Automatic durian quality evaluating and sorting system

Abstract

The invention provides an automatic durian quality evaluating and sorting system which comprises a conveying unit, a protective box body unit, a processing unit, an appearance image collecting unit, an x-ray detecting unit, a near-infrared detecting unit and a weight measuring unit, and the conveying unit is used for conveying durian from one side of the protective box body unit to the other side of the protective box body unit; the appearance image acquisition unit is used for acquiring a multi-angle appearance image of the durian when the durian is transmitted to an acquisition area of the appearance image acquisition unit; the x-ray detection unit is used for collecting x-ray imaging data of the durian when the durian is conveyed to an x-ray detection area; the near-infrared detection unit is used for collecting a near-infrared characteristic spectrum of the durian when the durian is transmitted to a near-infrared detection area; the weight measuring unit is used for collecting the weight of the durian; and the processing unit is used for comprehensively evaluating the durians and sorting the durians according to a comprehensive evaluation result. According to the method, the durian evaluation and sorting accuracy, efficiency and standard consistency can be improved.

Description

Technical Field

[0001] The invention relates to the technical field of fruit quality evaluation, and in particular to an automatic durian quality evaluation and sorting system. Background Art

[0002] Durian is a high-end fruit, mainly produced in Thailand, Malaysia, Vietnam and other countries. The size, pulp percentage, maturity and other qualities of durian directly determine the market price, affecting the interests of traders and consumers. At present, the judgment and sorting of durian quality in the market mainly rely on the experience of fruit pickers, who mainly judge based on the appearance of the durian, smell, shake, pinch the thorns, touch and knock the sound and feel. Traditional manual fruit selection methods have problems such as high error rate, difficulty in making high-sensitivity judgments for a long time, uneven levels, and human corruption, among which:

[0003] (1) High error rate: It is difficult to accurately judge the number of durian cells, dead bags, empty bags, moldy cores, pulp percentage, maturity, etc. in manual selection without loss. (2) Difficult to make high-sensitivity judgments for a long time: The sensitivity of the fruit selector to sound and touch will decrease significantly over time. If it lasts for more than 2 hours, the sensitivity will be less than 10% of the initial level. (3) Uneven levels: The ability and experience levels of the fruit selectors are uneven, and there is a large uncertainty in the knocking position, and there is a large proportion of inaccurate knocking positions. (4) Human corruption: There is corruption among fruit farmers, fruit selectors, and factories, and there is a subjective situation of passing off inferior goods as good ones.

[0004] Based on the above situation, a solution is needed that can automatically evaluate and sort durian quality to improve the accuracy, efficiency and standard consistency of durian evaluation and sorting. Summary of the invention

[0005] In order to solve at least one of the above technical problems, an embodiment of the present invention provides a system for automatically evaluating and sorting the quality of durian.

[0006] According to the first aspect, the automatic durian quality evaluation and sorting system provided by the embodiment of the present invention includes a transmission unit, a protective box unit, a processing unit, and an appearance image acquisition unit, an X-ray detection unit, a near infrared detection unit and a weight measurement unit arranged in the protective box unit, wherein:

[0007] The conveying unit is used to convey the durian from one side of the protective box unit to the other side;

[0008] The appearance image acquisition unit is used to acquire multi-angle appearance images of the durian when the durian is transferred to the acquisition area of ​​the appearance image acquisition unit;

[0009] The x-ray detection unit is used to collect x-ray imaging data of the durian when the durian is conveyed to the x-ray detection area;

[0010] The near infrared detection unit is used to collect the near infrared characteristic spectrum of the durian when the durian is transferred to the near infrared detection area;

[0011] The weight measuring unit is used to collect the weight of the durian;

[0012] The processing unit is used to perform multi-faceted evaluation on the durian according to the multi-angle appearance image, the X-ray imaging data, the near-infrared characteristic spectrum and the weight, perform comprehensive evaluation on the durian according to the multi-faceted evaluation results, and sort the durian according to the comprehensive evaluation results.

