Seabuckthorn Fruit Screening Equipment Based on Machine Vision

Abstract

This application relates to a seabuckthorn fruit screening device based on machine vision, including a transport table, a first conveyor belt and a second conveyor belt wound around the transport table, screening holes evenly distributed on the second conveyor belt, an image acquisition and analysis module arranged on the transport table and facing the screening holes on the second conveyor belt, and a rejector arranged on the transport table. The image acquisition and analysis module is used to acquire the object to be measured in the screening holes and analyze the surface of the object to be measured, discover surface defects of the object to be measured. The rejector is used to reject unqualified objects to be measured falling from the second conveyor belt according to the feedback of the image acquisition and analysis module. The moving directions of the first conveyor belt and the second conveyor belt are opposite. This application provides a seabuckthorn fruit screening device based on machine vision. By actively driving and rotating, a complete surface image of the object to be measured can be obtained, and then an automated analysis method is used to determine the scars and reject accordingly, which can effectively improve the product quality of the object to be measured.

Description

Technical Field

[0001] This application relates to the field of automated screening technologies, and particularly to a sea buckthorn fruit screening device based on machine vision. Background Art

[0002] Due to its cold tolerance, drought tolerance, salt-alkali tolerance, and economic value, sea buckthorn is widely used in the field of desert control because it can endow desert control projects with economic attributes, as sea buckthorn fruits can be made into economic commodities (food, beverages, medicinal materials).

[0003] The production process generally includes procedures such as harvesting, cleaning, screening and grading, and processing. Currently, most harvesting methods are carried out in a low-temperature environment, aiming to reduce the harvesting difficulty and ensure the integrity of the fruits by freezing the sea buckthorn fruits at low temperature. Cleaning, screening, and grading have also achieved mechanical automation with the help of bubble cleaning machines, vibrating screening machines, and air separators.

[0004] However, after screening and grading, it is still necessary to remove defective fruits. This is mainly because the skin of sea buckthorn fruits is thin and prone to being bruised by pressing and squeezing. Such types of bruises are not obvious. If directly sold as commodities or used for processed foods, it will directly affect the final product quality and price. Summary of the Invention

[0005] This application provides a sea buckthorn fruit screening device based on machine vision. By actively driving rotation to obtain a complete surface image of the object to be measured, and then using an automated analysis method to determine the bruises and perform rejection accordingly, the quality of the product to be measured can be effectively improved.

[0006] The above object of this application is achieved through the following technical solutions:

[0007] This application provides a sea buckthorn fruit screening device based on machine vision, including:

[0008] A transport table;

[0009] A first conveyor belt and a second conveyor belt, both wound around the transport table, with the first conveyor belt located inside the second conveyor belt;

[0010] Screening holes, evenly distributed on the second conveyor belt;

[0011] An image acquisition and analysis module, arranged on the transport table and facing the screening holes on the second conveyor belt. The image acquisition and analysis module is used to acquire the object to be measured in the screening holes and analyze the surface of the object to be measured to detect surface defects of the object to be measured;

[0012] A rejector, arranged on the transport table. The rejector is used to reject unqualified objects to be measured that fall from the second conveyor belt according to the feedback of the image acquisition and analysis module;

[0013] Among them, the moving directions of the first conveyor belt and the second conveyor belt are opposite.

[0014] In a possible implementation manner of the present application, the surface defects detected by the image acquisition and analysis module include:

[0015] Collect the images of the coverage area and extract the object to be measured in the images, denoted as the object image to be measured;

[0016] Perform gray-scale processing on the object image to be measured to obtain a gray-scale image;

[0017] Statistically analyze the gray-scale value distribution in the gray-scale image, delete the areas corresponding to the gray-scale values with a proportion exceeding the set ratio, and obtain suspected surface defects;

[0018] Determine the types of the suspected surface defects, and the types include adhesion, discoloration, wrinkle, and depression;

[0019] Screen the suspected surface defects according to the types and areas of the suspected surface defects to obtain surface defects.

