Agricultural product distribution automatic identification system and distribution method thereof

Abstract

The invention discloses an agricultural product distribution automatic identification system and a distribution method thereof, and relates to the technical field of agricultural product identification and distribution, and the system comprises a basic information module which is used for obtaining basic identification features of agricultural products; the information screening module is used for obtaining visual data, spectral data and texture data; the visual identification module is used for performing first identification on the agricultural products according to the difference quantity to obtain a first identification result; the spectrum identification module is used for carrying out second identification on the agricultural product according to the internal quality and the maturity to obtain a second identification result; the texture recognition module is used for performing third recognition on the agricultural products according to the product texture data and the packaging texture data to obtain a third recognition result; and the general management distribution module is used for distributing the agricultural products according to the final identification result. The method has the effects of improving the accuracy in the automatic identification process of the agricultural products and reducing the adverse effects caused by conditions such as time, stains and damage.

Description

Technical Field

[0001] This application relates to the technical field of agricultural product identification and distribution, and in particular to an automatic identification system for agricultural product distribution and its distribution method. Background Art

[0002] Agricultural products refer to primary products derived from agriculture, that is, plants, animals, microorganisms and their products obtained in agricultural activities, including both edible and non-edible aspects. For example, rice, peanuts, corn, as well as vegetables, fruits, poultry, aquatic products, and so on.

[0003] The current automatic identification system for agricultural product distribution can only identify agricultural products based on the preset external characteristics of each agricultural product or the pre-provided barcodes, etc. Different agricultural products have different properties, and the external characteristics of some agricultural products gradually change over time, which will lead to a decrease in the accuracy of the automatic identification system for agricultural product distribution during the identification process. At the same time, there are also problems such as non-standard sticking, stains, damage of the barcode labels of agricultural products and interference from environmental information, resulting in inaccurate information being recognized. Summary of the Invention

[0004] The purpose of the present invention is to provide an automatic identification system for agricultural product distribution and its distribution method to solve the problems raised in the above background art.

[0005] In a first aspect, this application provides an automatic identification system for agricultural product distribution, and the system includes: Basic Information Module: used to obtain the basic external information and packaging information of agricultural products, and obtain the basic identification characteristics of agricultural products according to the basic external information and the packaging information; Information Screening Module: used to obtain the visual information, spectral information and texture information of agricultural products, and perform information screening on the visual information, spectral information and texture information based on the basic external information to obtain visual data, spectral data and texture data; Visual Recognition Module: used to analyze the visual data according to the visual data, obtain the difference amount between different individuals in the agricultural products, and perform the first identification on the agricultural products according to the difference amount to obtain the first identification result; Spectral Recognition Module: used to analyze the spectral data according to the spectral data, obtain the internal quality and maturity of agricultural products, and perform the second identification on the agricultural products according to the internal quality and the maturity to obtain the second identification result; Texture Recognition Module: used to perform refined analysis on the texture data according to the texture data, respectively obtain the product texture data and packaging texture data of agricultural products, and perform the third identification on the agricultural products according to the product texture data and the packaging texture data to obtain the third identification result; Prime delivery module: used to combine the first recognition result, the second recognition result, and the third recognition result to obtain the final recognition result of agricultural products, and deliver the agricultural products according to the final recognition result.

[0006] Preferably, the step of obtaining the basic appearance information and packaging information of agricultural products and obtaining the basic recognition features of agricultural products according to the basic appearance information and the packaging information is specifically as follows: Take pictures and samples of the whole process of agricultural products from the breeding ground to the completion of packaging and then to decay to obtain a full-line photo sample; Extract the appearance color information, basic spectral information, and basic texture information of the agricultural products themselves in the photo sample; Combine the appearance color information, basic spectral information, and the basic texture information to obtain the basic appearance information of agricultural products; Extract the basic packaging information of the agricultural products in the photo sample, and subdivide the basic packaging information to obtain packaging material information and barcode information; Obtain the packaging information of agricultural products according to the packaging material information and the barcode information, combine the basic appearance information and the packaging information, and perform feature extraction to obtain basic recognition features.

