A method and system for evaluating glassware cleanliness scores for dishwasher performance testing

Abstract

The application discloses a glass cup washing score evaluation method and system for testing the performance of a dishwasher, comprising: obtaining a glass cup washing sample of a dishwasher to be tested, collecting images of the washing sample based on a preset washing score evaluation system, obtaining surface images of the washing sample, extracting a background axis region image, distinguishing two modes of coincident area subset matching and full sequence matching between adjacent images, discriminating and removing stain residual points of repeated residual areas, calculating physical areas and point numbers of independent residual areas, and outputting washing performance scores. The method uses stain fluorescence to overcome the interference of fully transparent materials and surface texture characteristics on imaging, improves the collection quality and image clarity of fluorescence signals, and removes repeated images, thereby improving the precision and objectivity of dishwasher washing performance detection.

Description

Technical Field

[0001] This invention relates to the field of environmental science, and in particular to a method and system for evaluating the cleaning score of a glass used in dishwasher performance testing. Background Technology

[0002] The performance evaluation of household electric dishwashers falls under the category of performance testing technology for household cleaning appliances. With the popularization of home appliances and the development of industry standardization, performance evaluation has become a key link in dishwasher product research and development, quality control, and national standard certification. According to the performance testing method for household electric dishwashers, the industry needs to conduct washing tests using standard glass cups contaminated with typical food contaminants such as milk. The cleaning effect of the glass cup is an important basis for measuring the cleaning ability of the dishwasher. Currently, manual visual observation is still the mainstream evaluation method. Although intelligent detection technologies such as machine vision and fluorescence imaging are gradually being applied in industrial cleaning testing, a mature testing solution adapted to transparent glass cups has not yet been formed. Preliminary explorations have only been carried out in some laboratory scenarios, which cannot meet the industry's actual application needs for efficient, objective, and quantitative testing.

[0003] The manual evaluation of dishwasher glass cleaning performance is heavily influenced by subjective judgment, resulting in poor consistency and low efficiency. Furthermore, it cannot output precise quantitative scores, hindering the scientific optimization of testing standards. Additionally, the transparent nature of the glass makes image acquisition prone to issues such as overlapping images of both sides of the glass and interference from background noise with milk residue signals. Existing visual inspection technologies cannot fully capture images of the entire glass surface, leading to an inability to accurately identify and remove duplicate residue areas caused by rotating the image acquisition process. Therefore, effectively acquiring full-surface images of the glass and overcoming the interference of the transparent material and surface texture features on imaging, in order to achieve intelligent and objective evaluation of dishwasher cleaning performance, has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for evaluating the cleaning score of a glass used in dishwasher performance testing.

[0005] To achieve the above objectives, the present invention is implemented according to the following technical solution: The first aspect of this invention provides a method for evaluating the washing cleanliness score of a glass used in dishwasher performance testing, comprising: Obtain a glass washing sample from the dishwasher to be tested; Based on a pre-set cleaning score evaluation system, images of the washed samples are acquired to obtain surface images of the washed samples; Extract the background axis region image from the side wall image of the glass, and obtain the step shooting angle corresponding to the background axis region image; Calculate the overlapping area between adjacent sidewall images based on the step shooting angle, background axis size and glass size, remove duplicate stain residue areas in the overlapping area, and obtain an optimized image; Based on the optimized image, a cleaning performance score is output according to the number of stain residue points and the residual area inside the glass.

[0006] Furthermore, the method for obtaining the surface image includes: Place the washing sample upside down on the O-ring of the washing score evaluation system, set the electric rotating worktable to jog mode, fix the mobile phone to the mobile phone holder, the camera sensor plane of the mobile phone is parallel to the central axis of the washing sample, the spectral response characteristics of the camera sensor are matched with the fluorescence band of the food contaminant, and the rotation angle of the jog mode is 90 degrees. According to the jog mode, the washing sample is rotated 360 degrees by an electric rotary table, and the side wall image of the washing sample is acquired by the camera sensor. The mobile phone is adjusted above the LED ultraviolet ring light source, and the plane of the camera sensor is parallel to the bottom of the washing sample to acquire the bottom image of the washing sample. The surface image is obtained based on the side wall image and the bottom image. The LED ultraviolet ring light source is set above the washing sample. The surface image includes the fluorescence information of food contaminants. When the mobile phone is acquiring the image, it is set to professional mode, with an ISO sensitivity of 1250, a shutter speed of 1 / 20, and a manual focus mode.