[0020] In a possible implementation manner of the present application, it further includes:

[0021] Decompose the object image to be measured into monochromatic images, a red object image to be measured, a green object image to be measured, and a blue object image to be measured;

[0022] Perform gray-scale processing on the monochromatic images to obtain a red gray-scale object image to be measured, a green gray-scale object image to be measured, and a blue gray-scale object image to be measured;

[0023] Search for suspected surface defects in the red gray-scale object image to be measured, the green gray-scale object image to be measured, and the blue gray-scale object image to be measured.

[0024] In a possible implementation manner of the present application, determining the types of the suspected surface defects includes:

[0025] Determine the height change of the suspected surface defect;

[0026] When there is no height change in the suspected surface defect, the type of the suspected surface defect is determined to be discoloration;

[0027] When there is a height change in the suspected surface defect and the height change is located outside the suspected surface defect, the type of the suspected surface defect is determined to be adhesion;

[0028] When there is a height change in the suspected surface defect and the height change is located inside the suspected surface defect, the type of the suspected surface defect is determined to be wrinkle;

[0029] When there is a height change and an overall color change or the overall color remains unchanged in the suspected surface defect, the type of the suspected surface defect is determined to be depression.

[0030] In a possible implementation manner of the present application, determining the height change of a suspected surface defect includes:

[0031] Determine a tracking point on the suspected surface defect, and the color value of the tracking point is the maximum or minimum color value within the corresponding range of the suspected surface defect;

[0032] Based on the tracking point, establish a tracking range, and the tracking point is located at the center position of the tracking range;

[0033] Dynamically track the tracking point and record the color change situation within the tracking range. If there is a color change situation within the tracking range, it is determined that the suspected surface defect includes a height change.

[0034] In a possible implementation manner of the present application, the shape of the tracking range is a rectangle; the length direction of the tracking range is consistent with the color change direction of the suspected surface defect.

[0035] In a possible implementation manner of the present application, determining a tracking point on the suspected surface defect includes:

[0036] Establish an extraction domain according to the maximum or minimum color value within the corresponding range of the suspected surface defect;

[0037] Use the extraction domain to extract on the suspected surface defect to obtain an extraction map;

[0038] Take the content on the extraction map as the tracking point.

[0039] In a possible implementation manner of the present application, taking the content on the extraction map as the tracking point includes:

[0040] Establish a closed circle so that the content on the extraction map is located inside the closed circle;

[0041] Use the content at the edge, center position, and middle position of the closed circle as the tracking point.

[0042] In a possible implementation manner of the present application, after taking the content on the extraction map as the tracking point, it further includes recording the shape of each tracking point.

[0043] In a possible implementation manner of the present application, it further includes using the tracking point to draw a tracking grid, and the number of sides of each grid in the tracking grid is the same.

[0044] The beneficial effects of the present application are:

[0045] The seabuckthorn fruit screening device based on machine vision provided by the present application uses a continuous rolling method to obtain the surface image of the object to be measured, and then analyzes the obtained surface image. During the analysis process, the defect area and the type of the defect area on the surface of the object to be measured can be determined. Based on this, the unqualified objects to be measured can be removed, and the qualified objects to be measured (seabuckthorn fruits) with better consistency can be obtained. Description of the Drawings

[0046] Figure 1 is a structural schematic diagram of a seabuckthorn fruit screening device provided by the present application. The arrow in the figure indicates the moving direction of the object to be measured.

[0047] Figure 2 is a schematic diagram of the relative positions of a first conveyor belt and a second conveyor belt provided by the present application. The arrow in the figure indicates the rotation direction.

[0048] Figure 3 is a schematic diagram of the distribution of screening holes on a second conveyor belt provided by the present application.

[0049] Figure 4 is a schematic block diagram of the steps for discovering surface defects provided by the present application.

[0050] Figure 5 is a schematic diagram of decomposing the image of the object to be measured provided by the present application.

[0051] Figure 6 is a schematic diagram of a tracking range provided by the present application.