[0007] Preferably, the step of performing information screening on the visual information, the spectral information, and the texture information based on the basic appearance information to obtain visual data, spectral data, and texture data is specifically as follows: According to the basic appearance information, obtain the body data of agricultural products, and perform background separation on the visual information, the spectral information, and the texture information according to the body data to obtain body visual information, body spectral information, and body texture information; Add different lighting conditions to the basic appearance information and generate an appearance information set under different lighting conditions; Screen the body visual information, the body spectral information, and the body texture information according to the appearance information set to obtain visual data, spectral data, and texture data.

[0008] Preferably, the step of analyzing the visual data to obtain the difference amount between different individuals in agricultural products and performing the first recognition on agricultural products according to the difference amount to obtain the first recognition result is specifically as follows: Based on the visual data, obtain the shape data and color data of each agricultural product individual, and obtain the appearance data of each agricultural product individual according to the shape data and the color data; Compare the appearance data corresponding to each individual agricultural product one by one to obtain multiple difference data values, and perform weighted average calculation on the difference data values to obtain the difference amount between different individuals; Determine whether the difference amount is greater than a preset difference threshold; If it is determined that the difference amount is greater than the difference threshold, perform a first identification on the agricultural product according to the shape data and the color data to obtain a first identification result; If it is determined that the difference amount is less than or equal to the difference threshold, input the shape data and the color data into a convolutional neural network for learning to obtain a learned target convolutional neural network; Call the target convolutional neural network to perform a first identification on the agricultural product to obtain a first identification result.

[0009] Preferably, the step of inputting the shape data and the color data into a convolutional neural network for learning to obtain a learned target convolutional neural network is specifically as follows: Perform data normalization processing on the shape data and the color data of each individual agricultural product to obtain processed shape data and processed color data with the same data size; Perform data augmentation on the processed shape data by rotating, flipping, scaling, and cropping to generate a shape-varied data set; Perform data augmentation on the processed color data by color transformation to generate a color-varied data set; Input the shape-varied data set and the color-varied data set into a convolutional neural network for learning to obtain a learned target convolutional neural network.

[0010] Preferably, the step of analyzing the spectral data to obtain the internal quality and maturity of the agricultural product and performing a second identification on the agricultural product according to the internal quality and the maturity to obtain a second identification result is specifically as follows: Based on the spectral data, identify the baseline drift in the spectral data by polynomial fitting and delete the baseline drift to obtain target spectral data; Establish a spectral absorption characteristic comparison table, and perform band division on the target spectral data based on the spectral absorption characteristic comparison table to obtain a quality band spectrum and a maturity band spectrum; Use independent component analysis to perform dimensionality reduction processing on the quality band spectrum and the maturity band spectrum, and extract quality spectral features and maturity spectral features; Obtain the internal quality and maturity of the agricultural product based on the quality spectral features and the maturity spectral features; Perform a second identification of the agricultural products based on the internal quality and the maturity level to obtain a second identification result.

[0011] Preferably, the step of performing a second identification of the agricultural products based on the internal quality and the maturity level to obtain a second identification result is specifically as follows: Establish a quality standard range for the agricultural products and divide the standard range to form multiple segments of standard ranges; Obtain the internal chemical composition of the agricultural products based on the internal quality, and obtain the first quality parameter of the agricultural products based on the internal chemical composition; Obtain the physical characteristics and cell composition of the agricultural products based on the maturity level, and obtain the second quality parameter of the agricultural products based on the physical characteristics and the cell composition; Obtain the quality parameter of the agricultural products based on the first quality parameter and the second quality parameter, and perform calibration identification on the quality parameter according to the multiple segments of standard ranges to obtain a second identification result.

[0012] Preferably, the step of performing a refinement analysis on the texture data to separately obtain the product texture data and the packaging texture data of the agricultural products, and performing a third identification of the agricultural products based on the product texture data and the packaging texture data to obtain a third identification result is specifically as follows: Perform a refinement analysis on the texture data to separately obtain the product texture data and the packaging texture data of the agricultural products; Based on the product texture data, obtain the texture orientation, texture depth, and texture interference of each individual agricultural product; Perform multi-modal fitting on the texture orientation, the texture depth, and the texture interference to obtain the texture feature of each individual agricultural product, and perform weighted averaging on the texture features of each individual agricultural product to obtain the product weighted texture feature of the agricultural products; Based on the packaging texture data, obtain the barcode texture data and the packaging bag texture data of each packaging, perform weighted averaging on the packaging bag texture data of multiple packagings to obtain the packaging weighted texture feature, and mark each barcode texture data on each packaging to obtain a barcode mark; Perform a third identification of the agricultural products based on the product weighted texture feature and the packaging weighted texture feature, and add a barcode mark after the identification to obtain a third identification result.