[0007] Further, the method for obtaining the background axis region image includes: Distortion correction is performed on the surface image, and the corrected surface image is converted to HSV space. A mask is generated based on the hue H channel to obtain the sidewall image of the washed sample after removing the scale area. The color range of the hue H channel is [0.023, 0.622]. The HSV space includes hue H, saturation S and lightness V. The surface image is in RGB format. Based on the sidewall image of the washed sample in the rejection scale area, the background axis pixel diameter is extracted, the background axis pixel diameter is rounded down to obtain the width of the background axis region, and the minimum vertical distance between the two arcs is taken as the height of the background axis region according to the background axis image. Using the coordinates of the upper left corner of the background axis region as the origin, and the coordinates corresponding to the point 50 pixels to the right of the origin as the starting point, the background axis region image is cropped according to the origin, starting point, width, and height to obtain the background axis region image corresponding to each surface image. The background axis region image is smaller than the projected area of ​​the background axis region.

[0008] Further, the method for obtaining the corrected surface image includes: Sub-pixel edge points of the left and right generatrices of the background axis cylinder and the left and right boundary points of the inner wall of the glass are extracted from the surface image. The surface image is vertically aligned, and the sampling index is obtained through row-by-row correspondence. The pixel coordinates of the edge point set and boundary point set are remapped using the Brownian-Conraddy distortion model to obtain the corrected abscissa. Based on the corrected abscissa, the center abscissa of the background axis and the center abscissa of the inner wall of the glass in each surface image are calculated, and an optimization objective function for the distortion parameter vector is constructed. The calculation formula of the optimization objective function is as follows: ; Among them, Let be the distortion parameter vector, and be , For the first and second order radial distortion coefficients, These are the first-order and second-order tangential distortion coefficients. The coordinates of the principal point are given, and the initial values ​​are the coordinates of the image center. The ideal busbar parameters for the background axis are... , The x-coordinate of the left generatrix of the background axis in the corrected image. The x-coordinate of the right generatrix of the background axis in the corrected image. Image sequence number The left and right vertical boundaries of the background axis in the image. For vertical sampling index, For the first Frame background axis middle Side Line edge point Corrected x-coordinate For the first The x-coordinate of the background axis center in the m-th line of frame. For the first Frame number The x-coordinate of the center of the inner wall of the glass. To correct the weighting coefficients; The objective function is minimized using the Levenberg-Marquardt algorithm to obtain the optimal distortion parameter vector. Based on the optimal distortion parameter vector, the surface image is remapped pixel by pixel and resampled bilinearly to obtain the corrected surface image.

[0009] Furthermore, the method for obtaining the optimized image includes: The width of the annular slit is calculated based on the concentric circle fit between the diameter of the upper cylinder of the background axis and the inner diameter of the washing sample. The imageable arc surface through the side wall of the washing sample in a single imaging is determined according to the width of the annular slit, and the field of view of a single imaging is obtained. The rotation angle of the electric rotary stage is used as the stepping shooting angle. The corresponding overlapping field of view is determined according to the difference between the imaging field of view and the stepping shooting angle. The overlapping field of view between adjacent side wall images is used as a specific overlapping region, and the width of the specific overlapping region is obtained. The overlapping region image is cropped from the edge of the adjacent background axis region image according to the width of the specific overlapping region, and the overlapping region image pair is obtained. Set the non-zero pixels of the overlapping region image pair to 255 to obtain a binarized image pair. Extract the 8 extreme points of the connected region in the binarized image. Subtract the adjacent point pairs between the extreme points in turn to obtain a two-dimensional difference vector. Use the two-dimensional difference vector as a contour feature. Perform cyclic shift on the contour feature to generate a sequence variant. Calculate the cosine similarity between the two-dimensional difference vector and the sequence variant. Take the maximum value as the similarity score. The number of sequence variants is the same as the length of the two-dimensional difference vector sequence. Based on the contour features of the binarized image, the template contour with the highest similarity score is extracted from another binarized image to obtain contour pairs between the binarized image pairs. Candidate contour pairs with a vertical coordinate deviation of less than 200 pixels are selected. If the similarity score between the candidate contour pairs is greater than or equal to a preset similarity threshold, the vertical coordinates of the vertices of the two connected regions are obtained. If the absolute value of the difference between the vertical coordinates of the vertices is greater than or equal to 50 pixels, it is determined to be non-repeating. If the absolute value of the difference between the vertical coordinates of the vertices is less than 50 pixels, it is determined to be a repeated stain residue area. The pixels of the repeated stain residue area are set to 0 to complete the removal and obtain the optimized image.

[0010] Furthermore, the method for obtaining the contour pairs includes: If the cosine similarity of one of the first-to-last connected vectors in the two-dimensional difference vector is greater than or equal to the preset similarity threshold, then the two connected regions in the binarized image pair overlap with each other. The two connected regions are used as subset similarity matching regions. The number of sequence variants is set to 1, and the maximum similarity value is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair. The similarity score is calculated using the following formula: ; in Contour feature vector and eigenvectors Similarity score between them Image of overlapping regions The first connected region in the middle A two-dimensional difference vector, Image of overlapping regions The first one aligned with it in the shifted variant A two-dimensional difference vector, The length of the two-dimensional difference vector sequence; If the two connected regions in the binarized image pair overlap completely, the number of sequence variants is set to 8, the two connected regions are used as the whole sequence similarity matching regions, and the maximum similarity is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair.