[0052] Figure 7 is a schematic diagram of the principle of color change occurring within the tracking range provided by the present application.

[0053] In the figure, 1 is a transportation platform, 2 is a first conveyor belt, 3 is a second conveyor belt, 4 is a screening hole, 5 is an image acquisition and analysis module, and 6 is an ejector. Detailed Embodiments

[0054] The following further elaborates on the technical solutions in the present application in conjunction with the accompanying drawings.

[0055] The seabuckthorn fruit screening device based on machine vision disclosed in the present application is deployed after the screening and grading process in the production process. The purpose is to remove bad fruits. Further elaborating in combination with the content mentioned in the background technology, the harvested seabuckthorn fruits go through winnowing (removing dust and impurities) and water washing (removing floating soil on the surface), and then enter the screening and grading process. In this process, the seabuckthorn fruits are graded according to their diameters through a layered vibration method, generally graded into three to four levels, and the particle sizes of the seabuckthorn fruits in each level are within a set range. However, the bad fruits cannot be removed during the grading process, and potential surface damage may also be caused during the grading process. Therefore, after the grading process, it is necessary to remove the bad seabuckthorn fruits.

[0056] Please refer to Figure 1 , in some examples, the machine vision-based seabuckthorn fruit screening device disclosed in the present application includes a transport table 1, a first conveyor belt 2, a second conveyor belt 3, an image acquisition and analysis module 5, and an eliminator 6. The first conveyor belt 2 and the second conveyor belt 3 are both wound around the transport table 1, and the first conveyor belt 2 is located inside the second conveyor belt 3.

[0057] Both the first conveyor belt 2 and the second conveyor belt 3 are driven by motors. Please refer to Figure 2 , the moving directions of the two conveyor belts are opposite. The purpose of the opposite moving directions is to drive the measured objects (seabuckthorn fruits) in the screening holes 4 to be in a rotating state, because only in this way can the entire surface of the measured objects be exposed within the field of view of the image acquisition and analysis module 5.

[0058] On both sides above the second conveyor belt 3 (near the two ends of the first conveyor belt 2), it is supported by the first conveyor belt 2. The drive motor of the first conveyor belt 2 is located on the left side of the first conveyor belt 2, and the drive motor of the second conveyor belt 3 is located below the middle position of the second conveyor belt 3.

[0059] Please refer to Figure 3 , the screening holes 4 are evenly distributed on the second conveyor belt 3 and arranged in the form of an MxN matrix, where both M and N are natural numbers greater than zero.

[0060] The image acquisition and analysis module 5 is installed on the transport table 1 and is used to acquire the measured objects in the screening holes 4 and analyze the surface of the measured objects to detect surface defects of the measured objects. In some examples, the image acquisition and analysis module 5 mainly consists of two parts: an image sensor and a graphics processor. Generally, a high-speed camera is used as the image sensor.

[0061] It should be noted that generally, the machine vision-based seabuckthorn fruit screening device disclosed in the present application needs to be deployed in an enclosed space, and an artificial light source is deployed inside the enclosed space (to improve the uniformity of illumination). The first conveyor belt 2 and the second conveyor belt 3 are made of non-reflective materials, and their colors need to have a distinct difference from the color of the seabuckthorn fruits. Generally, black is used, so that the seabuckthorn fruits can be clearly displayed in the image.

[0062] The eliminator 6 is arranged on the transport table 1 and is used to eliminate the unqualified measured objects falling from the second conveyor belt 3 according to the feedback of the image acquisition and analysis module 5. The specific implementation method is as follows:

[0063] Since the screening holes 4 are arranged in a matrix, on the surface of the second conveyor belt 3, digital marks can be made on one side. At this time, the row numbers of the screening holes 4 can be determined, and through the column numbers of the screening holes 4, the coordinates of the screening holes 4 can be obtained.

[0064] That is, if the measured object in a screening hole 4 is determined to be a bad fruit, it has a specific number (x, y). Refer to Figure 3 the number (1, 5, 10) in. The number represents the row number. At this time, the serial number from left to right or from right to left can be used to obtain (x, y).