[0013] In a second aspect, the present application provides an automatic identification and distribution method for agricultural product distribution, and the method includes: Obtain the basic appearance information and packaging information of the agricultural products, and obtain the basic identification features of the agricultural products based on the basic appearance information and the packaging information; Obtain the visual information, spectral information, and texture information of agricultural products, and perform information screening on the visual information, spectral information, and texture information based on the basic appearance information to obtain visual data, spectral data, and texture data; According to the visual data, analyze the visual data to obtain the difference amount between different individuals in the agricultural products, and perform the first identification on the agricultural products according to the difference amount to obtain the first identification result; According to the spectral data, analyze the spectral data to obtain the internal quality and maturity of the agricultural products, and perform the second identification on the agricultural products according to the internal quality and maturity to obtain the second identification result; According to the texture data, perform refined analysis on the texture data to respectively obtain the product texture data and packaging texture data of the agricultural products, and perform the third identification on the agricultural products according to the product texture data and the packaging texture data to obtain the third identification result; Combine the first identification result, the second identification result, and the third identification result to obtain the final identification result of the agricultural products, and distribute the agricultural products according to the final identification result.

[0014] In summary, the present application includes at least one of the following beneficial technical effects: By obtaining the basic appearance information and packaging information of the agricultural products to be distributed, generating the basic identification features of the agricultural products; then obtaining the visual information, spectral information, and texture information of the agricultural products to be identified, and screening the above three information contents with the basic identification features as the screening conditions to obtain visual data, spectral data, and texture data. By comparing the visual data between agricultural products, the difference amount between each agricultural product individual is obtained, and then different identification methods are selected according to the difference amount to obtain the first identification result of the agricultural products; by the spectral data of the agricultural products, the internal quality and maturity of each agricultural product during distribution are obtained to perform the second identification on the agricultural products to obtain the second identification result; by the texture data of the agricultural products, the product weighted texture features of the agricultural products themselves, the packaging weighted texture features on the packaging bags, and the barcode marks on each packaging are obtained, and the third identification is performed according to the above two features and one mark to obtain the third identification result. Summing up the above three identification results to obtain the final identification result of the agricultural products, and distributing the agricultural products according to the final identification result. The accuracy in the automatic identification process of agricultural products is improved, and the adverse effects caused by situations such as time, stains, and damage are reduced. Brief Description of the Drawings

[0015] Figure 1 is a module block diagram of an automatic identification system for agricultural product distribution provided by an embodiment of the present application; Figure 2It is a flowchart of the steps of an automatic recognition and distribution method for agricultural product distribution provided by an embodiment of the present application.

[0016] Explanation of reference numerals: 1. Basic information module; 2. Information screening module; 3. Visual recognition module; 4. Spectral recognition module; 5. Texture recognition module; 6. Prime distribution module. Detailed implementation manners

[0017] The following combines Figure 1 - Figure 2 to further elaborate on the present application in detail, but the implementation manners of the present invention are not limited thereto.

[0018] An embodiment of the present application discloses an automatic recognition system for agricultural product distribution and its distribution method.

[0019] In this embodiment, an automatic recognition system for agricultural product distribution, the system includes: Basic information module: used to obtain the basic appearance information and packaging information of agricultural products, and obtain the basic recognition features of agricultural products according to the basic appearance information and packaging information; Information screening module: used to obtain the visual information, spectral information and texture information of agricultural products, and perform information screening on the visual information, spectral information and texture information based on the basic appearance information to obtain visual data, spectral data and texture data; Visual recognition module: used to analyze the visual data according to the visual data, obtain the difference amount between different individuals in the agricultural products, and perform the first recognition on the agricultural products according to the difference amount to obtain the first recognition result; Spectral recognition module: used to analyze the spectral data according to the spectral data, obtain the internal quality and maturity of agricultural products, and perform the second recognition on the agricultural products according to the internal quality and maturity to obtain the second recognition result; Texture recognition module: used to refine and analyze the texture data according to the texture data, respectively obtain the product texture data and packaging texture data of agricultural products, and perform the third recognition on the agricultural products according to the product texture data and packaging texture data to obtain the third recognition result; Prime distribution module: used to combine the first recognition result, the second recognition result and the third recognition result to obtain the final recognition result of agricultural products, and distribute the agricultural products according to the final recognition result.