[0011] Furthermore, the method for obtaining the cleaning performance score includes: The bottom image of the washed sample is converted to the HSV color space, and a mask is created based on a threshold for binarization to extract the physical area of ​​the stain residue at the bottom of the washed sample. The threshold has H channel values ​​of [0.448, 0.586], S channel values ​​of [0.392, 0.944], and V channel values ​​of [0.904, 1.000]. Based on the optimized image, the connected regions are taken as the side wall stain residue area. The number of connected regions is taken as the number of residue points. According to the pinhole camera model, the side wall stain residue area is converted from pixel area to physical area, and the physical area unit is square millimeter. The cleaning performance score is output by the cleaning score evaluation system based on the physical area of ​​stain residue on the bottom and the area of ​​stain residue on the side walls.

[0012] A second aspect of the present invention provides a glass cleaning score evaluation system for dishwasher performance testing, comprising: 1. Glass cup; 2. Mobile phone; 3. Mobile phone holder; 4. LED ultraviolet ring light source; 5. Background axis; 6. Central fixed axis; 7. Rotating tray; 8. O-ring; 9. Electric rotating worktable; and 10. Optical breadboard. The glass cup 1 is placed on the upper end of the rotating tray 7. An O-ring 8 is provided between the glass cup 1 and the rotating tray 7. There is no relative sliding between the glass cup 1 and the rotating tray 7. The background shaft 5 is located inside the glass cup 1. The inner wall of the glass cup 1 does not contact the surface of the background shaft 5. One end of the central fixing shaft 6 is fixed to the bottom optical breadboard 10, and the other end is fixedly connected to the background shaft 5. An LED ultraviolet ring light source 4 is provided above the rotating tray 7. The bottom of the rotating tray 7 is fixedly connected to the electric rotating worktable 9. An optical breadboard 10 is installed at the bottom of the electric rotating worktable 9. A positioning hole is provided on the optical breadboard 10. The positioning hole is aligned with the central axis of the background axis 5. The background axis 5 coincides with the geometric center of the glass cup 1 cavity in the vertical direction.

[0013] Furthermore, including: The electric rotary worktable 9 is a two-phase stepper motor. The background shaft 5 is a matte black opaque cylinder with a surface roughness of less than or equal to 3.2 micrometers. The background shaft 5 has a light absorption rate of greater than or equal to 95% and a specular reflectivity of less than or equal to 5% for the LED ultraviolet ring light source 4. The background shaft 5 includes upper and lower cylinders. The upper cylinder has a diameter of 47 mm and is adapted to a glass cup with an inner diameter of 54 mm. The lower cylinder has a diameter of 18 mm and is smaller than the diameter of the central hole of the electric rotary worktable 9. The outer diameter of the glass cup 1 is less than or equal to 65.5 mm. The outer diameter of the rotating tray 7 is 100 mm and the diameter of the central hole is 30 mm. The outer diameter and the diameter of the central hole are the same as the outer diameter and the diameter of the central hole of the electric rotary table 9. The surface of the rotating tray 7 is provided with an annular groove, and a through hole is machined in the annular groove. The rotating tray 7 is fixedly connected to the mounting surface of the electric rotary table 9 through the through hole and the thread. The diameter of the annular groove is 68 mm. The outer diameter of the O-ring 8 is the same as the diameter of the annular groove and the wire diameter is 4 mm. The O-ring 8 is fixedly installed in the annular groove.

[0014] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: This invention uses an ultraviolet ring light source to excite food stain residue to produce fluorescence. Combined with a background axis placed inside the glass, it avoids the overlap of images on both sides of the glass caused by the transmission characteristics of the glass material. It overcomes the interference of the fully transparent material and surface texture features on imaging, improves the acquisition quality of fluorescence signals and image clarity. The overlapping areas of adjacent side wall images are measured by the cyclic shift cosine similarity of the difference vector of 8 extreme points. It combines two modes, region set matching and full sequence matching, to identify and remove repeated images of the same stain in different frames during the rotation shooting process, avoiding the repeated counting of residual points and areas. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of a method for evaluating the cleaning score of a glass used in dishwasher performance testing, as described in an embodiment of the present invention.