[0013] In one embodiment, the conveyor belt unit includes a bracket and a conveyor belt arranged on the bracket; the protective box unit includes a box, shielding curtains arranged on both sides of the box, and a lead protective body arranged in the box; the conveyor belt is used to drive the durian into the box from the shielding curtain on one side of the box and drive the durian out from the shielding curtain on the other side of the box.

[0014] In one embodiment, the appearance image acquisition unit includes a multi-angle light source, a photoelectric sensor and a multi-directional color industrial camera, wherein:

[0015] The multi-angle light source is used to provide the light source environment required for appearance image acquisition;

[0016] The photoelectric sensor is used to be triggered and counted when the durian passes through the photoelectric sensor;

[0017] The multi-directional color industrial camera is used to: when the photoelectric sensor is triggered, calculate the first start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the center position of the multi-directional color industrial camera, and take a photo when the first start time arrives to obtain the multi-angle appearance image.

[0018] In one embodiment, the x-ray detection unit comprises an x-ray source, a detector and a first control subunit; wherein:

[0019] The first control subunit is used to: calculate a second start time according to a conveying speed of the conveying unit and a distance between the photoelectric sensor and the x-ray detection area, and start the x-ray source and the detector when the second start time arrives;

[0020] The x-ray source is used to emit x-rays when activated;

[0021] The detector is used to receive, on the opposite side of the X-ray source, X-ray imaging data obtained after the X-rays pass through the durian.

[0022] In one embodiment, the near infrared detection unit includes a near infrared light source, a spectrum analyzer and a second control subunit; wherein:

[0023] The second control subunit is used to calculate a third start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the near-infrared detection area, and start the near-infrared light source and the spectrum analyzer when the third start time arrives;

[0024] The near-infrared light source is used to emit near-infrared light when started;

[0025] The spectrum analyzer is used for receiving, on the opposite side of the near-infrared light source, the near-infrared characteristic spectrum obtained after the near-infrared light passes through the durian.

[0026] In one embodiment, the processing unit is also used to: take the counting number, detection time, the multi-angle appearance image, the x-ray imaging data, the near-infrared characteristic spectrum and the weight corresponding to each durian as the detection data of the durian and store them.

[0027] In one embodiment, the processing unit is specifically used to: perform corresponding evaluation processing using multiple models based on the multi-angle appearance image, the x-ray imaging data and the near-infrared characteristic spectrum to obtain multiple evaluation results; wherein the multiple evaluation results include the appearance evaluation result, internal evaluation result and maturity evaluation result of the durian; determine the comprehensive evaluation result of the durian based on the appearance evaluation result, the internal evaluation result, the maturity evaluation result and the weight.

[0028] In one embodiment, the processing unit uses multiple target detection models to perform corresponding target detection according to the multi-angle appearance image, the x-ray imaging data and the near-infrared characteristic spectrum, and performs corresponding evaluation processing according to the target detection results, including: preprocessing the multi-angle appearance image and the x-ray imaging data, inputting the preprocessed multi-angle appearance image into a pre-trained first target detection model to obtain the appearance evaluation result of the durian; wherein the first target detection model is a target detection model for detecting and evaluating the shape contour, wormholes, mildew and / or damage of the durian; inputting the preprocessed x-ray imaging data into a second target detection model to obtain the internal evaluation result of the durian; wherein the second target detection model is a target detection model for detecting and evaluating the number of durian cells, shell contour, empty bag, raw bag, dead bag, insect damage and / or moldy core of the durian; extracting characteristic wavelengths related to sugar content from the near-infrared characteristic spectrum, and inputting the extracted characteristic wavelengths related to sugar content into a maturity evaluation model to obtain the maturity evaluation result.

[0029] In one embodiment, the dead bag evaluation result output by the second target detection model used to detect the dead bag is a suspected dead bag evaluation result; the processing unit is also used to: predict the actual dead bag situation of the durian according to the suspected dead bag evaluation result and the maturity evaluation result.

[0030] In one embodiment, the method further comprises:

[0031] The processing unit is also used to: generate a label for the durian according to a preset template based on the comprehensive evaluation result and the sorting processing result, send the label to a color printer so that the color printer prints the label of the durian, and after printing the label, personnel can stick the label on the durian.

