Method for detecting appearance quality of dried Chinese prickly ash based on machine vision

Through the machine vision-based dried pepper detection method, using loose arrangement and image processing technology, a high-accuracy classification neural network model is constructed, which solves the problem of low accuracy in the detection of dried pepper in the existing technology, and realizes the use of pepper seeds Accurate distinction and efficient detection of packaging and shedding patterns.

CN119090846APending Publication Date: 2024-12-06NORTHWEST A & F UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411191730.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing dried pepper detection method has difficulty in constructing samples, resulting in inaccurate classification results. It is difficult to distinguish between uncracked pepper seeds and completely cracked pepper seeds when distinguishing based on color texture, which reduces the detection accuracy.

Method used

Using machine vision-based detection method, by loosely arranging and placing dried pepper samples on the container, multi-sample images were collected, and image binarization threshold segmentation method was used to crop into a single-sample image to build a training classification neural network model. Including VGG-16, ResNet50, MobileNet, SwinTransformer or Vision Transformer network models, perform high-accuracy detection.

Benefits of technology

The individual equidistance control of dried pepper seeds is achieved, which reduces the difficulty of sample collection and improves the detection accuracy, especially the distinction between the morphology of the pepper seeds being completely wrapped and completely shed, and improves the accuracy of classification results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090846A_ABST
    Figure CN119090846A_ABST
Patent Text Reader

Abstract

The invention relates to a method for detecting the appearance quality of dried Chinese prickly ash based on machine vision, which comprises the following steps of: firstly, tiling a to-be-detected dried Chinese prickly ash grain group on any plane, then collecting an image of the to-be-detected dried Chinese prickly ash grain group, and then inputting the image of the to-be-detected dried Chinese prickly ash grain group into a trained classification neural network model; and obtaining an appearance quality detection result of the to-be-detected dried zanthoxylum bungeanum granule group. When the sample set is constructed, a dried pepper sample particle group containing a plurality of components is firstly obtained, and each individual in the dried pepper sample particle group is classified according to different components; then, for each category component, loosely arranging and placing for at least one time based on a container for detecting the dried Chinese prickly ash, and collecting a multi-sample image containing all the loosely arranged and placed individuals each time; and finally, cutting all multi-sample images of all category components into single-sample images of target pixels by using an image binarization threshold segmentation method. The method reduces the sample construction difficulty and is high in detection accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of pepper detection, and in particular to a method for detecting the appearance quality of dried pepper based on machine vision. Background Art

[0002] Sichuan pepper has a lot of use value. In addition to the shell can be used as a spice, the seeds can be used to extract oil, and its fruit, roots, stems, and leaves can also be used as medicine. Dried Sichuan pepper is the product of Sichuan pepper picked after maturity and then dried or dried.

[0003] At present, the main way to pick Sichuan pepper is still manual picking, and some impurities such as Sichuan pepper tree stalks, Sichuan pepper thorns and leaves will be picked up during the picking process. Even after manual inspection, specific gravity air separation and screening and impurity removal by vibrating screen, some impurities will inevitably remain. In addition, during the drying and storage process of Sichuan pepper, due to the different growth forms, temperature conditions and storage conditions of each Sichuan pepper, the final dried Sichuan pepper products including impurities will have different appearances. For dried Sichuan pepper, the better the appearance, the higher the quality of the dried Sichuan pepper, the higher the selling price, which can increase the selling price of dried Sichuan pepper and bring higher economic benefits to growers, so it is very important to detect the appearance quality of dried Sichuan pepper.

[0004] Traditional methods of dried peppercorn appearance quality inspection mostly rely on manual sensory evaluation, such as visual inspection, hand holding, nose smelling and mouth tasting. In the quality assessment of peppercorn appearance quality inspection, appearance quality indicators such as particle uniformity, closed eye rate, pepper seed rate, pepper stem rate and color uniformity are all determined by human visual perception, which is difficult to accurately quantify and has low inspection efficiency.

