Dining table dish satisfaction degree evaluation method based on image recognition
The rotating disc area of the dining table is calibrated through image recognition technology, and the dishes are automatically tracked and evaluated, which solves the problem that the restaurant is difficult to accurately obtain customer satisfaction and provides an efficient dish evaluation method.
Patent Information
- Application Number
- CN202510728042.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-03
- Publication Date
- 2025-07-18
AI Technical Summary
It is difficult for the restaurant to accurately obtain the satisfaction of guests with the dishes on the table, and the existing methods are not accurate.
The rotating disc area of the dining table is calibrated through image recognition technology, and the dishes images at the end of the meal are obtained, and the position information of the rotating disc is used for image pairing to evaluate the satisfaction of the dishes.
Automatic dish status tracking and satisfaction analysis without manual intervention is realized, providing restaurants with objective and efficient dish evaluation methods.
Smart Images
Figure CN120339660A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of image processing, and particularly relates to a method for evaluating the satisfaction of table dishes based on image recognition. Background Art
[0002] For the catering industry, the satisfaction of guests with table dishes is particularly important for the subsequent business improvement of the restaurant. The restaurant can increase the purchase volume of ingredients for the dishes that guests like, and can improve the taste of the dishes that guests don't like very much, etc. However, currently, the way for the restaurant to obtain the satisfaction of table dishes is basically to learn from the guests after the meal. However, this way of obtaining has low accuracy, resulting in the restaurant being difficult to accurately obtain the satisfaction of guests with the table dishes for each meal. Summary of the Invention
[0003] In order to overcome the defects existing in the prior art, the present invention provides a method for evaluating the satisfaction of table dishes based on image recognition to solve the above problems.
[0004] The technical solution adopted by the present invention to solve its technical problems is: A method for evaluating the satisfaction of table dishes based on image recognition, comprising the following steps: S1: Calibrate and train the image obtained by the camera, and calibrate the rotating disc image area in the image; S2: During the process of serving dishes, obtain the first image calibrated with the information of the rotating disc image area, associate the position corresponding to the rotating disc image area of the table with each dish for the current meal in the first image, and obtain the initial dish image of each dish when serving; S3: At the end of the meal, obtain the second image calibrated with the information of the rotating disc image area, and obtain the end dish image of each dish in the second image; S4: Pair the initial dish image and the end dish image belonging to the same dish through the position of the rotating disc image area of the table; S5: Use the degree of change obtained by comparing the paired initial dish image and end dish image as the satisfaction of the table dishes.
[0005] Preferably, in the step S5, extract the key point features from the paired initial dish image and end dish image; According to the key point features corresponding to the initial dish image and the key point features corresponding to the end dish image, obtain the key point features that match the key point features corresponding to the end dish image in the initial dish image through the BF algorithm; A projection transformation matrix is calculated based on the matched key-point features and the end dish image. If the calculation of the projection transformation matrix fails, it indicates that the degree of change of the dish is 100%. If the calculation of the projection transformation matrix is successful, after projecting the matched key-point features onto the end dish image, the points whose distances between the matched key-point features and the original key-point features in the end dish image are less than the threshold are taken as inliers, and the ratio of the number of inliers to the total number of the matched key-point features is calculated as the degree of change.
[0006] Optionally, in the BF algorithm of the step S5, the matching mechanism is to set a distance threshold, and the key-point features in the initial dish image whose distances from the key-point features of the end dish image are within the distance threshold are taken as the matched key-point features.
[0007] Specifically, in the step S2, for the rotating disk image region in the first image, with the center of the rotating disk image region as the origin, the rotating disk image region is evenly divided into multiple fan-shaped sub-regions at a preset angle, and a unique position code is assigned to each fan-shaped sub-region. When serving dishes, a unique dish code is assigned to each dish. After the dish is placed in the fan-shaped sub-region, the dish code corresponding to the dish is associated with the position code corresponding to the fan-shaped sub-region, and the associated dish code and position code are marked on the initial dish image corresponding to the dish.
[0008] It should be noted that in the step S4, the position code corresponding to the fan-shaped sub-region in the end dish image is obtained, and the position code and the associated dish code are marked on the end dish image. The initial dish image and the end dish image with the same position code and dish code are paired as the same dish.