[0065] At the same time, by appropriately controlling the rotation speed of the second conveyor belt 3, one row of seabuckthorn fruits can be made to fall during each dropping process. When the ejector 6 at the corresponding column number is activated, this bad fruit can be removed.

[0066] In some examples, the ejector 6 is composed of a cylinder and a swing arm. The cylinder drives the swing arm to rotate and pushes the bad fruit away from the established dropping trajectory.

[0067] In other examples, the ejector 6 can use a linear motor.

[0068] At the same time, a camera or a proximity sensor can also be configured for the ejector 6 to detect the position of the measured object and determine a clear activation time for the ejector 6.

[0069] In some examples, please refer to Figure 4 , the specific steps for the image acquisition and analysis module 5 to detect the surface defects of the measured object are as follows:

[0070] S101, Collect the image of the coverage area and extract the measured object in the image, denoted as the measured object image;

[0071] S102, Perform gray-scale processing on the measured object image to obtain a gray-scale image;

[0072] S103, Statistically analyze the gray-scale value distribution in the gray-scale image, delete the area corresponding to the gray-scale value whose proportion exceeds the set ratio, and obtain suspected surface defects;

[0073] S104, Determine the type of the suspected surface defect, and the types include attachment, discoloration, wrinkle, and depression;

[0074] S105, Screen the suspected surface defects according to the type and area of the suspected surface defects to obtain surface defects.

[0075] In step S101, first collect the image of the coverage area and extract the measured object in the image, denoted as the measured object image. Combining the content in the previous text, it can be seen that because there is an obvious color difference between the measured object and the background, only color discrimination can be used to achieve extraction here, or it can be described as deleting the conveyor belt in the collected image of the coverage area.

[0076] In step S102, it is necessary to perform gray-scale processing on the measured object image. At this time, the obtained image is called a gray-scale image. The purpose of gray-scale processing is to reduce the amount of data processing.

[0077] In step S103, the distribution of gray values in the grayscale image is counted, and the areas corresponding to the gray values whose proportion exceeds the set ratio are deleted to obtain suspected surface defects. The purpose here is to eliminate the normal areas on the object under test.

[0078] It should be noted that the object under test entering this step mainly has a small amount or even a trace amount of surface defects, and the bad fruits and unripe fruits have been eliminated in the previous processing steps. The fruits in this step have the obvious characteristics of being basically ripe and having a uniform surface color.

[0079] Generally, the set ratio is controlled within 15%-20% here, and the areas smaller than this ratio are regarded as suspected surface defects. It should be noted that for the corresponding areas of the suspected surface defects on the grayscale image, the content of the corresponding areas is not deleted, but all retained.

[0080] In step S104, it is necessary to determine the types of the suspected surface defects. In this application, the types are divided into four categories, namely attachment, discoloration, wrinkle and depression. For the sea buckthorn fruits that have been cleaned and screened, the attachment is mainly black spots, which are formed during the growth process of the sea buckthorn fruits. Generally, they are not used as the criteria for quality evaluation. However, for some orders, the number of black spots will be restricted. Here, it is mainly related to the appearance problem.

[0081] Discoloration, wrinkle and depression are the manifestations after the surface of the sea buckthorn fruits is damaged.

[0082] Finally, in step S105, the suspected surface defects are screened according to the types and areas of the suspected surface defects to obtain the surface defects.

[0083] For example, generally two criteria are implemented for screening. The first criterion is the cumulative value. For example, the area value of the black spots is not allowed to exceed a preset value. The second criterion is the maximum value. For example, the area value of the largest black spot among all the black spots is not allowed to exceed a preset value.

[0084] Of course, these two criteria can also be used simultaneously.

[0085] In some examples, the following steps are also added:

[0086] The image of the object under test is decomposed into monochromatic images, a red image of the object under test, a green image of the object under test and a blue image of the object under test, as Figure 5 shown;

[0087] The monochromatic images are subjected to grayscale processing to obtain a red grayscale image of the object under test, a green grayscale image of the object under test and a blue grayscale image of the object under test;

[0088] Search for suspected surface defects in the red grayscale image to be measured, the green grayscale image to be measured, and the blue grayscale image to be measured.