[0016] Figure 2 This is a cross-sectional structural schematic diagram and a physical image of a glass cleaning score evaluation system for dishwasher performance testing according to an embodiment of the present invention; Figure 3 This is a sidewall binarized image of a glass cleaning score evaluation method for dishwasher performance testing according to an embodiment of the present invention. Figure 4 This is an example diagram of the whole sequence similarity matching region for a method of evaluating the cleaning score of a glass used in dishwasher performance testing according to an embodiment of the present invention; Figure 5 This is an example diagram of subset similarity matching regions for a glass washing score evaluation method for dishwasher performance testing in an embodiment of the present invention; Figure 6 This is a comparison image before and after removing repeated milk residue areas in a glass cleaning score evaluation method for dishwasher performance testing according to an embodiment of the present invention. Figure 7 The image shows the bottom of a glass and its binarized image, which are part of a method for evaluating the cleaning score of a glass used in dishwasher performance testing according to an embodiment of the present invention. In the diagram: 1-Glass cup; 2-Mobile phone; 3-Mobile phone holder; 4-LED ultraviolet ring light source; 5-Background axis; 6-Center fixed axis; 7-Rotating tray; 8-O-ring; 9-Electric rotating worktable; 10-Optical breadboard. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 As shown, this invention provides a method for evaluating the washing cleanliness score of a glass used in dishwasher performance testing, comprising: Obtain a glass washing sample from the dishwasher to be tested; In the actual evaluation, milk was selected as a food contaminant according to the GB / T 20290-2024 standard "Performance Test Method for Household Electric Dishwashers". A transparent glass cup with an inner diameter of 54mm was selected and milk was used to contaminate the glass cup to prepare a contaminated sample. The contaminated sample was then placed in the dishwasher to be tested for washing to obtain a glass cup washing sample. Based on a pre-set cleaning score evaluation system, images of the washed samples are acquired to obtain surface images of the washed samples; In the actual evaluation, the washed sample was placed upside down on the O-ring of the washing score evaluation system, with the rim facing downwards. A mobile phone was fixed to a phone holder, and the phone's camera was set to professional mode, ISO 1250, shutter speed 1 / 20, and focus mode manual. An LED ultraviolet ring light source was positioned above the washed sample with a center wavelength of 365nm. The LED ultraviolet ring light source was turned on to illuminate the residual milk stains on the washed sample, exciting the stains to fluoresce. The phone's camera sensor plane was parallel to the central axis of the washed sample, and the spectral response characteristics of the camera sensor covered a spectral response range of 400nm– The 500nm band spectral response characteristics match the fluorescence band of milk stimulated emission. The electric rotary stage is set to jog mode, rotating 90° each time it is triggered. According to the jog mode, the washing sample is rotated 360° by the electric rotary stage. Every 90° rotation, an image of the glass side wall with stains in fluorescent state is acquired. Four images of the glass side wall are obtained for each glass. Then, the mobile phone is adjusted above the LED ultraviolet ring light source so that the camera sensor plane is parallel to the bottom of the washing sample, and one image of the bottom of the glass with stains in fluorescent state is acquired. Five images represent the full surface image of each glass. Extract the background axis region image from the side wall image of the glass, and obtain the step shooting angle corresponding to the background axis region image; In the actual evaluation, Canny edge detection was performed on four surface images of the glass sidewalls, with a low threshold of 50 and a high threshold of 150. Edge points along the rim and background axis were extracted. Typical parameters for each image were obtained through least-squares ellipse fitting. The major axis of the rim projection ellipse was 1063px, the minor axis was 1058px, and the center was (1512, 2016). The major axis of the background projection ellipse was 925px, the minor axis was 920px, and the center was (1505, 2020). The initial eccentricity offset was calculated to be 8.1px, representing the deviation caused by adjusting the fixed position of the phone when shooting from the sidewall and bottom. Constraints were extracted based on the washing samples. The pixel coordinates of the edge point set and boundary point set are remapped using the Brown-Conrady radial distortion model according to constraints to obtain the corrected abscissas. The constraints include coaxial constraints (corrected cup rim center = background axis center) and scale constraints (corrected pixel diameter ratio = physical diameter ratio 54mm / 47mm). Based on the corrected abscissas, the center abscissas of the background axis and the center abscissa of the inner wall of the glass are calculated in each surface image. An optimization objective function for the distortion parameter vector is constructed, and the objective function is minimized using the LM algorithm to obtain the distortion parameter vector. The first-order radial distortion coefficient is -1.1 × 10⁻⁶. -7 px -2Based on the physical diameter of the background axis (47mm) and the corrected pixel diameter (925px), the pixel-physical scaling factor was determined to be 0.0508mm / px. Four surface images of the glass sidewalls were corrected using the distortion parameter vector and an eccentricity compensation of 8.1px, resulting in a corrected surface image of 3024px (width) × 4032px (height), where the distortion correction error was ≤0.5px and the concentricity residual error was ≤0.5px. The corrected surface image in RGB format was converted to HSV space, and a mask was generated based on the hue H channel to obtain the sidewall image of the washed sample excluding the scale area. The hue H channel color range was [0.023, 0.622]. The background axis was extracted based on the sidewall image of the washed sample excluding the scale area. The background axis pixel diameter was 925, rounded down to 900, which was used as the width w of the background axis region to avoid differences between images. There are slight displacement errors. Since the part of the background axis inside the glass is a cylinder, its two ends are projected onto the sensor plane as arcs. The minimum vertical distance between the two arcs in the background axis image is taken as the height h of the background axis region. The measured pixel length of h is 2234. Let the origin coordinates of the upper left corner of the background axis be (x0, y0). Taking (x0+50, y0) in each image as the starting point, the background axis region image is cropped according to the starting point, width w and height h. The area of ​​the cropped background axis region image is smaller than the actual projection area of ​​the background axis region to avoid imaging errors caused by slight deviations between the outer contour of the background axis and the vertical direction. The four background axis region images obtained for each glass are recorded as B1, B2, B3 and B4. The step shooting angles corresponding to the four background axis region images are 0°, 90°, 180° and 270° respectively. Calculate the overlapping area between adjacent sidewall images based on the step shooting angle, background axis size and glass size, remove duplicate stain residue areas in the overlapping area, and obtain an optimized image; In the actual evaluation, based on the concentric circle fit relationship of the stepping shooting angle of 90°, the diameter of the upper cylinder of the background axis of 47mm, and the inner diameter of the washed sample of 54mm, the width of the annular slit was calculated to be 54 / 2-47 / 2=3.5mm, thus determining the imageable arc surface through the side wall of the washed sample in a single imaging. The single-image field of view is set to 120° to facilitate setting the acquisition angle to an integer value when the glass is rotated. The corresponding overlapping field of view is determined based on the difference between the imaging field of view and the step-by-step shooting angle. The overlapping field of view between adjacent sidewall images is taken as a specific overlapping region. The pixel width of the overlapping region is approximately 206.74. To avoid missing the detection of milk residue areas, the width w of the specific overlapping region between the two images is determined. g It is 220 pixels, based on the width w of the specific overlapping area. g From adjacent background axis region image B i and B jThe overlapping region image is cropped from the edges to obtain the overlapping region image for I. {c,i} and I {c,j} ; In the actual evaluation, the non-zero pixels of the overlapping region image pairs are set to 255 to obtain binarized image pairs. Eight extreme points of the connected regions are extracted from each binarized image pair. These eight extreme points are [top left, top right, top right, bottom right, bottom right, bottom left, bottom left, top left], defined as [1, 2, 3, 4, 5, 6, 7, 8]. Adjacent point pairs between these eight extreme points are subtracted sequentially to form eight two-dimensional difference vectors, namely the difference vectors of the 1→2, 2→3, ..., 7→8, 8→1 extreme points, denoted as I. {c,i} The difference vectors of the eight extreme points are V 12,i V 23,i V 34,i V 45,i V 56,i V 67,i V 78,i V 81,i I {c,j} The difference vectors of the eight extreme points are V 12,j V 23,j V 34,j V 45,j V 56,j V 67,j V 78,j V 81,j Using the two-dimensional difference vector as contour features, the contour features are cyclically shifted to generate sequence variants. The feature vectors of the two contours are defined as V. i and V j To assess their similarity, V is not directly calculated. i and V j Instead of simply evaluating the cosine similarity of V, we systematically assess V. i and V j The similarity between cyclic shift variants, specifically for a sequence V of length L. i L different sequence variants can be generated through cyclic shift operations. Where s = 0, 1, ..., L-1, then, the eigenvectors V are calculated respectively. i With each shift variant Cosine similarity between them If the cosine similarity of one of the first-to-last vectors [3→4] or [8→7] among the difference vectors of the eight extrema of two connected regions satisfies the preset similarity threshold of 0.85, then the two are subset similarity matching regions. In this case, L is set to 1, and the maximum value of the similarity calculated under the shift position is taken as the subset similarity score of the two contours, where the subset similarity matching region is represented by I. {c,i}Let I be a connected region. {c,j} Part of a connected region, and located in I {c,i} The left boundary in, or I {c,j} Let I be a connected region. {c,i} Part of a connected region, and located in I {c,j} If the right boundary is in the middle, then both cases involve a "part-whole" matching relationship and require the region to be close to the boundary. In this case, subset similarity matching is performed. If I {c,i} A certain connected region in I {c,j} If a connected region completely overlaps, the two connected regions are considered as the whole-sequence similarity matching region. In this case, L is set to 8. The maximum value of the similarity calculated at all shift positions is taken as the whole-sequence similarity score of the two contours, thus obtaining the contour pair. Traversal I {c,i} and I {c,j} In the binarized image pair, connected regions are used to calculate the similarity score between contour pairs. For the binarized I... {c,i} Each contour in I {c,j} The system searches for the template contour with the highest similarity score and determines whether the similarity score of the contour pair exceeds a preset similarity threshold of 0.85. If it exceeds 0.85, the contour pair is considered a successful match, and the I... {c,i} The outline region in I is marked as {c,j} For matching regions, if a subset similarity matching region has already been determined, the corresponding region will not undergo full sequence similarity matching. Successfully matched contour regions and their similarity scores will be recorded and output. Contour pairs with a y-direction deviation of less than 200 pixels will be selected as candidate matching objects. For candidate matching objects, if |y2-y1|>50 pixels, they will be determined as non-repeating stain residue regions and this pair will be removed. If |y2-y1|≤50 pixels, they will be determined as repeating stain residue regions and the repeating stain residue regions S will be removed. {c,j} All pixels are set to 0 to complete the culling and obtain the optimized image, where the top left vertex of contour 1 is (x1, y1) and the top left vertex of contour 2 is (x2, y2). Based on the optimized image, a cleaning performance score is output according to the number of stain residue points and the residual area inside the glass.