[0009] Specifically, in the step S1, a first color curtain is set in the rotating disk area of the table, an image of the table is collected, and the image area of the first color curtain is extracted to obtain the rotating disk image area.
[0010] Preferably, the step S1 includes: S11: Perform hsv conversion on the collected image of the table, convert the image of the table from an RGB image to an hsv image, and then convert the pixel points of the first color extracted from the hsv image into a binary image; S12: Extract connected regions in the binary image, remove the individual green pixel points generated by noise, and obtain an annular edge polygon corresponding to the shape of the rotating disk; S13: Dilate and erode the annular edge polygon to fill the inside of the annular edge polygon and obtain the rotating disk image area.
[0011] Optionally, in the step S11, the first color is green. For the green curtain, the green pixel points extracted from the hsv image are converted into a binary image through the function cv2.inrange(hsv, minGreen, maxGreen) in opencv, where hsv is the hsv image corresponding to the dining table, minGreen = (50, 50, 50), and maxGreen = (70, 255, 255).
[0012] It should be noted that in the step S12, the connected regions are extracted through the function cv2.connectedComponents(image, connectivity, ltype) in opencv, where image is the binary image, connectivity is the connected domain, connectivity = 8, and ltype is the type of the function output, ltype = CV_32.
[0013] Preferably, in the step S13, the annular edge polygon corresponding to the rotating disk is inflated by 200% through the function shapely.buffer(2.) to fill the interior of the annular edge polygon, obtaining an inflated circular region; The inflated circular region is eroded by 200% through the function shapely.buffer(-2.) to reduce the area of the circular region to be consistent with the actual area of the rotating disk, obtaining the rotating disk image region.
[0014] The beneficial effect of the present invention is as follows: In the method for evaluating the satisfaction of dining table dishes based on image recognition, a calibrated image of the rotating disk area of the dining table is obtained through a camera. When serving dishes, the initial dish images of each dish and their positions on the rotating disk are recorded. After the meal, the end dish images are taken again. The present invention utilizes the position information of the rotating disk to pair the initial dish images and the end dish images of the same dish, and evaluates the satisfaction of customers with the dish by comparing the change degree of the two images. This method requires no manual intervention and can automatically complete the tracking of the dish status and the analysis of satisfaction during the entire dining process, providing an objective and efficient dish evaluation means for the restaurant. Description of the Drawings
[0015] Figure 1 It is a flowchart of the method for evaluating the satisfaction of dining table dishes based on image recognition in an embodiment of the present invention; Figure 2 It is a schematic diagram of the division of the rotating disk image region and the sector sub-regions in the dining table image in an embodiment of the present invention. Detailed Embodiments
[0016] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted here that the description of these embodiments is for helping to understand the present invention, but does not constitute a limitation to the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0017] As Figure 1 and 2 shown, a method for evaluating the satisfaction of table dishes based on image recognition includes the following steps: S1: Calibrate and train the image obtained by the camera, and calibrate the rotating disc image area in the image; S2: During the process of serving dishes, obtain the first image calibrated with the information of the rotating disc image area, associate the position corresponding to the rotating disc image area of the table with each dish currently being eaten in the first image, and obtain the initial dish image of each dish when it is served; S3: At the end of the meal, obtain the second image calibrated with the information of the rotating disc image area, and obtain the end dish image of each dish in the second image; in this solution, the mark of the end of the meal can be obtained in the following way: recognize the end-of-meal instruction through voice recognition and form an end-of-meal instruction after the time when the projection of the human hand does not overlap with the rotating disc image area exceeds a preset time; specifically, an end-of-meal instruction is formed after the time when the projection of the human hand does not overlap with the rotating disc image area exceeds a preset time threshold. During the meal, the diners will pick up dishes, so within the set preset time threshold, the projection of the human hand will overlap with the rotating disc image area. When the time when the projection of the human hand does not overlap with the rotating disc image area exceeds the preset time threshold, it means that no one is picking up dishes, that is, the meal is over; in this embodiment, the relationship between the projection of the human hand and the rotating disc image area can be recognized through human body key points; S4: Pair the initial dish image and the end dish image belonging to the same dish through the position of the rotating disc image area of the table; S5: Use the degree of change of the paired initial dish image and end dish image as the satisfaction of the table dishes.