[0005] The detection method of dried pepper using artificial intelligence classification can improve the detection efficiency of dried pepper and realize quantitative detection. However, the existing detection method of dried pepper using artificial intelligence classification has two disadvantages: first, the difficulty in constructing training samples leads to a small number of samples in the sample set used for training, thereby reducing the accuracy of the classification results. Due to the small volume of dried pepper and its impurities, when the training samples are individuals, the existing detection method of dried pepper using artificial intelligence classification needs to be manually placed in an individual separation state. The individual separation density is difficult to control during manual placement, which increases the difficulty of collecting individual samples; when the training samples contain multiple dried peppers, the existing method of detecting dried pepper using artificial intelligence classification only uses a single type of sample after artificial clustering for classification research, which does not match the fact that the dried pepper to be detected contains multiple random categories, and has poor practicality. Second, dried pepper is mostly distinguished based on color texture. When distinguishing by color texture, for dried pepper whose peel is not cracked and the pepper seeds are completely wrapped in the peel, its color is consistent with that of dried pepper whose peel is completely cracked. When distinguishing based on color texture, the two types of pepper are often classified into one category, which reduces the detection accuracy. Summary of the invention

[0006] Based on this, it is necessary to provide a method for detecting the appearance quality of dried peppercorns based on machine vision to address the above-mentioned technical problems. When constructing samples, the samples are loosely arranged in a container, which reduces the difficulty of sample construction. The samples are individually separated and random. The dried peppercorns with the peppercorns completely wrapped in the peel and those with the peppercorns completely detached can be distinguished, thereby improving the detection accuracy.

[0007] The present invention provides a method for detecting the appearance quality of dried prickly ash based on machine vision, comprising:

[0008] Spreading the dried pepper grains to be detected on an arbitrary plane and collecting an image of the dried pepper grains to be detected;

[0009] Inputting the image of the dried pepper grains to be tested into the trained classification neural network model to obtain the appearance quality test result of the dried pepper grains to be tested;

[0010] The construction of a sample set for training a classification neural network model includes the following steps:

[0011] A dried pepper sample particle group containing multiple components is obtained, and the multiple components include single open-eye pepper, plum blossom open-eye pepper, completely closed-eye pepper, incompletely closed-eye pepper, pepper stem, pepper thorn, pepper seed, pepper leaf and variegated pepper;

[0012] Classify the individuals in the dried pepper sample particle group according to different components;

[0013] For each category component, at least one loose arrangement is performed based on a container for dried pepper detection, and multiple sample images containing all loosely arranged individuals are collected each time, wherein the container includes a neutral gray panel with a plurality of equidistant array-arranged grooves on the top;

[0014] The image binarization threshold segmentation method is used to crop all multi-sample images of all category components into single-sample images of target pixels. The single-sample image only includes one dried pepper individual, and all single-sample images constitute the sample set for training the classification neural network model.

[0015] In one embodiment, the container further comprises an enclosure frame, which is rectangular and is composed of four vertical baffles fixedly connected, the panel is horizontally fixed inside the enclosure frame, and the top surface of the enclosure frame is higher than the top surface of the panel;

[0016] A pair of horizontally opposite side walls of the panel are respectively fixedly connected with a first guide plate and a second guide plate, and another pair of horizontally opposite side walls of the panel, the side wall of the first guide plate away from the panel and the side wall of the second guide plate away from the panel are all in contact with a baffle of the enclosure frame.

[0017] In one embodiment, the top surfaces of the first guide plate and the second guide plate are both inclined surfaces, and the top surfaces of the first guide plate and the second guide plate are both inclined downward from the horizontal panel to the enclosure frame, and the inclination angle is 15° to 30°; or

[0018] The top surfaces of the first guide plate and the second guide plate are both V-shaped surfaces or U-shaped surfaces.

[0019] In one of the embodiments, the first guide plate and the second guide plate are each provided with a leak hole;

[0020] The distance between the center points of any two grooves is 5 mm to 20 mm, and the number of grooves is 200 to 500;

[0021] Any pair of two opposite baffles of the enclosure frame is provided with a hand groove.