[0032] The automatic evaluation and sorting system for durian quality provided by the embodiment of the present invention transports durian from one side of the protective box unit to the other side through the transmission unit, so that the durian passes through the appearance image acquisition unit, the x-ray detection unit, the near-infrared detection unit and the weight measurement unit inside the protective box unit, so that the appearance image acquisition unit acquires the multi-angle appearance image of the durian, the x-ray detection unit acquires the x-ray imaging data of the durian, the near-infrared detection unit acquires the near-infrared characteristic spectrum of the durian, and the weight measurement unit acquires the weight of the durian. Then the processing unit uses the multi-angle appearance image, x-ray imaging data, near-infrared characteristic spectrum and the weight of a durian to evaluate the durian in many aspects, and then comprehensively evaluates the durian according to the evaluation results in many aspects, and finally sorts the durian according to the comprehensive evaluation results, so as to realize the evaluation and sorting of the durian. It can be seen that the embodiment of the present invention uses industrial cameras, X-rays, and near-infrared spectra to collect data images of the internal and external conditions of the durian, and evaluates the quality of the durian based on the model, which has a wider range of detection and evaluation capabilities than the existing automatic durian quality evaluation technology, that is, the embodiment of the present invention has a more comprehensive detection and evaluation capability. Moreover, the detection process is carried out in a non-destructive manner, and no damage is caused to the durian. In addition, since the present invention does not detect and sort by manual methods, it does not have the various disadvantages of manual methods. In short, the embodiments of the present invention can improve the accuracy, efficiency and standard consistency of durian evaluation and sorting. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The figure is a schematic diagram of the structure of a system for automatically evaluating and sorting durian quality in one embodiment of the present invention. DETAILED DESCRIPTION

[0034] In a first aspect, an embodiment of the present invention provides a system for automatically evaluating and sorting the quality of durian, the system comprising a transmission unit, a protective box unit, a processing unit, and an appearance image acquisition unit, an X-ray detection unit, a near infrared detection unit, and a weight measurement unit arranged in the protective box unit, wherein:

[0035] The conveying unit is used to convey the durian from one side of the protective box unit to the other side;

[0036] The appearance image acquisition unit is used to acquire multi-angle appearance images of the durian when the durian is transferred to the acquisition area of ​​the appearance image acquisition unit;

[0037] The x-ray detection unit is used to collect x-ray imaging data of the durian when the durian is conveyed to the x-ray detection area;

[0038] The near infrared detection unit is used to collect the near infrared characteristic spectrum of the durian when the durian is transferred to the near infrared detection area;

[0039] The weight measuring unit is used to collect the weight of the durian;

[0040] The processing unit is used to perform multi-faceted evaluation on the durian according to the multi-angle appearance image, the X-ray imaging data, the near-infrared characteristic spectrum and the weight, perform comprehensive evaluation on the durian according to the multi-faceted evaluation results, and sort the durian according to the comprehensive evaluation results.

[0041] It can be seen that in the embodiment of the present invention, the durian is transported from one side of the protective box unit to the other side through the transmission unit, so that the durian passes through the appearance image acquisition unit, the x-ray detection unit, the near-infrared detection unit and the weight measurement unit inside the protective box unit, so that the appearance image acquisition unit acquires multi-angle appearance images of the durian, the x-ray detection unit acquires x-ray imaging data of the durian, the near-infrared detection unit acquires the near-infrared characteristic spectrum of the durian, and the weight measurement unit acquires the weight of the durian. Then, the processing unit uses the multi-angle appearance images, x-ray imaging data, near-infrared characteristic spectrum and the weight of a durian to evaluate the durian in many aspects, and then comprehensively evaluates the durian according to the evaluation results in many aspects, and finally sorts the durian according to the comprehensive evaluation results to achieve evaluation and sorting of the durian.

[0042] In one embodiment, see Figure 1 The conveyor belt unit includes a bracket 2 and a conveyor belt 1 arranged on the bracket 2; the protective box unit includes a box 11, shielding curtains 13 arranged on both sides of the box, and a lead protective body 12 arranged in the box; the conveyor belt 1 is used to drive the durian into the box 11 from the shielding curtain 13 on one side of the box 11 and drive the durian out of the shielding curtain 13 on the other side of the box 11.