[0018] In the method for evaluating the satisfaction of table dishes based on image recognition, a calibrated image of the table rotating disk area is obtained through a camera. When serving dishes, the initial dish image of each dish and its position on the rotating disk are recorded. After the meal, the end dish image is taken again. The present invention utilizes the position information of the rotating disk to pair the initial dish image and the end dish image of the same dish, and evaluates the satisfaction of the customer with the dish by comparing the change degree of the two images. This method requires no manual intervention and can automatically complete the tracking of the dish status and the analysis of satisfaction during the entire dining process, providing an objective and efficient dish evaluation means for the restaurant.
[0019] It should be noted that in step S5, key point features are extracted from the paired initial dish image and end dish image. According to the key point features corresponding to the initial dish image and the key point features corresponding to the end dish image, the key point features that match the key point features corresponding to the end dish image in the initial dish image are obtained through the BF algorithm (brute force algorithm). A projective transformation matrix is performed with the matched key point features and the end dish image. If the calculation of the projective transformation matrix fails, it means that the change degree of the dish is 100%. If the calculation of the projective transformation matrix is successful, after the matched key point features are projected onto the end dish image, the points whose distance between the matched key point features and the original key point features in the end dish image is less than the threshold are taken as inliers, and the ratio of the number of inliers to the total number of the matched key point features is calculated as the change degree.
[0020] In this embodiment, the key point features are ORB key point features. The code for extracting ORB key point features from the paired initial dish image and end dish image is as follows: orb = cv2.ORB_create() #find the keypoints and descriptors with ORB kp1, des1 = orb.detectAndCompute(img1) kp2, des2 = orb.detectAndCompute(img2) Among them, img1 is the initial dish image, des1 is the ORB key point feature corresponding to the initial dish image, img2 is the end dish image, and des2 is the ORB key point feature corresponding to the end dish image.
[0021] Preferably, in the BF algorithm of step S5, the matching mechanism is to set a distance threshold, and the key point features in the initial dish image whose distance from the key point features of the end dish image is within the distance threshold are used as the matching key point features.
[0022] The code of the BF algorithm is as follows: bf = cv2.BFMatcher() matches = bf.match(des1, des2) des1 is the ORB key point features corresponding to the initial dish image, des2 is the ORB key point features corresponding to the end dish image; matches is the initial dish image with the matching key point features. In this embodiment, the distance threshold is 0.7.
[0023] In this embodiment, the inliers are obtained through the following code: H, mask = cv2.findHomography(pts1, pts2, cv2.RANSAC, 5.0) Among them, pts1 are the matching key point features in the initial dish image in the matches returned by the BF algorithm. pts2 is the end dish image; H is the Homography matrix; Mask is the inliers that match successfully; cv2.RANSAC means that the findHomography function automatically ignores the points that are too outlier; the real value 5.0 means that the difference before and after the transformation exceeds 5 pixels is recognized as an outlier point.
[0024] Optionally, in step S2, for the rotating disk image area in the first image, with the center of the rotating disk image area as the origin, the rotating disk image area is evenly divided into multiple fan-shaped sub-areas at a preset angle, and a unique position code is assigned to each fan-shaped sub-area; When serving dishes, a unique dish code is assigned to each dish. After the dish is placed in the fan-shaped sub-area, the dish code corresponding to the dish is associated with the position code corresponding to the fan-shaped sub-area, and the associated dish code and position code are marked on the initial dish image corresponding to the dish.
[0025] For the division of the image area of the rotating disk, it is first necessary to determine the coordinates of the center of the rotating disk as the origin point, which can be achieved through image recognition technology. For example, the geometric center of the disk can be found using a circular detection algorithm. Assuming the diameter of the rotating disk is 1 meter and the center point coordinates are a certain pixel point in the image, based on this, the disk is evenly divided into multiple sector sub-regions at a preset angle. If the preset angle is 30 degrees, the disk can be divided into 12 sector sub-regions, and each sub-region covers 1 / 12 of the area of the disk. A unique position code is assigned to each sector sub-region. For example, starting from the clockwise direction, the codes are P1 to P12, which facilitates subsequent positioning and management.