[0020] It should be noted that the above modules are only the basic modules of this embodiment. In the specific implementation process, without affecting the overall implementation effect, some modules can be appropriately added, reduced or modified.

[0021] The steps of obtaining the basic appearance information and packaging information of agricultural products, and obtaining the basic recognition features of agricultural products according to the basic appearance information and packaging information are specifically as follows: Take photos and samples of the entire process of agricultural products from breeding to packaging and then to decay, and obtain full-line photo samples; Extract the appearance color information, basic spectrum information and basic texture information of the agricultural products in the photo samples; Combining appearance color information, basic spectrum information and basic texture information to obtain basic appearance information of agricultural products; Extract the basic packaging information of agricultural products in the photo samples, and subdivide the basic packaging information to obtain packaging material information and barcode information; The packaging information of agricultural products is obtained based on the packaging material information and barcode information, and the basic appearance information and packaging information are combined and feature extraction is performed to obtain basic recognition features.

[0022] In application, taking the automatic identification and distribution of carrots as an example, the whole process of carrot harvesting, packaging in the workshop, and stacking some carrots until they rot is photographed and sampled, and photo samples of the whole process of carrot harvesting to rotting are obtained. According to these photo samples, the appearance color information of the carrots themselves is red, dark red, light red, orange, light orange, etc. The basic spectrum information is the spectrum information of each carrot in each time period of the whole process. The basic texture information includes the depth of the grain on each carrot, the direction of the grain, and the grain changes of the carrots in different time periods. According to the above information and data content, the basic appearance information of the carrots is obtained. Then, the photos of the carrots when they are packaged are obtained. After identifying the photos, the packaging of the carrots is 6 pieces per bag, which is packaged in transparent plastic bags and the barcode information on the packaging bags. Feature extraction is performed based on the above content to obtain the basic identification features of the carrots.

[0023] The steps of filtering visual information, spectral information and texture information based on basic appearance information to obtain visual data, spectral data and texture data are specifically as follows: According to the basic appearance information, the ontology data of the agricultural product is obtained, and the background of the visual information, the spectral information and the texture information is separated according to the ontology data to obtain the ontology visual information, the ontology spectral information and the ontology texture information; Add different lighting conditions to the basic appearance information and generate appearance information sets under different lighting conditions; The entity visual information, entity spectral information and entity texture information are screened according to the appearance information set to obtain visual data, spectral data and texture data.

[0024] In operation, taking the automatic identification and distribution of carrots as an example, content extraction is performed on the obtained visual information, spectral information, and texture information of carrots according to the appearance color information, basic spectral information, and basic texture information of carrots in the basic appearance information of carrots. The useless background information is removed, and the obtained body visual information, body spectral information, and body texture information that only contain carrots are obtained. Then, different lighting conditions are added to the basic appearance information of carrots, including weak light to strong light, from red light to purple light, from monochromatic light to seven-color light, etc., and an appearance information set under different lighting conditions is collected. According to this information set, the body visual information, body spectral information, and body texture information are screened to obtain visual data, spectral data, and texture data.

[0025] The steps of analyzing the visual data to obtain the difference amount between different individuals in agricultural products and performing the first identification of agricultural products according to the difference amount to obtain the first identification result are specifically as follows: Based on the visual data, the shape data and color data of each agricultural product individual are obtained, and the appearance data of each agricultural product individual are obtained according to the shape data and color data; The appearance data corresponding to each agricultural product individual are compared one by one to obtain a plurality of difference data values, and the weighted average calculation is performed on the difference data values to obtain the difference amount between different individuals; Judge whether the difference amount is greater than a preset difference threshold; If it is judged that the difference amount is greater than the difference threshold, the agricultural products are first identified according to the shape data and color data to obtain the first identification result; If it is judged that the difference amount is less than or equal to the difference threshold, the shape data and color data are input into a convolutional neural network for learning to obtain the learned target convolutional neural network; The target convolutional neural network is called to perform the first identification of agricultural products to obtain the first identification result.

