisition data and wolfberry characteristic hyperspectral standard data, establishes a wolfberry characteristic hyperspectral optimization model, optimizes wolfberry characteristic spectral data through the model, improves the recognizability of wolfberry characteristic hyperspectra, thereby effectively improving the practicality of wolfberry hyperspectral imaging acquisition, greatly reducing the impact of light source and other factors, and thus improving the accuracy of the final judgment of wolfberry quality grading.

[0115] Reference Figure 4 As shown, the method of extracting the spectrum optimization data values corresponding to different wolfberry characteristics based on the wolfberry characteristic hyperspectral optimization data values and analyzing the size, saturation, integrity, color and composition characteristics of wolfberry specifically includes:

[0116] According to the whole-body scanning results of wolfberry by the hyperspectral imaging acquisition device, all coordinate information of the spatial dimension of the wolfberry characteristic hyperspectral acquisition data cube and the optimized data value of the wolfberry characteristic hyperspectral are obtained;

[0117] According to the hyperspectral data of wolfberry, all coordinate information of the spatial dimension of the data cube is collected to obtain the overall two-dimensional image of wolfberry;

[0118] The size and integrity of wolfberries are judged by the area and boundary curvature of the overall two-dimensional image of wolfberries;

[0119] Based on the wolfberry characteristic hyperspectral optimization data value, a wolfberry characteristic hyperspectral analysis model was established to analyze the saturation, color and composition characteristics of wolfberry;

[0120] The method for judging the size and integrity of wolfberries specifically includes:

[0121] Based on big data, we obtain the area value range of the two-dimensional image of wolfberries at different wolfberry grades, and set the size classification thresholds of wolfberries at different grades respectively.

[0122] Determine whether the area of the entire two-dimensional image of the wolfberry is larger than the wolfberry size classification threshold of the current grade. If so, it means that the size of the wolfberry meets the current grade. If not, it means that the size of the wolfberry does not meet the current grade and needs to flow to the next layer for judgment;

[0123] Based on big data, a threshold value for the change rate of the boundary curve of the overall two-dimensional image of wolfberries is set. The actual change rate of the boundary curve of the overall two-dimensional image of wolfberries is judged to be greater than the threshold. If so, it means that the integrity of the wolfberries is unqualified and needs to be eliminated. If not, it means that the integrity of the wolfberries meets the standard.

[0124] The wolfberry characteristic hyperspectral analysis model expression is:

[0125]

[0126] Where, γ i is the weight of the i-th feature, is the optimized data value of the wolfberry feature hyperspectral corresponding to the jth position of the i-th feature, N is the number of all positions corresponding to the i-th feature, that is, the number of all coordinates of the wolfberry feature hyperspectral acquisition data cube spatial dimension, α i is the standard judgment threshold, ai is the minimum value of the hyperspectral optimization data value corresponding to the wolfberry feature i, b i The maximum value of the hyperspectral optimized data value corresponding to the wolfberry feature i.

[0127] It can be explained that by collecting all the coordinate information of the spatial dimension of the data cube according to the wolfberry characteristic hyperspectral data, the overall two-dimensional image of the wolfberry is obtained, and the area and boundary curvature of the overall two-dimensional image of the wolfberry are determined, so that the wolfberries that meet the wolfberry size and integrity conditions are screened out by setting a threshold. Secondly, by establishing a wolfberry characteristic hyperspectral analysis model, the saturation, color and composition characteristics of the wolfberry are analyzed. By analyzing the proportion of the number of positions of the wolfberry characteristic hyperspectral optimized data values that meet these three characteristics in the total number of positions, by setting a threshold, the wolfberries that meet the single condition standard are screened out, and the wolfberries that do not meet the standard are eliminated, thereby improving the single condition wolfberry screening and ensuring the reliability and accuracy of the single condition wolfberry screening results.

[0128] In summary, the advantages of the present invention are: effectively improving the processing efficiency and reliability of wolfberry quality grading and reducing the complexity of wolfberry quality grading.

[0129] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A wolfberry quality grading system integrating hyperspectral imaging and computer vision, characterized in that: include: A data acquisition module, wherein the data acquisition module is used to set up a preliminary wolfberry quality grading device by using a channel transmission screening network, and to set up a hyperspectral imaging acquisition device at the diversion point of the transmission screening network; A feature extraction module is used to grade the quality of wolfberries according to size, saturation, integrity, color and composition characteristics based on big data, and to establish a wolfberry quality grading list; A data optimization module is used to establish a wolfberry characteristic hyperspectral optimization model based on the wolfberry characteristic hyperspectral acquisition curve graph and the standard curve graph generated by the hyperspectral imaging acquisition device to improve the recognizability of the wolfberry characteristic hyperspectral; A feature analysis module is used to extract the spectral optimization data values corresponding to different wolfberry characteristics based on the wolfberry characteristic hyperspectral optimization data values, and analyze the size, saturation, integrity, color and composition characteristics of the wolfberry; A comprehensive evaluation module is used to establish a comprehensive evaluation model for wolfberry quality grading, and to sort wolfberries of different grades based on the analysis results of wolfberry size, saturation, integrity, color, and composition characteristics; The platform display module is used to set up a wolfberry quality grading display platform, to monitor and display the diversion of wolfberries of different grades, and to record the processing data of the diversion of wolfberries of different grades.

2. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 1 is characterized in that: The method of using a channel transmission screening net to set up a preliminary wolfberry quality grading device and setting up a hyperspectral imaging acquisition device at the diversion point of the transmission screening net specifically includes: A multi-layer channel transmission and screening net is set up to divert and preliminarily screen the piles of wolfberries through multi-layer, vibration and limited hole methods. The multi-level, vibration and pore-limiting methods for diversion and preliminary screening specifically include: Set up multiple filters with holes at fixed distances, arrange them vertically from large to small according to the size of the filter holes, and layer them; According to the layered filter screens, filters with holes at fixed intervals are installed on the conveyor belt at each layer to build a multi-layer channel transmission and screening network. Each layer of the channel transmission and screening network is composed of multiple individual filter screens arranged in parallel. The area of the filter screen is determined by the width of the channel transmission and screening network and the acquisition range of the hyperspectral imaging acquisition device. A vibration device is installed under or at both ends of the conveyor belt, which can make the multi-layer channel conveyor screening net vibrate up and down and left and right at a certain frequency; At the entrance of the multi-layer channel conveying and screening net, a limited channel for the entry of wolfberries is set to ensure that the wolfberries can be evenly and thinly covered on the multi-layer channel conveying and screening net; At the exit of the multi-layer channel transmission and screening network, a hyperspectral imaging acquisition device and a hyperspectral imaging acquisition stationary platform are set up to collect spectral data of the filtered wolfberries transmitted by the multi-layer channel transmission and screening network; According to the size of wolfberries and the limitation of the filter holes, the wolfberries are separated and preliminarily screened; The specific steps of collecting the spectral data of the filtered wolfberry transmitted by the multi-layer channel transmission and screening network include: The individual filter screens are conveyed through a multi-layer channel conveying screening net to a stationary stage for hyperspectral imaging acquisition; When all the filters conveyed by the multi-layer channel transmission screening network reach the hyperspectral imaging acquisition stationary platform, the conveyor belt stops conveying and waits for a fixed time before continuing to convey, wherein the fixed time is the time for wolfberry spectrum acquisition and data processing; The collection, diversion and preliminary screening of wolfberry spectral data are completed in a cycle.

3. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 2 is characterized in that: Based on big data, wolfberry quality is graded according to size, saturation, integrity, color and composition characteristics, and a wolfberry quality grading list is established, specifically including: Based on big data, the size, saturation, integrity, color and composition characteristics of wolfberry are extracted; According to the spectral values corresponding to the size, saturation, integrity, color and composition characteristics of wolfberry, the quality grades corresponding to different spectral value ranges are determined; A quality grading list for wolfberries was established, which detailed the correspondence between the size, saturation, integrity, color and composition characteristics, spectral values and the quality grading of wolfberries.

4. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 3 is characterized in that: The data optimization module specifically includes: An acquisition data processing unit, configured to generate multiple sets of wolfberry characteristic hyperspectral acquisition data cubes based on multiple sets of wolfberry characteristic spectral data acquired by the hyperspectral imaging acquisition device; A standard data processing unit, the standard data processing unit is used to obtain standard data of the wolfberry characteristic spectral data based on a control experiment, and to establish a wolfberry characteristic hyperspectral standard data cube; A data segmentation unit, which is used to segment the data cube by controlling a single variable, extract the spectral curve at a specific position, and establish a wolfberry characteristic hyperspectral acquisition curve graph and a standard curve graph; A data optimization unit, the data optimization unit is used to filter the wolfberry characteristic hyperspectral acquisition curve and the wolfberry characteristic hyperspectral standard curve to reduce the influence of noise and peaks and improve the smoothing quality of the image; A reference data unit, which is used to obtain reference data of dark current and background noise of the acquisition device, as well as reference data of light source non-uniformity and response difference of the acquisition device based on big data or test experiments; The optimization model unit is used to establish a wolfberry characteristic hyperspectral optimization model, optimize wolfberry characteristic spectral data, and improve the recognizability of wolfberry characteristic hyperspectral data.

5. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 4 is characterized in that: The wolfberry characteristic hyperspectral optimization model expression is: Where R is the optimized data value of wolfberry characteristic hyperspectral, g z is the standard data value of wolfberry characteristic hyperspectral, L is the brightness compensation value, g is the collected data value of wolfberry characteristic hyperspectral, δ is the balance weight, I d is the reference data of dark current and background noise of the acquisition device, I w It is the reference data for the non-uniformity of light source and the response difference of acquisition device.

6. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 5, characterized in that: The feature analysis module specifically includes: A data acquisition unit, configured to acquire all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral acquisition data cube and optimized data values of the wolfberry characteristic hyperspectral according to the whole-body scanning result of the wolfberry by the hyperspectral imaging acquisition device; A two-dimensional image acquisition unit, configured to acquire a two-dimensional image of the entire wolfberry according to all coordinate information of the spatial dimensions of the wolfberry characteristic hyperspectral data cube; a size and integrity judgment unit, configured to judge the size and integrity of the wolfberries based on the area and boundary curvature of the wolfberry's overall two-dimensional image; The analysis model unit is used to optimize the data value according to the wolfberry characteristic hyperspectral, establish a wolfberry characteristic hyperspectral analysis model, and analyze the saturation, color and composition characteristics of wolfberry.

7. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 6, characterized in that: The method for judging the size and integrity of wolfberries specifically includes: Based on big data, we obtain the area value range of the two-dimensional image of wolfberries at different wolfberry grades, and set the size classification thresholds of wolfberries at different grades respectively. Determine whether the area of the entire two-dimensional image of the wolfberry is larger than the wolfberry size classification threshold of the current grade. If so, it means that the size of the wolfberry meets the current grade. If not, it means that the size of the wolfberry does not meet the current grade and needs to flow to the next layer for judgment; Based on big data, a threshold value for the change rate of the boundary curve of the overall two-dimensional image of wolfberries is set. The actual change rate of the boundary curve of the overall two-dimensional image of wolfberries is judged to be greater than the threshold. If so, it means that the integrity of the wolfberries is unqualified and needs to be eliminated. If not, it means that the integrity of the wolfberries meets the standard. The wolfberry characteristic hyperspectral analysis model expression is: Where, γ i is the weight of the i-th feature, is the optimized data value of the wolfberry feature hyperspectral corresponding to the jth position of the i-th feature, N is the number of all positions corresponding to the i-th feature, that is, the number of all coordinates of the wolfberry feature hyperspectral acquisition data cube spatial dimension, α i is the standard judgment threshold, a i is the minimum value of the hyperspectral optimization data value corresponding to the wolfberry feature i, b i The maximum value of the hyperspectral optimized data value corresponding to the wolfberry feature i.

8. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 7, characterized in that: The establishment of a comprehensive evaluation model for wolfberry quality grading, combined with the analysis results of wolfberry size, saturation, integrity, color and composition characteristics, to sort wolfberries of different grades specifically includes: Based on the analysis results of wolfberry's size, saturation, integrity, color and composition characteristics, a sample training set for the wolfberry quality grading and comprehensive evaluation model was established; According to the evaluation results of wolfberry characteristic hyperspectral standard data, the target training set of wolfberry quality grading comprehensive evaluation model was established; Based on the Pearson correlation coefficient, the effects of wolfberry size, saturation, integrity, color and composition characteristics on wolfberry quality were determined. The correlation between wolfberry size, saturation, integrity, color and composition characteristics and wolfberry quality was used as the weight value of each characteristic in the comprehensive evaluation model of wolfberry quality grading; According to the sample training set, target training set, weight value of each feature and wolfberry quality grading list of the wolfberry quality grading comprehensive evaluation model, a wolfberry quality grading comprehensive evaluation model was established based on the machine learning algorithm. The quality grading of wolfberries was comprehensively evaluated, the quality grade of each wolfberry was determined, and wolfberries of different grades were diverted.

9. The wolfberry quality grading system integrating hyperspectral imaging and computer vision according to claim 8, characterized in that: The wolfberry quality grading display platform is set up to monitor and display the diversion of wolfberries of different grades and record the processing data of the diversion of wolfberries of different grades, specifically including: Set up a wolfberry quality grading display platform to build the operating environment of the wolfberry quality grading comprehensive evaluation model and ensure the normal operation of the system; Based on the wolfberry quality grading display platform, it is used to monitor and display the diversion of wolfberries of different grades, and provide feedback and early warning of abnormal situations during wolfberry quality grading; Based on the wolfberry quality grading display platform, the IoT technology is used to receive and store the processing data of wolfberries of different grades, and mark the data of wolfberries that do not meet the standards; Based on the wolfberry quality grading display platform, it is used to update and optimize the evaluation results of the wolfberry quality grading comprehensive evaluation model to improve system reliability and system applicability.

Images