[0026] Specifically, in the link of assigning codes to dishes and associating position codes, a simple mapping mechanism can be designed. Each dish will be assigned a unique dish code after being made, such as D001, D002, etc. When the dish is placed in a certain sector sub-region, the system will record the corresponding relationship between the dish code and the position code. For example, if the dish D001 is placed at position P3, the system will associate D001 with P3. This associated information can be stored in the database for convenient subsequent query and management. At the same time, the associated information is marked on the initial dish image. For example, "P3 - D001" is marked in the corner of the dish image. The advantage of this approach is to improve the accuracy of dish management and avoid the inefficiency of manual search.
[0027] In this embodiment, the way to obtain the initial dish image of each dish when serving is to first use a camera to obtain the first image corresponding to the entire dining table, then obtain the image corresponding to the rotating disk image area from the first image, and then obtain the image of each sector sub-region from the rotating disk image area. The image of each sector sub-region represents the initial dish image of each dish. Similarly, the way to obtain the end dish image of each dish at the end of the meal is to first use a camera to obtain the second image corresponding to the entire dining table, then obtain the image corresponding to the rotating disk image area from the second image, and then obtain the image of each sector sub-region from the rotating disk image area. The image of each sector sub-region represents the end dish image of each dish.
[0028] Specifically, in step S4, the position code corresponding to the sector sub-region in the end dish image is obtained, and the position code and the associated dish code are marked on the end dish image; The initial dish image and the end dish image with the same position code and dish code are paired as the same dish.
[0029] At the end of the meal, after obtaining the end-dish image, the system will identify the position code of the fan-shaped sub-region again, such as P3, and mark the dish code D001 associated with the position code P3 on the end-dish image. In this way, the images with the same position code and dish code in the initial dish image and the end-dish image will be paired as the same dish.
[0030] For example, for the pairing process of the initial dish image and the end-dish image, it can be achieved by comparing the coding consistency. Suppose there are three dishes with codes D001, D002, and D003 respectively, and the corresponding position codes are P3, P5, and P7. In the end-dish image, the system identifies the fan-shaped sub-regions with position codes P3, P5, and P7, finds the associated dish codes, marks them respectively, and then pairs the images with the same codes in the initial dish image and the end-dish image as the same dish. The advantage of this pairing method is that it can clearly track the changes of each dish from serving to the end of the meal, which provides convenience for subsequent analysis.
[0031] It should be noted that in the step S1, a first color curtain is set in the rotating disk area of the dining table, an image of the dining table is collected, and the image area of the first color curtain is extracted to obtain the rotating disk image area.
[0032] Optionally, the step S1 includes: S11: Perform hsv conversion on the collected image of the dining table, convert the image of the dining table from an RGB image to an hsv image, and then convert the pixel points of the first color extracted from the hsv image into a binary image; S12: Extract connected regions in the binary image, remove the individual green pixel points caused by noise, and obtain the annular edge polygon corresponding to the shape of the rotating disk; S13: Perform dilation and erosion on the annular edge polygon to fill the inside of the annular edge polygon and obtain the rotating disk image area.
[0033] Specifically, for the processing of the first color curtain, first convert the extracted pixel points of the first color into a binary image, that is, mark the area of the first color as white and other areas as black. Suppose the proportion of the pixel points of the first color in the image is 30%, but 5% of the pixel points may be isolated points caused by noise interference. Through connected region extraction, these noise points can be effectively removed, and only the larger connected regions are retained to form an annular edge polygon similar to the rotating disk. Then, perform dilation and erosion operations on this polygon. The purpose is to fill the gaps inside the edge and form a complete disk area. The advantage of this method is that even if there are small-scale missing areas in the area of the first color in the original image, the complete shape can be restored through morphological operations, and finally an accurate calibration area of the rotating disk can be obtained.
[0034] Specifically, in the step S11, the first color is green. For the green curtain, the green pixel points extracted from the hsv image are converted into a binary image through the function cv2.inRange(hsv, minGreen, maxGreen) in opencv, where hsv is the hsv image corresponding to the dining table, minGreen = (50, 50, 50), and maxGreen = (70, 255, 255).