[0022] In one embodiment, a loose arrangement is performed based on a container for dry pepper testing, wherein a number of dry pepper individuals are taken from components of the same category that have not been loosely arranged and placed on a neutral gray panel, and the container is shaken left and right to arrange the dry pepper individuals loosely in the grooves.

[0023] In one embodiment, the number of single open-eyed peppers in the dry pepper sample particle group is greater than or equal to 3000, the number of plum-shaped open-eyed peppers is greater than or equal to 2000, the number of completely closed-eyed peppers is greater than or equal to 1000, the number of incompletely closed-eyed peppers is greater than or equal to 1000, the number of pepper stalks is greater than or equal to 1500, the number of pepper thorns is greater than or equal to 1000, the number of pepper seeds is greater than or equal to 500, the number of pepper leaves is greater than or equal to 1000, and the number of variegated peppers is greater than or equal to 1000.

[0024] In one of the embodiments, the classification neural network model is a VGG-16, ResNet50, MobileNet, SwinTransformer or Vision Transformer network model, and the number of training iterations of the classification neural network model is greater than 300.

[0025] The beneficial effects of the present invention are:

[0026] (1) The present invention uses a container for dried pepper detection, which includes a neutral gray panel with a plurality of equidistant array grooves on the top, to loosely arrange dried peppers of the same category, thereby achieving equidistant control of individual dried peppers and reducing the difficulty of collecting samples when the training samples are individuals.

[0027] (2) The present invention uses a large number of single sample images to train the classification neural network model. The single sample images include various appearance forms such as completely closed-eye peppers and incompletely closed-eye peppers, which can be used to distinguish between dried peppercorns with completely peeled peppercorns and peppers with completely detached peppercorns. The classification neural network model finally generated has a high detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is one of the flow charts of the method for detecting the appearance quality of dried prickly ash based on machine vision provided in an embodiment of the present invention;

[0029] Figure 2 This is one of the flowcharts of constructing a sample set for training a classification neural network model in an embodiment of the present invention;

[0030] Figure 3 It is a schematic diagram of the structure of a container for dried pepper detection provided by an embodiment of the present invention;

[0031] Figure 4 It is a schematic structural diagram of another angle of the container for dried prickly ash provided in an embodiment of the present invention.

[0032] Explanation of reference numerals: 100, panel; 200, enclosure frame; 300, first guide plate; 400, second guide plate; 500, leakage hole; 600, hand groove; 700, groove. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0034] In one embodiment, Figure 1 As shown, Figure 1 This is one of the flow charts of the method for detecting the appearance quality of dried pepper based on machine vision provided by an embodiment of the present invention. The method for detecting the appearance quality of dried pepper based on machine vision in this embodiment comprises the following steps:

[0035] S101, collecting images of the dry pepper grains to be detected after spreading the dry pepper grains to be detected on any plane.

[0036] Specifically, the dried pepper grains to be tested should be spread out to ensure that the individuals in each component do not overlap.

[0037] S102, inputting the image of the dry pepper grain group to be detected into the trained classification neural network model to obtain the appearance quality detection result of the dry pepper grain group to be detected.

[0038] Among them, Figure 2 As shown, Figure 2 : is a flow chart of the construction of a sample set for training a classification neural network model in an embodiment of the present invention. In this example, the construction of the sample set includes the following steps:

[0039] S201. Obtain a dry pepper sample particle group containing multiple components, wherein the multiple components include single-grain open-eye pepper, plum-shaped open-eye pepper, completely closed-eye pepper, incompletely closed-eye pepper, pepper stems, pepper thorns, pepper seeds, pepper leaves and variegated peppers.

[0040] Among them, variegated peppers include individual dried peppercorns that are inconsistent with the variety of the dried peppercorn group to be tested or have changed color due to improper storage and preparation, fried peppers, and moldy peppers.

[0041] S202, classifying each individual in the dried pepper sample particle group according to different components.

[0042] S203. For each category component, perform at least one loose arrangement based on a container for dried pepper detection, and collect multiple sample images of all loosely arranged individuals each time, wherein the container includes a neutral gray panel 100 with a plurality of equidistant array grooves 700 on the top.