[0026] In operation, taking the automatic identification and distribution of carrots as an example, shape data of each carrot individual is obtained based on visual data, with some being long and some short, some thick and some thin; color data, mostly orange-red, with some having a darker color and some a lighter color, and some even having black notches, etc.; the appearance data of each carrot is obtained based on the shape data and color data. The appearance data of each carrot is compared one by one to obtain multiple sets of difference data values. For example, if 4 carrots are compared, 6 sets of difference data values are obtained. These difference data values are weighted and averaged to obtain the difference values between different carrot individuals, and the preset difference threshold is 5. If the difference value is 6, that is, the difference value is greater than the difference threshold, the carrots are directly identified based on the shape data and color data to obtain the first identification result. If the difference value is 3, that is, the difference value is less than the difference threshold, all the shape data and color data are input into the convolutional neural network for learning to obtain the target convolutional neural network after learning, and then the target convolutional neural network is used to perform the first identification on the carrots to obtain the first identification result.

[0027] The steps of inputting the shape data and color data into the convolutional neural network for learning to obtain the target convolutional neural network after learning are specifically as follows: The shape data and color data of each agricultural product individual are subjected to data normalization processing to obtain the to-be-processed shape data and to-be-processed color data with the same data size; The to-be-processed shape data is rotated, flipped, scaled, and cropped for data augmentation to generate a dataset with diverse shapes; The to-be-processed color data is subjected to color transformation for data augmentation to generate a dataset with diverse colors; The dataset with diverse shapes and the dataset with diverse colors are input into the convolutional neural network for learning to obtain the target convolutional neural network after learning.

[0028] In operation, taking the automatic identification and distribution of carrots as an example, the shape data and color data of each carrot are subjected to data normalization processing to keep each data at the same size and type. Then, operations such as rotation, flipping, scaling, and cropping are performed on the shape data, so that one data can be expanded into N data to generate a dataset with diverse shapes. At the same time, the color of one carrot is subjected to color transformation to obtain M data to generate a dataset with diverse colors. The above two datasets are input into the convolutional neural network to allow the convolutional neural network to have sufficient sample data for learning to obtain the target convolutional neural network after learning.

[0029] The steps of analyzing the spectral data to obtain the internal quality and maturity of the agricultural product and performing the second identification on the agricultural product based on the internal quality and maturity to obtain the second identification result are specifically as follows: Based on spectral data, identify the baseline drift in the spectral data through polynomial fitting, and delete the baseline drift to obtain the target spectral data; Establish a spectral absorption characteristic comparison table, and divide the target spectral data based on the spectral absorption characteristic comparison table to obtain the quality band spectrum and the maturity band spectrum; Use independent component analysis to perform dimensionality reduction on the quality band spectrum and the maturity band spectrum, and extract the quality spectral features and the maturity spectral features; Obtain the internal quality and maturity of the agricultural product based on the quality spectral features and the maturity spectral features; Perform a second identification on the agricultural product according to the internal quality and maturity to obtain the second identification result.

[0030] In application, taking the automatic identification and distribution of carrots as an example, the baseline drift in the spectral data of carrots is fitted through polynomial fitting, and then the baseline drift is deleted to obtain the target spectral data. Relevant data is obtained through the network and a spectral absorption characteristic comparison table is established. According to this comparison table, the target spectral data of carrots is divided into bands to obtain four quality band spectra and four maturity band spectra, corresponding to four stages of carrots, namely high-quality fresh products, medium-quality sub-fresh products, low-quality non-fresh products, and poor-quality rotten products. Then, independent component analysis is used to perform dimensionality reduction on the quality band spectrum and the maturity band spectrum of carrots to obtain the quality spectral features and the maturity spectral features. Then, according to the above content, the internal quality and maturity of carrots are obtained, and a second identification is performed on carrots according to the internal quality and maturity to obtain the second identification result.

[0031] The steps of performing a second identification on the agricultural product according to the internal quality and maturity to obtain the second identification result are specifically as follows: Establish a quality standard interval for the agricultural product, and divide the standard interval to form multiple standard intervals; Obtain the internal chemical composition of the agricultural product according to the internal quality, and obtain the first quality parameter of the agricultural product according to the internal chemical composition; Obtain the physical characteristics and cell composition of the agricultural product according to the maturity, and obtain the second quality parameter of the agricultural product according to the physical characteristics and cell composition; Obtain the quality parameter of the agricultural product according to the first quality parameter and the second quality parameter, and calibrate and identify the quality parameter according to the multiple standard intervals to obtain the second identification result.