[0035] The performance of the rotating disc image area in the image is greatly affected by factors such as light and shadow. Based on this, after converting the image from the conventional format to a format more suitable for color separation, that is, the hsv image, the target pixel points can be screened by setting the color range. Green has a relatively concentrated numerical distribution in a specific color space. By limiting the upper and lower thresholds, the target area can be effectively distinguished from the background. The extraction of green pixel points can be achieved by setting the ranges of hue, saturation, and lightness. Assume that in an actual dining table scenario, the hue values of the green curtain usually concentrate between 50 and 70, and the saturation and lightness are respectively in the ranges of 50 to 255 and 50 to 255. By setting such ranges, the pixels that meet the conditions in the image can be marked as white, and other areas are marked as black, forming a clear binary result. The advantage of this method is that even if the light in the dining table environment changes, the green area can still be recognized relatively stably, providing a reliable basis for subsequent area division.
[0036] The formula for converting the image of the dining table from an RGB image to an hsv image is: ; After hsv conversion, the pixel points of the first color such as green and the second color such as blue can be extracted by setting the ranges of saturation and lightness. The technical effects brought by this conversion are significant. It reduces the influence of light changes on color recognition and improves the accuracy of subsequent processing.
[0037] Preferably, in the step S12, the connected regions are extracted through the function cv2.connectedComponents(image, connectivity, ltype) in opencv, where image is the binary image, connectivity is the connected domain, connectivity = 8, and ltype is the type of the function output, ltype = CV_32.
[0038] For the extraction of connected regions in binary images, the accurate division of the target region can be achieved by setting specific connectivity parameters. Suppose in a table scene, there are multiple small regions disconnected due to light interference in the binary image corresponding to the green curtain. At this time, by setting the connectivity parameter to 8, adjacent pixel points can be connected to form a complete region. The advantage of this method is that it can effectively integrate fragmented target regions and improve the accuracy of subsequent shape recognition.
[0039] It should be noted that in the step S12, the connected region of the annular edge polygon is extracted from the connected regions extracted by the function cv2.HoughCircles in opencv from the function cv2.connectedComponents(image, connectivity, ltype). After extracting the connected regions, it is necessary to further identify the regions with annular characteristics to correspond to the rotating disc in the table scene. In a possible implementation, the target region can be screened by detecting the annular edge polygon.
[0040] It should be noted that in the step S13, the annular edge polygon corresponding to the rotating disc is inflated by 200% through the function shapely.buffer(2.) to fill the interior of the annular edge polygon, and the inflated circular region is obtained; The inflated circular region is eroded by 200% through the function shapely.buffer(-2.) to reduce the area of the circular region to be consistent with the area of the actual rotating disc, and the rotating disc image region is obtained.
[0041] It should be noted that the inflation and erosion operations of the annular edge polygon can help to more accurately calibrate the target region. Especially when extracting the shape of the rotating disc against the background of the green curtain, this method is particularly important. By adjusting the area of the inflated circular region, it can better fit the contour of the actual disc.
[0042] Regarding the inflation process of the annular edge polygon, it can be understood as a way of boundary expansion. Its purpose is to fill the gaps inside the polygon to form a complete circular region. Suppose in a table image, there are some small breaks in the initially extracted annular edge polygon due to light or occlusion. Through the inflation operation, these broken parts can be connected to form a continuous region. Specifically, the inflation ratio can be set to 200%, which can ensure that even if there is a large gap in the edge, it can be effectively filled. The benefit of this operation is that it allows subsequent processing to be based on a more complete shape foundation and avoids misjudgment caused by incomplete edges.
[0043] The circular area obtained after completion of dilation often has a relatively large area. Therefore, it is necessary to reduce the area through erosion operation to be closer to the size of the actual rotating disc. It can be understood that the erosion operation is equivalent to shrinking the region boundary, aiming to remove the redundant extended part. By setting the erosion ratio to 200%, the boundary can be gradually retracted, and finally the area of the region is made consistent with the actual disc. The beneficial effect of this method is to improve the accuracy of the calibration region and ensure that subsequent analysis will not produce errors due to area deviation.
[0044] From another perspective, the combined use of dilation and erosion can also cope with the possible noise interference in the table image. Suppose there are some small impurities around the green curtain area in the image. Through the operation of dilation first and then erosion, while filling the edge break, the influence of these impurities can be minimized. The advantage of this combined front and back processing method is that it not only ensures the integrity of the shape but also improves the purity of the region calibration.
[0045] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principle and spirit of the present invention, various changes, modifications, substitutions, and variations made to these embodiments still fall within the protection scope of the present invention.