[0019] In the actual evaluation, step S4 is repeated for the sidewall images B1, B2, B3, and B4 of the washed sample. The resulting connected regions are the stain residue areas on the sidewalls, and the number of connected regions is the number of residue points on the sidewalls. Based on the pinhole camera model, the stain residue areas on the sidewalls are converted from pixel area to physical area (mm²). 2The bottom RGB image of the washed sample was converted to the HSV color space, and a mask was created based on the thresholds H channel [0.448, 0.586], S channel [0.392, 0.944], and V channel [0.904, 1.000]. Binarization was performed using the mask to extract the bottom stain residue area. The number of connected regions in the binarized bottom image represents the number of bottom residue points. Using a pinhole camera model, the milk residue area at the bottom was converted from pixel area to physical area. The washing performance score was calculated based on the sum of the physical areas of the bottom stain residue and the sidewall stain residue. The total physical area of ​​the stains in the washed sample was 2.9 mm². 2 The number of residual points is 3. According to the washing performance evaluation method specified in GB / T 20290-2024, based on the grading logic of glass cleaning performance evaluation, the qualitative grading standard of manual visual inspection is transformed into a quantitative scoring rule based on the physical area of ​​residual milk. The area and number of points of milk residue are mapped to a washing performance score of 0-5 points. The physical area of ​​bottom stain residue and the physical area of ​​side wall stain residue, as well as the number of residual points, are substituted into the scoring calculation of the evaluation method to output the washing performance score. Therefore, the washing performance score of the dishwasher under test is 4 points.