[0032] In operation, taking the automatic identification and distribution of carrots as an example, a high-standard interval for agricultural products is established through data such as the standard quality parameters of carrots, and the standard interval is divided into four standard intervals, namely high-quality fresh products, medium-quality semi-fresh products, low-quality non-fresh products, and poor-quality rotten products. Then, according to the internal quality, the internal chemical composition of carrots is obtained, including chemical compositions such as water, sugar, and organic acids. Based on these chemical compositions, the first quality parameters of agricultural products are high moisture, high sugar, and sufficient organic matter. Then, according to the physical characteristics and cell composition of carrots, that is, the appearance color, texture, etc. of carrots, the second quality parameters of carrots are obtained as orange-red, relatively hard texture, plant cell composition, etc. Then, based on the above content, the quality parameters of carrots are obtained, and the quality parameters are substituted into the multi-segment standard interval to calibrate and identify the quality parameters, and the second identification result of high-quality fresh products is obtained.

[0033] The steps of performing a refined analysis on the texture data to separately obtain the product texture data and packaging texture data of the agricultural product, and performing a third identification on the agricultural product based on the product texture data and packaging texture data to obtain the third identification result are as follows: Perform a refined analysis on the texture data to separately obtain the product texture data and packaging texture data of the agricultural product; Based on the product texture data, obtain the texture direction, texture depth, and texture interference of each agricultural product individual; Perform multi-modal fitting on the texture direction, texture depth, and texture interference to obtain the texture characteristics of each agricultural product individual, and perform weighted averaging on the texture characteristics of each agricultural product individual to obtain the product weighted texture characteristics of the agricultural product; Based on the packaging texture data, obtain the barcode texture data and packaging bag texture data of each packaging, perform weighted averaging on the packaging bag texture data of multiple packagings to obtain the packaging weighted texture characteristics, and mark each barcode texture data on each packaging to obtain barcode markings; Perform a third identification on the agricultural product according to the product weighted texture characteristics and packaging weighted texture characteristics, and add barcode markings after the identification to obtain the third identification result.

[0034] In operation, taking the automatic identification and distribution of carrots as an example, the texture data is analyzed to obtain the product texture data of carrots and the packaging texture data of the packaging bags. According to the product texture data, it is obtained that the texture direction of each carrot is in circles, the texture depth ranges from 0.1 mm to 0.3 mm, and there is texture interference generated between different textures. The above content is subjected to multimodal fitting to obtain the texture features of each carrot, and then the texture features of multiple carrots are weighted and averaged to obtain the product weighted texture features of each carrot. According to the packaging texture data, the barcode texture data and the packaging bag texture data on each packaging bag are obtained. The packaging bag texture data on the packaging is weighted and averaged to obtain the packaging weighted texture features, and then the barcode texture data is marked to obtain the barcode mark. The third identification is performed based on the above two texture features, and the barcode mark is added after the identification to obtain the third identification result.

[0035] An embodiment of the present invention provides an automatic identification and distribution method for agricultural product distribution. Using an automatic identification and distribution system for agricultural product distribution as described in any one of the above, the method includes the following steps: Obtain the basic appearance information and packaging information of the agricultural product, and obtain the basic identification features of the agricultural product according to the basic appearance information and packaging information; Obtain the visual information, spectral information, and texture information of the agricultural product, and perform information screening on the visual information, spectral information, and texture information based on the basic appearance information to obtain visual data, spectral data, and texture data; According to the visual data, analyze the visual data to obtain the difference amount between different individuals in the agricultural product, and perform the first identification on the agricultural product according to the difference amount to obtain the first identification result; According to the spectral data, analyze the spectral data to obtain the internal quality and maturity of the agricultural product, and perform the second identification on the agricultural product according to the internal quality and maturity to obtain the second identification result; According to the texture data, perform refined analysis on the texture data to respectively obtain the product texture data and the packaging texture data of the agricultural product, and perform the third identification on the agricultural product according to the product texture data and the packaging texture data to obtain the third identification result; Combine the first identification result, the second identification result, and the third identification result to obtain the final identification result of the agricultural product, and perform distribution on the agricultural product according to the final identification result.