Claims
1. A method for evaluating the satisfaction of table dishes based on image recognition, characterized in that, Including the following steps: S1: Calibrate and train the images obtained by the camera, and calibrate the rotating disc image area in the images; S2: During the process of serving dishes, obtain the first image calibrated with the information of the rotating disc image area. In the first image, associate the positions corresponding to the rotating disc image area of the table for each dish being eaten currently, and obtain the initial dish images of each dish when serving; S3: At the end of the meal, obtain the second image calibrated with the information of the rotating disc image area, and obtain the final dish images of each dish in the second image; S4: Pair the initial dish images and the final dish images belonging to the same dish through the positions of the rotating disc image areas of the table; S5: Use the degree of change obtained by comparing the paired initial dish images and final dish images as the satisfaction of the dishes on the table.
2. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 1, wherein: In the step S5, extract key point features from the paired initial dish images and final dish images; According to the key point features corresponding to the initial dish images and the key point features corresponding to the final dish images, use the BF algorithm to obtain the key point features in the initial dish images that match the key point features corresponding to the final dish images; Perform a projective transformation matrix through the matched key point features and the final dish images. If the calculation of the projective transformation matrix fails, it means the degree of change of the dish is 100%; if the calculation of the projective transformation matrix is successful, after projecting the matched key point features onto the final dish images, take the points whose distance between the matched key point features and the original key point features in the final dish images is less than the threshold as inliers, and calculate the ratio of the number of inliers to the total number of the matched key point features as the degree of change.
3. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 2, characterized in that: In the BF algorithm of the step S5, the matching mechanism is to set a distance threshold, and take the key point features whose distance from the key point features of the final dish images in the initial dish images is within the distance threshold as the matched key point features.
4. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 1, wherein: In the step S2, for the rotating disc image area in the first image, with the center of the rotating disc image area as the origin, evenly divide the rotating disc image area into multiple sector sub-areas at a preset angle, and assign a unique position code to each sector sub-area; When serving dishes, assign a unique dish code to each dish. After placing the dish in the sector sub-area, associate the dish code corresponding to the dish with the position code corresponding to the sector sub-area, and mark the associated dish code and position code on the initial dish image corresponding to the dish.
5. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 4, wherein: In the step S4, obtain the position code corresponding to the sector sub-area in the image in the final dish image, and mark the position code and the associated dish code on the final dish image; Pair the initial dish images and the final dish images with the same position code and dish code as the same dish.
6. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 1, characterized in that: In the step S1, set a first color curtain in the rotating disc area of the table, collect the image of the table, extract the image area of the first color curtain, and obtain the rotating disc image area.
7. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 1, wherein: The step S1 includes: S11: Perform HSV conversion on the collected image of the dining table, convert the image of the dining table from an RGB image to an HSV image, and then convert the pixel points of the first color extracted from the HSV image into a binary image; S12: Extract connected regions in the binary image, remove the individual green pixel points generated by noise, and obtain the annular edge polygon corresponding to the shape of the rotating disc; S13: Dilate and erode the annular edge polygon to fill the interior of the annular edge polygon and obtain the image region of the rotating disc.
8. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 7, characterized in that: In the step S11, the first color is green. For the first color curtain, the green pixel points extracted from the HSV image are converted into a binary image through the function cv2.inRange(hsv, minGreen, maxGreen) in OpenCV, where hsv is the HSV image corresponding to the dining table, minGreen=(50, 50, 50), and maxGreen=(70, 255, 255).
9. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 7, characterized in that: In the step S12, the connected regions are extracted through the function cv2.connectedComponents(image, connectivity, ltype) in OpenCV, where image is the binary image, connectivity is the connected domain, connectivity = 8, and ltype is the type of the function output, ltype = CV_32.
10. The method for evaluating the satisfaction of table dishes based on image recognition according to claim 7, characterized in that: In the step S13, the annular edge polygon corresponding to the rotating disc is dilated by 200% through the function shapely.buffer(2.) to fill the interior of the annular edge polygon and obtain the dilated circular region; The dilated circular region is eroded by 200% through the function shapely.buffer(-2.) to reduce the area of the circular region to be consistent with the area of the actual rotating disc and obtain the image region of the rotating disc.