[0043] Among them, the lead protection body inside the box can prevent the x-rays and near-infrared rays inside the box from causing harm to people outside the box, thereby improving the safety of durian detection and sorting.

[0044] In one embodiment, see Figure 1 , the appearance image acquisition unit may include a multi-angle light source 4, a photoelectric sensor 3 and a multi-directional color industrial camera 5, wherein:

[0045] The multi-angle light source 4 is used to provide the light source environment required for appearance image acquisition;

[0046] The photoelectric sensor 3 is used to be triggered and count when the durian passes through the photoelectric sensor;

[0047] The multi-directional color industrial camera 5 is used to: when the photoelectric sensor is triggered, calculate the first start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the center position of the multi-directional color industrial camera, and take a photo when the first start time arrives to obtain the multi-angle appearance image.

[0048] Among them, the multi-directional color industrial camera can adopt a multi-directional high-resolution color CCD industrial camera.

[0049] Among them, multi-angle light sources can ensure the consistency of the light source environment when capturing appearance images without shadows.

[0050] It can be seen that when a durian is placed on the conveyor belt and passes through the photoelectric sensor, the photoelectric sensor is triggered and counted. The distance between the photoelectric sensor and the center position of the multi-directional color industrial camera is determined, and the distance is divided by the conveyor belt speed to obtain the first start time. When the first start time after the photoelectric sensor is triggered is reached, the multi-directional color industrial camera is started to take pictures, thereby obtaining a multi-angle appearance image of the durian. The processing unit uses the various collected data to evaluate the durian in various aspects. After the evaluation of various aspects is completed, the evaluation results of the durian in various aspects can be calculated by combining expert experience and entropy weight method to obtain a comprehensive evaluation result, so as to perform sorting processing. Among them, the calculation method, for example, the weighted summation calculation method.

[0051] In one embodiment, see Figure 1 , the x-ray detection unit may include an x-ray source 6, a detector and a first control subunit 7; wherein:

[0052] The first control subunit is used to: calculate a second start time according to a conveying speed of the conveying unit and a distance between the photoelectric sensor and the x-ray detection area, and start the x-ray source and the detector when the second start time arrives;

[0053] The x-ray source is used to emit x-rays when activated;

[0054] The detector is used to receive, on the opposite side of the X-ray source, X-ray imaging data obtained after the X-rays pass through the durian.

[0055] Among them, multiple x-ray sources and detectors can be set respectively, for example, an x-ray source is set at the bottom and one side of the box, and a detector is set at the top and the other side. The first control subunit calculates the distance between the photoelectric sensor and the x-ray detection area, divides the distance by the transmission speed, and obtains the second start time. When the second start time is reached after the photoelectric sensor is triggered, each x-ray source and detector is started, so that the x-ray sources at the bottom and one side emit x-rays, and the detectors at the top and the other side receive the x-ray signals after passing through the durian, thereby obtaining x-ray imaging data.

[0056] In one embodiment, see Figure 1 , the near infrared detection unit may include a near infrared light source 8, a spectrometer 9 and a second control subunit; wherein:

[0057] The second control subunit is used to calculate a third start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the near-infrared detection area, and start the near-infrared light source and the spectrum analyzer when the third start time arrives;

[0058] The near-infrared light source 8 is used to emit near-infrared light when started;

[0059] The spectrum analyzer 9 is used for receiving the near-infrared characteristic spectrum obtained after the near-infrared light passes through the durian at the opposite side of the near-infrared light source.

[0060] It can be seen that the second control subunit can determine the distance between the photoelectric sensor and the near-infrared detection area, divide the distance by the conveyor belt speed, and obtain the third start time. When the third start time is reached after the photoelectric sensor is triggered, the near-infrared light source and the spectrometer are started, so that the near-infrared light source shines on the durian from one side, and the spectrometer receives the transmitted near-infrared spectrum from the other side, thereby obtaining the near-infrared characteristic spectrum of the durian.

[0061] See also Figure 1 When the durian is brought to the weighing area by the conveyor belt, the weight measuring unit 10 will weigh the durian to obtain the weight.