[0036] The above are all preferred embodiments of the present application. The protection scope of the present application is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. An automatic identification system for agricultural product distribution, characterized in that: include: Basic information module: used to obtain basic appearance information and packaging information of agricultural products, and obtain basic identification features of agricultural products according to the basic appearance information and packaging information; Information screening module: used to obtain visual information, spectral information and texture information of agricultural products, and screen the visual information, spectral information and texture information based on the basic appearance information to obtain visual data, spectral data and texture data; Visual recognition module: used for analyzing the visual data according to the visual data to obtain the difference between different individuals in the agricultural products, and performing a first recognition on the agricultural products according to the difference to obtain a first recognition result; Spectral recognition module: used for analyzing the spectral data according to the spectral data to obtain the internal quality and maturity of the agricultural product, and performing a second recognition on the agricultural product according to the internal quality and the maturity to obtain a second recognition result; Texture recognition module: used for performing a detailed analysis on the texture data according to the texture data, respectively obtaining product texture data and packaging texture data of the agricultural product, performing a third recognition on the agricultural product according to the product texture data and the packaging texture data, and obtaining a third recognition result; Prime distribution module: used to combine the first recognition result, the second recognition result and the third recognition result to obtain the final recognition result of the agricultural products, and distribute the agricultural products according to the final recognition result.

2. The automatic identification system for agricultural product distribution according to claim 1 is characterized in that: The steps of obtaining basic appearance information and packaging information of the agricultural product and obtaining basic identification features of the agricultural product according to the basic appearance information and the packaging information are specifically as follows: Take photos and samples of the entire process of agricultural products from breeding to packaging and then to decay, and obtain full-line photo samples; Extracting appearance color information, basic spectrum information, and basic texture information of the agricultural products in the photo samples; Combining the appearance color information, the basic spectrum information and the basic texture information to obtain basic appearance information of the agricultural product; Extracting basic packaging information of agricultural products in the photo samples, and subdividing the basic packaging information to obtain packaging material information and barcode information; The packaging information of the agricultural product is obtained according to the packaging material information and the barcode information, and feature extraction is performed in combination with the basic appearance information and the packaging information to obtain basic identification features.

3. The automatic identification system for agricultural product distribution according to claim 2 is characterized in that: The step of filtering the visual information, the spectral information and the texture information based on the basic appearance information to obtain the visual data, the spectral data and the texture data is specifically: Obtaining ontology data of the agricultural product according to the basic appearance information, and performing background separation on the visual information, the spectral information and the texture information according to the ontology data to obtain ontology visual information, ontology spectral information and ontology texture information; Adding different lighting conditions to the basic appearance information, and generating appearance information sets under different lighting conditions; The main body visual information, the main body spectral information and the main body texture information are screened according to the appearance information set to obtain visual data, spectral data and texture data.

4. The automatic identification system for agricultural product distribution according to claim 3 is characterized in that: The steps of analyzing the visual data to obtain the difference between different individuals in the agricultural products, and performing a first identification on the agricultural products according to the difference to obtain a first identification result are specifically: Based on the visual data, shape data and color data of each individual agricultural product are obtained, and appearance data of each individual agricultural product is obtained according to the shape data and the color data; Comparing the appearance data corresponding to each individual agricultural product one by one to obtain a plurality of difference data values, and performing weighted average calculation on the difference data values ​​to obtain the difference between different individuals; Determining whether the difference is greater than a preset difference threshold; If it is determined that the difference amount is greater than the difference threshold, the agricultural product is first identified according to the shape data and the color data to obtain a first identification result; If it is determined that the difference amount is less than or equal to the difference threshold, the shape data and the color data are input into a convolutional neural network for learning to obtain a target convolutional neural network after learning; The target convolutional neural network is called to perform a first recognition on the agricultural product to obtain a first recognition result.

5. The automatic identification system for agricultural product distribution according to claim 4 is characterized in that: The step of inputting the shape data and the color data into a convolutional neural network for learning to obtain a target convolutional neural network after learning is specifically as follows: Normalizing the shape data and the color data of each individual agricultural product to obtain to-be-processed shape data and to-be-processed color data of the same data size; Rotate, flip, scale, and crop the shape data to be processed to expand the data and generate a data set with diverse shapes; Performing color transformation on the color data to be processed to perform data expansion, and generating a color diverse data set; The shape-diverse data set and the color-diverse data set are input into a convolutional neural network for learning to obtain a learned target convolutional neural network.