[0020] In this embodiment, the method for obtaining the washed sample includes: Food contaminants and glass cups were selected based on dishwasher performance test parameters. The glass cups were contaminated with food contaminants, and the contaminated samples were washed by the dishwasher under test to obtain a washed sample of the glass cups. The food contaminants included milk, tea, minced meat, eggs, oatmeal, spinach, and vegetable butter. The glass cups were made of a translucent material.

[0021] In this embodiment, the method for obtaining the surface image includes: Place the washing sample upside down on the O-ring of the washing score evaluation system, set the electric rotating worktable to jog mode, fix the mobile phone to the mobile phone holder, the camera sensor plane of the mobile phone is parallel to the central axis of the washing sample, the spectral response characteristics of the camera sensor are matched with the fluorescence band of the food contaminant, and the rotation angle of the jog mode is 90 degrees. According to the jog mode, the washing sample is rotated 360 degrees by an electric rotary table, and the side wall image of the washing sample is acquired by the camera sensor. The mobile phone is adjusted above the LED ultraviolet ring light source, and the plane of the camera sensor is parallel to the bottom of the washing sample to acquire the bottom image of the washing sample. The surface image is obtained based on the side wall image and the bottom image. The LED ultraviolet ring light source is set above the washing sample. The surface image includes the fluorescence information of food contaminants. When the mobile phone is acquiring the image, it is set to professional mode, with an ISO sensitivity of 1250, a shutter speed of 1 / 20, and a manual focus mode.

[0022] In this embodiment, the method for obtaining the background axis region image includes: Distortion correction is performed on the surface image, and the corrected surface image is converted to HSV space. A mask is generated based on the hue H channel to obtain the sidewall image of the washed sample after removing the scale area. The color range of the hue H channel is [0.023, 0.622]. The HSV space includes hue H, saturation S and lightness V. The surface image is in RGB format. Based on the sidewall image of the washed sample in the rejection scale area, the background axis pixel diameter is extracted, the background axis pixel diameter is rounded down to obtain the width of the background axis region, and the minimum vertical distance between the two arcs is taken as the height of the background axis region according to the background axis image. Using the coordinates of the upper left corner of the background axis region as the origin, and the coordinates corresponding to the point 50 pixels to the right of the origin as the starting point, the background axis region image is cropped according to the origin, starting point, width, and height to obtain the background axis region image corresponding to each surface image. The background axis region image is smaller than the projected area of ​​the background axis region.