[0089] The above method is to perform color separation on the image to be measured before grayscale processing. After obtaining the red image to be measured, the green image to be measured, and the blue image to be measured, grayscale processing is performed on the red image to be measured, the green image to be measured, and the blue image to be measured respectively.

[0090] The advantages of this method are as follows:

[0091] The gradient calculation of a monochromatic image is more direct. For example, the object contour can be quickly located through the Sobel operator or the Canny edge detector, avoiding interference between color channels. In addition, the homogeneity of the grayscale image (such as pixel brightness consistency) is convenient for region-based segmentation algorithms.

[0092] Suspected surface defects are composed of three colors: red, green, and blue, and the proportions of these three colors are not the same. For example, in some cases, when observed only with a certain color, its display may be more obvious, or a certain monochromatic image can be enhanced to obtain a higher recognition rate.

[0093] In some examples, the specific method for determining the type of suspected surface defect is as follows:

[0094] Determine the height change of the suspected surface defect;

[0095] When there is no height change in the suspected surface defect, the type of the suspected surface defect is determined as color change;

[0096] When there is a height change in the suspected surface defect and the height change is outside the suspected surface defect, the type of the suspected surface defect is determined as attachment;

[0097] When there is a height change in the suspected surface defect and the height change is inside the suspected surface defect, the type of the suspected surface defect is determined as wrinkle;

[0098] When there is a height change and an overall color change or the overall color remains unchanged in the suspected surface defect, the type of the suspected surface defect is determined as depression.

[0099] In the above method, the type of suspected surface defect is determined by color change and height change. The specific rules are:

[0100] Color change: There is no height change in the suspected surface defect;

[0101] Attachment: There is a height change in the suspected surface defect, and the height change is outside the suspected surface defect;

[0102] Wrinkle: There is a height change in the suspected surface defect, and the height change is inside the suspected surface defect;

[0103] Indentation: There is a height change, an overall color change, or no overall color change in the suspected surface defect.

[0104] In some examples, the way to determine the height change of the suspected surface defect is as follows:

[0105] S201. Determine a tracking point on the suspected surface defect. The color value of the tracking point is the maximum or minimum color value within the corresponding range of the suspected surface defect;

[0106] S202. Establish a tracking range based on the tracking point. The tracking point is located at the center of the tracking range;

[0107] S203. Dynamically track the tracking point and record the color change situation within the tracking range. If there is a color change within the tracking range, it is determined that the suspected surface defect includes a height change.

[0108] In steps S201 to S203, first, a tracking point needs to be determined on the suspected surface defect. The color value of the tracking point is the maximum or minimum color value within the corresponding range of the suspected surface defect.

[0109] In some possible implementation manners, the tracking point is a combination of multiple pixel points (MxN matrix).

[0110] Please refer to Figure 6 , and then establish a tracking range based on the tracking point. The tracking point is located at the center of the tracking range. Generally, a circular structure is used for the tracking range because if there is a color change around the tracking point, the direction of the color change is uncertain, and the circular structure ensures full coverage of the circumferential direction when the direction cannot be determined.

[0111] Finally, dynamically track the tracking point and record the color change situation within the tracking range. If there is a color change within the tracking range, it is determined that the suspected surface defect includes a height change. The reason for the change here is that as the measured object rotates, the illumination conditions on its surface will also change.

[0112] It should be clear here that there may be two situations for the color change within the tracking range. The first situation is that the color change within the tracking range is consistent with the color change outside the tracking range, and the second situation is that the color change within the tracking range is inconsistent with the color change outside the tracking range.

[0113] When the second situation occurs, it indicates that the suspected surface defect includes a height change, as shown in Figure 7 . Conversely, the suspected surface defect does not include a height change.