[0062] In one embodiment, the processing unit is also used to: take the counting number, detection time, the multi-angle appearance image, the x-ray imaging data, the near-infrared characteristic spectrum and the weight corresponding to each durian as the detection data of the durian and store them.

[0063] That is to say, after a durian is transported from one side of the box to the other side, the counting number, detection time, multi-angle appearance image, X-ray imaging data, near-infrared characteristic spectrum and weight of the durian can be obtained. In order to facilitate subsequent use, the above data of the durian are saved in the database.

[0064] In one embodiment, the processing unit can be specifically used to: perform corresponding evaluation processing using multiple models based on the multi-angle appearance image, the x-ray imaging data and the near-infrared characteristic spectrum to obtain multiple evaluation results; wherein the multiple evaluation results include the appearance evaluation result, internal evaluation result and maturity evaluation result of the durian; determine the comprehensive evaluation result of the durian based on the appearance evaluation result, the internal evaluation result, the maturity evaluation result and the weight.

[0065] It can be seen that different models can be used to obtain evaluation results of different indicators. Through multiple models, we can obtain multiple evaluation results of a durian, including appearance evaluation results, internal evaluation results and maturity evaluation results, and then conduct a comprehensive evaluation based on the evaluation results of multiple aspects to obtain a comprehensive evaluation result.

[0066] Specifically, the processing unit uses multiple models to perform corresponding evaluation processing according to the multi-angle appearance image, the X-ray imaging data and the near-infrared characteristic spectrum to obtain multiple evaluation results, which may include:

[0067] The multi-angle appearance image and the x-ray imaging data are preprocessed, and the preprocessed multi-angle appearance image is input into a pre-trained first target detection model to obtain the appearance evaluation result of the durian; wherein the first target detection model is a target detection model for detecting and evaluating the shape contour, wormholes, mildew and / or damage of the durian; the preprocessed x-ray imaging data is input into a second target detection model to obtain the internal evaluation result of the durian; wherein the second target detection model is a target detection model for detecting and evaluating the number of durian cells, shell contour, empty pods, raw pods, dead pods, insect pests and / or moldy cores of the durian; characteristic wavelengths related to sugar content are extracted from the near-infrared characteristic spectrum, and the extracted characteristic wavelengths related to sugar content are input into a maturity evaluation model to obtain the maturity evaluation result.

[0068] That is, the first target detection model may include target detection models for detecting and evaluating the shape contour, wormhole, mildew and / or damage of the durian. It can be seen that based on multiple first target detection models, the evaluation results of the shape contour, wormhole, mildew and / or damage of the durian can be obtained, and these evaluation results reflect the appearance of the durian. The second target detection model may include target detection models for detecting and evaluating the number of durian rooms, shell contour, empty package, raw package, dead package, insect pest and / or moldy heart of the durian. It can be seen that based on multiple second target detection models, the evaluation results of the number of durian rooms, shell contour, empty package, raw package, dead package, insect pest and / or moldy heart of the durian can be obtained, and these evaluation results reflect the internal situation of the durian. In addition to evaluating the appearance and internal situation, the maturity of the durian is evaluated according to the near-infrared characteristic spectrum and the maturity evaluation model.

[0069] Furthermore, the dead bag evaluation result output by the second target detection model for detecting the dead bag is a suspected dead bag evaluation result; and the processing unit is further used to predict the actual dead bag situation of the durian based on the suspected dead bag evaluation result and the maturity evaluation result.

[0070] It can be seen that the dead bag evaluation result output by the second target detection model for detecting the dead bag is a suspected dead bag evaluation result, and it cannot be determined that there must be dead bags in durian. Therefore, it can be further judged that the actual dead bag situation of durian can be further predicted based on the suspected dead bag evaluation result and the maturity evaluation result, and the decision tree method can be applied. It is understandable that the more mature the durian is, the fewer the dead bags are. Therefore, the maturity can be used to assist in predicting whether the suspected dead bag is a real dead bag, thereby improving the accuracy of determining the dead bag.