6. The automatic identification system for agricultural product distribution according to claim 4 is characterized in that: The steps of analyzing the spectral data to obtain the internal quality and maturity of the agricultural product, and performing a second identification on the agricultural product according to the internal quality and the maturity to obtain a second identification result are specifically: Based on the spectral data, identifying the baseline drift in the spectral data by polynomial fitting, and deleting the baseline drift to obtain target spectral data; Establishing a spectral absorption characteristic comparison table, dividing the target spectral data into bands based on the spectral absorption characteristic comparison table to obtain a quality band spectrum and a maturity band spectrum; Using independent component analysis to perform dimensionality reduction processing on the quality band spectrum and the maturity band spectrum to extract quality spectrum features and maturity spectrum features; Obtaining the internal quality and maturity of the agricultural product based on the quality spectral characteristics and the maturity spectral characteristics; The agricultural products are identified for a second time according to the internal quality and the maturity to obtain a second identification result.

7. The automatic identification system for agricultural product distribution according to claim 6 is characterized in that: The step of performing a second identification on the agricultural product according to the internal quality and the maturity to obtain a second identification result is specifically: Establishing quality standard intervals for agricultural products and dividing the standard intervals to form multiple standard intervals; Obtaining an internal chemical composition of the agricultural product based on the internal quality, and obtaining a first quality parameter of the agricultural product based on the internal chemical composition; Obtaining physical properties and cell composition of the agricultural product according to the maturity, and obtaining a second quality parameter of the agricultural product according to the physical properties and the cell composition; The quality parameter of the agricultural product is obtained according to the first quality parameter and the second quality parameter, and the quality parameter is calibrated and identified according to the multiple standard intervals to obtain a second identification result.

8. The automatic identification system for agricultural product distribution according to claim 6 is characterized in that: The steps of performing a detailed analysis on the texture data to obtain product texture data and packaging texture data of the agricultural product respectively, and performing a third recognition on the agricultural product according to the product texture data and the packaging texture data to obtain a third recognition result are specifically as follows: Performing detailed analysis on the texture data to obtain product texture data and packaging texture data of the agricultural product respectively; Based on the product texture data, the texture direction, texture depth and texture interference of each individual agricultural product are obtained; Performing multimodal fitting on the texture direction, the texture depth and the texture interference to obtain the texture features of each individual agricultural product, and performing weighted averaging on the texture features of each individual agricultural product to obtain the product weighted texture features of the agricultural product; Based on the package texture data, obtain the barcode texture data and the packaging bag texture data of each package, perform weighted averaging on the packaging bag texture data of multiple packages to obtain the package weighted texture feature, and mark each of the barcode texture data on each package to obtain a barcode mark; The agricultural products are identified for a third time according to the product weighted texture features and the package weighted texture features, and a barcode mark is added after the identification to obtain a third identification result.

9. A method for automatic identification and distribution of agricultural products, the method using an automatic identification system for distribution of agricultural products as claimed in any one of claims 1 to 8, characterized in that: The method comprises the following steps: Acquire basic appearance information and packaging information of the agricultural product, and obtain basic identification features of the agricultural product according to the basic appearance information and the packaging information; Acquire visual information, spectral information and texture information of the agricultural product, and screen the visual information, spectral information and texture information based on the basic appearance information to obtain visual data, spectral data and texture data; Analyzing the visual data to obtain differences between different individuals in the agricultural products, and performing a first identification on the agricultural products according to the differences to obtain a first identification result; Analyzing the spectral data to obtain the internal quality and maturity of the agricultural product, and performing a second identification on the agricultural product according to the internal quality and the maturity to obtain a second identification result; According to the texture data, the texture data is refined and analyzed to obtain product texture data and packaging texture data of the agricultural product respectively, and the agricultural product is thirdly identified according to the product texture data and the packaging texture data to obtain a third identification result; The first recognition result, the second recognition result and the third recognition result are combined to obtain a final recognition result of the agricultural product, and the agricultural product is distributed according to the final recognition result.

Images