[0023] In this embodiment, the method for obtaining the corrected surface image includes: Sub-pixel edge points of the left and right generatrices of the background axis cylinder and the left and right boundary points of the inner wall of the glass are extracted from the surface image. The surface image is vertically aligned, and the sampling index is obtained through row-by-row correspondence. The pixel coordinates of the edge point set and boundary point set are remapped using the Brownian-Conraddy distortion model to obtain the corrected abscissa. Based on the corrected abscissa, the center abscissa of the background axis and the center abscissa of the inner wall of the glass in each surface image are calculated, and an optimization objective function for the distortion parameter vector is constructed. The calculation formula of the optimization objective function is as follows: ; Among them, Let be the distortion parameter vector, and be , For the first and second order radial distortion coefficients, These are the first-order and second-order tangential distortion coefficients. The coordinates of the principal point are given, and the initial values ​​are the coordinates of the image center. The ideal busbar parameters for the background axis are... , The x-coordinate of the left generatrix of the background axis in the corrected image. The x-coordinate of the right generatrix of the background axis in the corrected image. Image sequence number The left and right vertical boundaries of the background axis in the image. For vertical sampling index, For the first Frame background axis middle Side Line edge point Corrected x-coordinate For the first The x-coordinate of the background axis center in the m-th line of frame. For the first Frame number The x-coordinate of the center of the inner wall of the glass. To correct the weighting coefficients; The objective function is minimized using the Levenberg-Marquardt algorithm to obtain the optimal distortion parameter vector. Based on the optimal distortion parameter vector, the surface image is remapped pixel by pixel and resampled bilinearly to obtain the corrected surface image.

[0024] In this embodiment, the method for obtaining the optimized image includes: The width of the annular slit is calculated based on the concentric circle fit between the diameter of the upper cylinder of the background axis and the inner diameter of the washing sample. The imageable arc surface through the side wall of the washing sample in a single imaging is determined according to the width of the annular slit, and the field of view of a single imaging is obtained. The rotation angle of the electric rotary stage is used as the stepping shooting angle. The corresponding overlapping field of view is determined according to the difference between the imaging field of view and the stepping shooting angle. The overlapping field of view between adjacent side wall images is used as a specific overlapping region, and the width of the specific overlapping region is obtained. The overlapping region image is cropped from the edge of the adjacent background axis region image according to the width of the specific overlapping region, and the overlapping region image pair is obtained. Set the non-zero pixels of the overlapping region image pair to 255 to obtain a binarized image pair. Extract the 8 extreme points of the connected region in the binarized image. Subtract the adjacent point pairs between the extreme points in turn to obtain a two-dimensional difference vector. Use the two-dimensional difference vector as a contour feature. Perform cyclic shift on the contour feature to generate a sequence variant. Calculate the cosine similarity between the two-dimensional difference vector and the sequence variant. Take the maximum value as the similarity score. The number of sequence variants is the same as the length of the two-dimensional difference vector sequence. Based on the contour features of the binarized image, the template contour with the highest similarity score is extracted from another binarized image to obtain contour pairs between the binarized image pairs. Candidate contour pairs with a vertical coordinate deviation of less than 200 pixels are selected. If the similarity score between the candidate contour pairs is greater than or equal to a preset similarity threshold, the vertical coordinates of the vertices of the two connected regions are obtained. If the absolute value of the difference between the vertical coordinates of the vertices is greater than or equal to 50 pixels, it is determined to be non-repeating. If the absolute value of the difference between the vertical coordinates of the vertices is less than 50 pixels, it is determined to be a repeated stain residue area. The pixels of the repeated stain residue area are set to 0 to complete the removal and obtain the optimized image.

[0025] In this embodiment, the method for obtaining the contour pair includes: If the cosine similarity of one of the first-to-last connected vectors in the two-dimensional difference vector is greater than or equal to the preset similarity threshold, then the two connected regions in the binarized image pair overlap with each other. The two connected regions are used as subset similarity matching regions. The number of sequence variants is set to 1, and the maximum similarity value is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair. The similarity score is calculated using the following formula: ; in Contour feature vector and eigenvectors Similarity score between Image of overlapping regions The first connected region in the middle A two-dimensional difference vector, Image of overlapping regions The first one aligned with it in the shifted variant A two-dimensional difference vector, The length of the two-dimensional difference vector sequence; If the two connected regions in the binarized image pair overlap completely, the number of sequence variants is set to 8, the two connected regions are used as the whole sequence similarity matching regions, and the maximum similarity is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair.

[0026] In this embodiment, the method for obtaining the cleaning performance score includes: The bottom image of the washed sample is converted to the HSV color space, and a mask is created based on a threshold for binarization to extract the physical area of ​​the stain residue at the bottom of the washed sample. The threshold has H channel values ​​of [0.448, 0.586], S channel values ​​of [0.392, 0.944], and V channel values ​​of [0.904, 1.000]. Based on the optimized image, the connected regions are taken as the side wall stain residue area. The number of connected regions is taken as the number of residue points. According to the pinhole camera model, the side wall stain residue area is converted from pixel area to physical area, and the physical area unit is square millimeter. The cleaning performance score is output by the cleaning score evaluation system based on the physical area of ​​stain residue on the bottom and the area of ​​stain residue on the side walls.