[0071] In one embodiment, in addition to using the second target detection model to evaluate the moldy core, the near-infrared characteristic spectrum can also be used for judgment. Specifically, the characteristic wavelength related to the moldy core is extracted from the near-infrared characteristic spectrum, and the extracted characteristic wavelength related to the moldy core is input into the moldy core evaluation model to obtain the moldy core evaluation result.

[0072] The preprocessing mentioned above may include format conversion, scaling, enhancement, denoising, image normalization, encoding and other preprocessing.

[0073] Among them, the target detection model or other models mentioned above are pre-trained based on training data. The target detection model can be constructed based on convolutional neural network deep learning algorithm, regression / classification target detection method, correlation coefficient method, uninformative variable elimination method, interval partial least squares method, reverse interval partial least squares method, forward interval partial least squares method, genetic algorithm, multivariate linear regression and other methods.

[0074] For example, the target detection model is constructed using the convolutional neural network-based deep learning algorithm architecture Fast R-CNN and the regression / classification-based target detection method SSD, and the model is trained using 7,000 positive samples and 50,000 negative samples.

[0075] In one embodiment, the processing unit can also be used to: generate a label for the durian according to a preset template based on the comprehensive evaluation results and the sorting processing results, send the label to a color printer so that the color printer prints the label of the durian, and after printing the label, personnel can stick the label on the durian.

[0076] That is to say, after the comprehensive evaluation and sorting of a durian is completed, a corresponding label is generated according to the comprehensive evaluation result and the sorting result, and the label is printed, so that the staff can stick the label on the durian to provide reference for subsequent operations. In order to ensure that the label is not easily damaged, the label can also be attached to the durian after being plasticized.

[0077] Of course, a monitoring unit can also be set up in the system to display the various detection processes, multi-faceted evaluation results, comprehensive evaluation results, and sorting results of durian. The system can also be operated and controlled by touch screen for personnel to understand intuitively.

[0078] Based on the above description, it can be known that the present invention can automatically evaluate the following aspects without damaging the durian, conduct comprehensive evaluation and sorting, and print labels to form one label for one fruit. A QR code can also be set. By scanning the QR code, specific evaluation content can be displayed, even including traceability monitoring content of the entire durian circulation process.

[0079] In summary, compared with the existing automatic evaluation and sorting technology, the embodiment of the present invention has a more comprehensive detection and evaluation capability for the internal and external quality of durian. Specifically, industrial cameras, X-rays, and near-infrared spectra are used to collect data images of the internal and external conditions of durian, and the quality of durian is evaluated based on the model, which has a wider detection and evaluation capability range than the existing automatic evaluation technology for durian quality. Moreover, the detection process is carried out under non-destructive conditions and will not cause harm to durian. Since the present invention is not detected and sorted by manual methods, it does not have the various shortcomings of manual methods. In short, the embodiment of the present invention can improve the accuracy, efficiency and standard consistency of durian evaluation and sorting.

[0080] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0081] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, widgets, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0082] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A durian quality automatic evaluation and sorting system, characterized in that: The invention comprises a transmission unit, a protective box unit, a processing unit, and an appearance image acquisition unit, an X-ray detection unit, a near infrared detection unit and a weight measurement unit arranged in the protective box unit, wherein: The conveying unit is used to convey the durian from one side of the protective box unit to the other side; The appearance image acquisition unit is used to acquire multi-angle appearance images of the durian when the durian is transferred to the acquisition area of ​​the appearance image acquisition unit; The x-ray detection unit is used to collect x-ray imaging data of the durian when the durian is conveyed to the x-ray detection area; The near infrared detection unit is used to collect the near infrared characteristic spectrum of the durian when the durian is transferred to the near infrared detection area; The weight measuring unit is used to collect the weight of the durian; The processing unit is used to perform multi-faceted evaluation on the durian according to the multi-angle appearance image, the X-ray imaging data, the near-infrared characteristic spectrum and the weight, perform comprehensive evaluation on the durian according to the multi-faceted evaluation results, and sort the durian according to the comprehensive evaluation results.