[0043] The panel 100 is a neutral grayscale, that is, 18 degrees gray, and the collected multi-sample images do not need to be further corrected for image color, color temperature, and brightness, which simplifies the single-sample image construction process. In this embodiment, the collection of multi-sample images is not limited to mobile phones and cameras.

[0044] S204, using an image binarization threshold segmentation method to crop all multi-sample images of all category components into a single sample image of a target pixel, wherein the single sample image only includes one dried pepper individual, and all single sample images constitute a sample set for training a classification neural network model.

[0045] Specifically, in this embodiment, single sample images belonging to the same type of comparative components are stored in a folder respectively, and then the data in all folders are combined into a classification neural network model training data set, which can provide open source data for the existing Zanthoxylum bungeanum detection algorithm.

[0046] For example, based on the mainstream mobile phone camera CMO—SONY IMX363 (12 million pixels, unit pixel 1.4μm) 5 years ago, according to the actual application scenario of the camera field of view A4 format (30cm×25cm) and the shooting distance of 15 to 30cm, the target pixel table of each category of single sample image is calculated using the Shaanxi Dahongpao pepper of the season as an example as shown in Table 1.

[0047] Table 1 Target pixel table of single sample images of each category of Shaanxi Dahongpao pepper

[0048]

[0049] In this embodiment, in order to facilitate the distinction of samples, each classification component, subclass and individual is named and numbered, and the specific names and numbers are shown in Table 2.

[0050] Table 2 Naming and numbering of each classification component of dried pepper sample particle group

[0051]

[0052] Specifically, in this embodiment, the number of single-grain open-eye peppers in the dry pepper sample particle group is greater than or equal to 3000, the number of plum-shaped open-eye peppers is greater than or equal to 2000, the number of completely closed-eye peppers is greater than or equal to 1000, the number of incompletely closed-eye peppers is greater than or equal to 1000, the number of pepper stalks is greater than or equal to 1500, the number of pepper thorns is greater than or equal to 1000, the number of pepper seeds is greater than or equal to 500, the number of pepper leaves is greater than or equal to 1000, and the number of variegated peppers is greater than or equal to 1000.

[0053] In this embodiment, samples are loosely arranged in containers when constructing samples, which reduces the difficulty of sample construction. The samples are individually separated and random, and the dried peppercorns with the peppercorn seeds completely wrapped in the peel and the dried peppercorns with the peppercorn seeds completely detached can be distinguished, thereby improving the detection accuracy.

[0054] Specifically, in this embodiment, each evaluation index is quantified into a specific calculation formula according to the national standard "Zanthoxylum bungeanum (GB / T 30391-2013)" and the Shaanxi Provincial Standard "Zanthoxylum bungeanum Quality Grade (DB61 / T72.5-2011)", and the evaluation grade of the dry Zanthoxylum bungeanum grain group to be tested can be calculated based on the number of classification results. The calculation formula and the rating relationship are shown in Table 3.

[0055] Table 3 is the formula for the detection and rating of the appearance quality of Sichuan pepper

[0056]

[0057] In one embodiment, if Figure 3 and Figure 4 As shown, the container further includes an enclosure frame 200 , which is rectangular and is composed of four vertical baffles fixedly connected. The panel 100 is horizontally fixed inside the enclosure frame 200 , and the top surface of the enclosure frame 200 is higher than the top surface of the panel 100 .

[0058] The function of the enclosure frame 200 is to prevent the dried pepper samples from escaping from the panel 100 due to shaking when the components of the dried pepper sample particle group are loosely arranged on the panel 100.

[0059] In this embodiment, a pair of horizontally opposite side walls of the panel 100 are respectively fixedly connected with a first guide plate 300 and a second guide plate 400, and another pair of horizontally opposite side walls of the panel 100, the side wall of the first guide plate 300 away from the panel 100 and the side wall of the second guide plate 400 away from the panel 100 are all abutted against a baffle of the enclosure frame 200.