[0114] In some possible implementations, the shape of the tracking range is rectangular, and the length direction of the tracking range is consistent with the color change direction of the suspected surface defect. That is, after the color of the suspected surface defect starts to change, the color change direction of the suspected surface defect can be determined. At this time, by adjusting the shape of the tracking range, on the one hand, accurate tracking of the color change can be achieved, and on the other hand, the area of the tracking range can be reduced to reduce the data processing volume and improve the data processing speed.

[0115] The specific method for determining the tracking points on the suspected surface defect is as follows:

[0116] S301, establish an extraction domain based on the maximum color value or the minimum color value within the corresponding range of the suspected surface defect;

[0117] S302, use the extraction domain to extract on the suspected surface defect to obtain an extraction map;

[0118] S303, use the content on the extraction map as the tracking points.

[0119] In steps S301 to S303, an extraction domain will be first established based on the maximum color value or the minimum color value within the corresponding range of the suspected surface defect. The maximum color value or the minimum color value is located at the midpoint or the end point within the corresponding extraction domain.

[0120] Then, use the extraction domain to extract on the suspected surface defect to obtain an extraction map. At this time, the obtained content are all pixel points related to the maximum color value or the minimum color value (extraction domain). Finally, use the content on these extraction maps as the tracking points.

[0121] Furthermore, the area of the tracking points was mentioned above. The content on the extraction map includes three cases: larger than the area of the tracking points, equal to the area of the tracking points, and smaller than the area of the tracking points. The content corresponding to the area smaller than the tracking points is deleted, and the content larger than and equal to the area of the tracking points is retained. For the content larger than the area of the tracking points, a part of its edge is selected as the tracking points for use.

[0122] Next, use the following steps for processing:

[0123] Establish a closed circle so that the content on the extraction map is located inside the closed circle;

[0124] Use the content at the edge, the center position, and the middle position of the closed circle as the tracking points.

[0125] The purpose of the above two steps is to ensure the uniformity of the distribution of the tracking points. The middle position is generally at 40%-70% of the radius length of the closed circle.

[0126] If the above method cannot be implemented, it is processed by adjusting the extraction field. The length of the extraction field is generally controlled at about 15 - 30. When adjusting, the length of the extraction area can be adjusted, or the position of the maximum or minimum color value in the extraction field can be adjusted, for example, from the midpoint to the endpoint or other positions.

[0127] In some examples, after taking the content on the extraction graph as the tracking points, it further includes recording the shape of each tracking point, aiming to facilitate continuous tracking of the tracking points.

[0128] Furthermore, it also includes using the tracking points to draw a tracking grid, and the number of sides of each grid in the tracking grid is the same. Its purpose is also to facilitate continuous tracking of the tracking points.

[0129] It should be understood that during the continuous movement of the tracking points within a small range, their shapes and relative positions basically do not change relatively. In this application, multiple images that appear sequentially in a time series are processed separately, and after the processing is completed, the matching of the tracking points is involved.

[0130] The matching work of this application is determined by the shape and relative position of the tracking points. When performing the matching, an error range also needs to be set for the shape and relative position. The error range of the relative position is generally 1% - 2%, and the error range of the shape is generally 1 - 2 pixel points.

[0131] The embodiments of this specific implementation manner are all preferred embodiments of this application, and do not limit the protection scope of this application accordingly. Therefore, all equivalent changes made according to the structure, shape, and principle of this application should be covered within the protection scope of this application.