2. The system according to claim 1, characterized in that The conveyor belt unit includes a bracket and a conveyor belt arranged on the bracket; the protective box unit includes a box, shielding curtains arranged on both sides of the box, and a lead protective body arranged in the box; the conveyor belt is used to drive the durian into the box from the shielding curtain on one side of the box and drive the durian out from the shielding curtain on the other side of the box.

3. The system according to claim 1, characterized in that The appearance image acquisition unit includes a multi-angle light source, a photoelectric sensor and a multi-directional color industrial camera, wherein: The multi-angle light source is used to provide the light source environment required for appearance image acquisition; The photoelectric sensor is used to be triggered and counted when the durian passes through the photoelectric sensor; The multi-directional color industrial camera is used to: when the photoelectric sensor is triggered, calculate the first start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the center position of the multi-directional color industrial camera, and take a photo when the first start time arrives to obtain the multi-angle appearance image.

4. The system according to claim 3, characterized in that The x-ray detection unit comprises an x-ray source, a detector and a first control subunit; wherein: The first control subunit is used to: calculate a second start time according to a conveying speed of the conveying unit and a distance between the photoelectric sensor and the x-ray detection area, and start the x-ray source and the detector when the second start time arrives; The x-ray source is used to emit x-rays when activated; The detector is used to receive, on the opposite side of the X-ray source, X-ray imaging data obtained after the X-rays pass through the durian.

5. The system according to claim 3, characterized in that The near infrared detection unit includes a near infrared light source, a spectrum analyzer and a second control subunit; wherein: The second control subunit is used to calculate a third start time according to the transmission speed of the transmission unit and the distance between the photoelectric sensor and the near-infrared detection area, and start the near-infrared light source and the spectrum analyzer when the third start time arrives; The near-infrared light source is used to emit near-infrared light when started; The spectrum analyzer is used for receiving, on the opposite side of the near-infrared light source, the near-infrared characteristic spectrum obtained after the near-infrared light passes through the durian.

6. The system according to claim 1, characterized in that The processing unit is also used to: store the counting number, detection time, multi-angle appearance image, X-ray imaging data, near-infrared characteristic spectrum and weight corresponding to each durian as the detection data of the durian.

7. The system according to claim 1, characterized in that The processing unit is specifically used for: performing corresponding evaluation processing using multiple models according to the multi-angle appearance image, the X-ray imaging data and the near-infrared characteristic spectrum to obtain multiple evaluation results; wherein the multiple evaluation results include the appearance evaluation result, internal evaluation result and maturity evaluation result of the durian; and determining the comprehensive evaluation result of the durian according to the appearance evaluation result, the internal evaluation result, the maturity evaluation result and the weight.

8. The system according to claim 7, characterized in that The processing unit uses a plurality of target detection models to perform corresponding target detection according to the multi-angle appearance image, the x-ray imaging data and the near-infrared characteristic spectrum, and performs corresponding evaluation processing according to the target detection results, including: preprocessing the multi-angle appearance image and the x-ray imaging data, inputting the preprocessed multi-angle appearance image into a first target detection model trained in advance, and obtaining the appearance evaluation result of the durian; wherein the first target detection model is a target detection model for detecting and evaluating the shape contour, wormholes, mildew and / or damage of the durian; inputting the preprocessed x-ray imaging data into a second target detection model, and obtaining the internal evaluation result of the durian; wherein the second target detection model is a target detection model for detecting and evaluating the number of durian cells, shell contour, empty pods, raw pods, dead pods, insect pests and / or moldy cores of the durian; extracting characteristic wavelengths related to sugar content from the near-infrared characteristic spectrum, and inputting the extracted characteristic wavelengths related to sugar content into a maturity evaluation model to obtain the maturity evaluation result.

9. The system according to claim 7, characterized in that The dead bag evaluation result output by the second target detection model for detecting the dead bag is a suspected dead bag evaluation result; the processing unit is also used to predict the actual dead bag situation of the durian according to the suspected dead bag evaluation result and the maturity evaluation result.

10. The system according to claim 1, characterized in that The processing unit is also used to: generate a label for the durian according to a preset template based on the comprehensive evaluation result and the sorting processing result, send the label to a color printer so that the color printer prints the label of the durian, and after printing the label, personnel can stick the label on the durian.

Images