[0060] The function of the guide plate is to guide the excess component individuals on the panel 100 after each loose arrangement or to change the position of the container after the multi-sample image acquisition is completed so that all individuals on the panel 100 enter the guide plate for drainage.

[0061] In one embodiment, the top surfaces of the first guide plate 300 and the second guide plate 400 are both inclined surfaces, and the top surfaces of the first guide plate 300 and the second guide plate 400 are both inclined downward from the horizontal panel 100 to the enclosure frame 200, with an inclination angle of 15° to 30°; or the top surfaces of the first guide plate 300 and the second guide plate 400 are both V-shaped surfaces or U-shaped surfaces.

[0062] In one embodiment, the first guide plate 300 and the second guide plate 400 are each provided with a leak hole 500; the center point distance between any two grooves 700 is 5 mm to 20 mm, and the number of the grooves 700 is 200 to 500; any pair of opposite baffles of the enclosure frame 200 is provided with a hand groove 600. The leak hole 500 is used to facilitate the individual to discharge the container. The hand groove 600 facilitates the holding of the container.

[0063] It should be noted that the groove 700 in this embodiment may be in a shape of, but not limited to, a columnar shape, a hemispherical shape, or a petal shape.

[0064] The container in this embodiment can ensure that there is a certain density between individual samples, and at the same time, there is no need for manual placement, and each sample body can be naturally in a loose arrangement state.

[0065] In one embodiment, a loose arrangement is performed based on the container used for dry pepper detection. A number of dry pepper individuals are taken from components of the same category that have not been loosely arranged and placed on a neutral gray panel 100, and the container is shaken left and right to make the dry pepper individuals loosely arranged in the groove 700.

[0066] It should be noted that if the number of placements is N times, then each time the first N times of loose arrangement are performed, individuals whose number is greater than 1 / 2 of the number of grooves 700 are randomly selected from the groups of the same category that have not been loosely arranged, until the last time when the number of remaining individuals is less than 1 / 2 of the number of grooves 700, all the remaining individuals are placed.

[0067] In one embodiment, the classification neural network model is a VGG-16, ResNet50, MobileNet, SwinTransformer or Vision Transformer network model, and the number of training iterations of the classification neural network model is greater than 300.

[0068] In this embodiment, 80% of the data of the sample set for training the classification neural network model constructed above is randomly selected as the training set, and the remaining 20% ​​is the test set, and five neural networks, including VGG-16, ResNet50, MobileNet, SwinTransformer, and Vision Transformer, are used for training to further evaluate the accuracy of the algorithm. The results show that Vision Transformer is the optimal neural network and can be used for the accurate classification of pepper and its inherent impurities.

[0069] Table 4 shows the effects and comparison of the five neural network algorithms on the classification of some selected categories of dried pepper. The meanings of the parameters of the selected categories are: BYJ_00 is completely closed-eye pepper, BYJ_10 is incompletely closed-eye pepper, ZYJ_00 is plum-shaped open-eye pepper, ZYJ_10 is single-grain open-eye pepper, ZSJ is variegated pepper, ZZ_JY is pepper leaves, and ZZ_JG&ZZ_JC are pepper stems and pepper thorns.

[0070] Table 4 Comparison of classification effects of five neural networks and dried pepper categories

[0071]

[0072] As can be seen from Table 4, the Vision Transformer (ViT) model has the best comprehensive classification effect on the categories of open-eye pepper, inherent impurities and variegated pepper in dried pepper. The classification algorithm based on the ViT model has an accuracy rate of 92.5% to 100%, an average accuracy rate of 97.83%, a recall rate of 94.44% to 99.87%, and an average recall rate of 97.61%, with the best classification effect.

[0073] The machine vision-based dried pepper appearance quality detection method of this embodiment constructs a sample set for training a classification neural network model that can adapt to a variety of classification network models and has a good classification effect.