Claims

1. A screening method for seabuckthorn fruits based on machine vision, characterized in that, Including: Transport platform (1); The first conveyor belt (2) and the second conveyor belt (3), both wound around the transport platform (1), and the first conveyor belt (2) is located inside the second conveyor belt (3); Sieving holes (4), evenly distributed on the second conveyor belt (3); Image acquisition and analysis module (5), arranged on the transport platform (1) and facing the sieving holes (4) on the second conveyor belt (3), and the image acquisition and analysis module (5) is used to acquire the object under test in the sieving holes (4) and analyze the surface of the object under test to detect surface defects of the object under test; Rejector (6), arranged on the transport platform (1), and the rejector (6) is used to reject unqualified objects under test falling from the second conveyor belt (3) according to the feedback of the image acquisition and analysis module (5); Wherein, the moving directions of the first conveyor belt (2) and the second conveyor belt (3) are opposite; When the image acquisition and analysis module (5) detects a surface defect of the object under test, it is necessary to determine the suspected surface defect and the type of the suspected surface defect, and the type of the suspected surface defect is determined according to the height change of the suspected surface defect; When determining the height change of the surface defect, it is necessary to determine tracking points on the suspected surface defect; Determining the height change of the suspected surface defect includes: Determining tracking points on the suspected surface defect, and the color value of the tracking points is the maximum or minimum color value within the corresponding range of the suspected surface defect; Establishing a tracking range based on the tracking points, and the tracking points are located at the center position of the tracking range; Dynamically tracking the tracking points and recording the color change situation within the tracking range, and if there is a color change situation within the tracking range, it is determined that the suspected surface defect includes a height change; Determining tracking points on the suspected surface defect includes: Establishing an extraction domain according to the maximum or minimum color value within the corresponding range of the suspected surface defect; Using the extraction domain to extract on the suspected surface defect to obtain an extraction map; Taking the content on the extraction map as the tracking points.

2. The method for screening seabuckthorn fruits based on machine vision according to claim 1, wherein The image acquisition and analysis module (5) detecting a surface defect of the object under test includes: Acquiring an image of the coverage area and extracting the object under test in the image, denoted as the object under test image; Performing gray-scale processing on the object under test image to obtain a gray-scale image; Statistically analyzing the gray-scale value distribution in the gray-scale image, deleting the area corresponding to the gray-scale value with a proportion exceeding the set ratio to obtain a suspected surface defect; Determining the type of the suspected surface defect, and the types include attachment, discoloration, wrinkle and depression; Screening the suspected surface defect according to the type of the suspected surface defect and the area of the suspected surface defect to obtain a surface defect.

3. The method for screening seabuckthorn fruits based on machine vision according to claim 2, wherein, It also includes: Decomposing the object under test image into monochromatic images, a red object under test image, a green object under test image and a blue object under test image; Performing gray-scale processing on the monochromatic images to obtain a red gray-scale object under test image, a green gray-scale object under test image and a blue gray-scale object under test image; Searching for suspected surface defects on the red gray-scale object under test image, the green gray-scale object under test image and the blue gray-scale object under test image.

4. The method for screening seabuckthorn fruits based on machine vision according to claim 2, wherein, Determining the type of the suspected surface defect includes: Determining the height change of the suspected surface defect; When there is no height change in the suspected surface defect, the type of the suspected surface defect is determined as discoloration; When there is a height change in the suspected surface defect and the height change is outside the suspected surface defect, the type of the suspected surface defect is determined to be adhesion; When there is a height change in the suspected surface defect and the height change is inside the suspected surface defect, the type of the suspected surface defect is determined to be a fold; When there is a height change and an overall color change or the overall color remains unchanged in the suspected surface defect, the type of the suspected surface defect is determined to be a depression.

5. The method for screening seabuckthorn fruits based on machine vision according to claim 1, characterized in that The shape of the tracking range is rectangular; the length direction of the tracking range is consistent with the color change direction of the suspected surface defect.

6. The method for screening seabuckthorn fruits based on machine vision according to claim 1, characterized in that, Taking the content on the extraction diagram as tracking points includes: Creating a closed circle with the content on the extraction diagram located inside the closed circle; Using the content at the edge, the center position, and the middle position of the closed circle as tracking points.

7. The method for screening seabuckthorn fruits based on machine vision according to claim 1, characterized in that After taking the content on the extraction diagram as tracking points, it also includes recording the shape of each tracking point.

8. The method for screening seabuckthorn fruits based on machine vision according to claim 6, wherein It also includes using the tracking points to draw a tracking grid, and the number of sides of each grid in the tracking grid is the same.

Images