[0074] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for detecting the appearance quality of dried pepper based on machine vision. It is characterized in that include: After the dried pepper grain group to be detected is laid out on an arbitrary plane, an image of the dried pepper grain group to be detected is collected; Inputting the image of the dried pepper grains to be tested into the trained classification neural network model to obtain the appearance quality test result of the dried pepper grains to be tested; The construction of the sample set for training the classification neural network model includes the following steps: Obtaining a dried pepper sample particle group containing multiple components, wherein the multiple components include single open-eye pepper, plum blossom open-eye pepper, completely closed-eye pepper, incompletely closed-eye pepper, pepper stem, pepper thorn, pepper seed, pepper leaf and variegated pepper; Classifying the individuals in the dried pepper sample particle group according to different components; For each category component, loosely arranging the dried peppercorns at least once is performed based on a container for dried peppercorn detection, and multiple sample images containing all loosely arranged individuals are collected each time, wherein the container includes a neutral gray panel (100) with a plurality of equidistant array-arranged grooves (700) on the top; All multi-sample images of all category components are cropped into a single-sample image of a target pixel using an image binarization threshold segmentation method. The single-sample image includes only one dried pepper individual, and all single-sample images constitute a sample set for training the classification neural network model.

2. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to claim 1, It is characterized in that The container further comprises an enclosure frame (200), the enclosure frame (200) is rectangular and is composed of four vertical baffles fixedly connected, the panel (100) is horizontally fixed inside the enclosure frame (200), and the top surface of the enclosure frame (200) is higher than the top surface of the panel (100); A pair of horizontally opposite side walls of the panel (100) are respectively fixedly connected to a first guide plate (300) and a second guide plate (400); another pair of horizontally opposite side walls of the panel (100), a side wall of the first guide plate (300) away from the panel (100) and a side wall of the second guide plate (400) away from the panel (100) are all in contact with a baffle of the enclosure frame (200).

3. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to claim 2, It is characterized in that The top surfaces of the first guide plate (300) and the second guide plate (400) are both inclined surfaces, and the top surfaces of the first guide plate (300) and the second guide plate (400) are both inclined downward from the horizontal panel (100) to the enclosure frame (200), and the inclination angle is 15° to 30°; or The top surfaces of the first guide plate (300) and the second guide plate (400) are both V-shaped surfaces or U-shaped surfaces.

4. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to claim 3, It is characterized in that The first guide plate (300) and the second guide plate (400) are both provided with a leakage hole (500); The distance between the center points of any two grooves (700) is 5 mm to 20 mm, and the number of the grooves (700) is 200 to 500; Any pair of two opposing baffles of the enclosure frame (200) is provided with a hand groove (600).

5. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to any one of claims 2 to 4, It is characterized in that The method of loosely arranging the dried peppercorns in a container for detecting the dried peppercorns is to take a number of dried peppercorns from components of the same category that have not been loosely arranged and place them on the neutral gray panel (100), and shake the container left and right to make the dried peppercorns loosely arranged in the groove (700).

6. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to claim 5, It is characterized in that The number of single-grain open-eye peppers in the dry pepper sample particle group is greater than or equal to 3000, the number of plum-shaped open-eye peppers is greater than or equal to 2000, the number of completely closed-eye peppers is greater than or equal to 1000, the number of incompletely closed-eye peppers is greater than or equal to 1000, the number of pepper stalks is greater than or equal to 1500, the number of pepper thorns is greater than or equal to 1000, the number of pepper seeds is greater than or equal to 500, the number of pepper leaves is greater than or equal to 1000, and the number of variegated peppers is greater than or equal to 1000.

7. The method for detecting the appearance quality of dried Chinese prickly ash based on machine vision according to claim 6, It is characterized in that The classification neural network model is a VGG-16, ResNet50, MobileNet, Swin Transformer or VisionTransformer network model, and the number of training iterations of the classification neural network model is greater than 300.

Citation Information

Patent Citations

  • Apple grading method and system based on machine vision

    CN110276386A

  • Intelligent fruit optimal selection equipment with intelligent fine grading function

    CN110743819A

  • Rice quality detection method, device, server and system

    CN113240653A

  • Fluorimeter sampling apparatus

    US4443104A