[0027] A second aspect of the present invention also provides a glass cleaning score evaluation system for dishwasher performance testing, comprising: 1. Glass cup; 2. Mobile phone; 3. Mobile phone holder; 4. LED ultraviolet ring light source; 5. Background axis; 6. Central fixed axis; 7. Rotating tray; 8. O-ring; 9. Electric rotating worktable; and 10. Optical breadboard. The glass cup 1 is placed on the upper end of the rotating tray 7. An O-ring 8 is provided between the glass cup 1 and the rotating tray 7. There is no relative sliding between the glass cup 1 and the rotating tray 7. The background shaft 5 is located inside the glass cup 1. The inner wall of the glass cup 1 does not contact the surface of the background shaft 5. One end of the central fixing shaft 6 is fixed to the bottom optical breadboard 10, and the other end is fixedly connected to the background shaft 5. An LED ultraviolet ring light source 4 is provided above the rotating tray 7. The bottom of the rotating tray 7 is fixedly connected to the electric rotating worktable 9. An optical breadboard 10 is installed at the bottom of the electric rotating worktable 9. A positioning hole is provided on the optical breadboard 10. The positioning hole is aligned with the central axis of the background axis 5. The background axis 5 coincides with the geometric center of the glass cup 1 cavity in the vertical direction.

[0028] Further, it is characterized by comprising: The electric rotary worktable 9 is a two-phase stepper motor. The background shaft 5 is a matte black opaque cylinder with a surface roughness of less than or equal to 3.2 micrometers. The background shaft 5 has a light absorption rate of greater than or equal to 95% and a specular reflectivity of less than or equal to 5% for the LED ultraviolet ring light source 4. The background shaft 5 includes upper and lower cylinders. The upper cylinder has a diameter of 47 mm and is adapted to a glass cup with an inner diameter of 54 mm. The lower cylinder has a diameter of 18 mm and is smaller than the diameter of the central hole of the electric rotary worktable 9. The outer diameter of the glass cup 1 is less than or equal to 65.5 mm. The outer diameter of the rotating tray 7 is 100 mm and the diameter of the central hole is 30 mm. The outer diameter and the diameter of the central hole are the same as the outer diameter and the diameter of the central hole of the electric rotary table 9. The surface of the rotating tray 7 is provided with an annular groove, and a through hole is machined in the annular groove. The rotating tray 7 is fixedly connected to the mounting surface of the electric rotary table 9 through the through hole and the thread. The diameter of the annular groove is 68 mm. The outer diameter of the O-ring 8 is the same as the diameter of the annular groove and the wire diameter is 4 mm. The O-ring 8 is fixedly installed in the annular groove.

[0029] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A method for evaluating the washing cleanliness score of a glass used in dishwasher performance testing, characterized in that, Includes the following steps: Obtain a glass washing sample from the dishwasher to be tested; Based on a pre-set cleaning score evaluation system, images of the washed samples are acquired to obtain surface images of the washed samples; Extract the background axis region image from the side wall image of the glass, and obtain the step shooting angle corresponding to the background axis region image; Calculate the overlapping area between adjacent sidewall images based on the step shooting angle, background axis size and glass size, remove duplicate stain residue areas in the overlapping area, and obtain an optimized image; Based on the optimized image, a cleaning performance score is output according to the number of stain residue points and the residual area inside the glass.

2. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 1, characterized in that, A method for obtaining the surface image includes: Place the washing sample upside down on the O-ring of the washing score evaluation system, set the electric rotating worktable to jog mode, fix the mobile phone to the mobile phone holder, the camera sensor plane of the mobile phone is parallel to the central axis of the washing sample, the spectral response characteristics of the camera sensor are matched with the fluorescence band of the food contaminant, and the rotation angle of the jog mode is 90 degrees. According to the jog mode, the washing sample is rotated 360 degrees by an electric rotary table, and the side wall image of the washing sample is acquired by the camera sensor. The mobile phone is adjusted above the LED ultraviolet ring light source, and the plane of the camera sensor is parallel to the bottom of the washing sample to acquire the bottom image of the washing sample. The surface image is obtained based on the side wall image and the bottom image. The LED ultraviolet ring light source is set above the washing sample. The surface image includes the fluorescence information of food contaminants. When the mobile phone is acquiring the image, it is set to professional mode, with an ISO sensitivity of 1250, a shutter speed of 1 / 20, and a manual focus mode.

3. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 1, characterized in that, A method for obtaining the background axis region image includes: Distortion correction is performed on the surface image, and the corrected surface image is converted to HSV space. A mask is generated based on the hue H channel to obtain the sidewall image of the washed sample after removing the scale area. The color range of the hue H channel is [0.023, 0.622]. The HSV space includes hue H, saturation S and lightness V. The surface image is in RGB format. Based on the sidewall image of the washed sample in the rejection scale area, the background axis pixel diameter is extracted, the background axis pixel diameter is rounded down to obtain the width of the background axis region, and the minimum vertical distance between the two arcs is taken as the height of the background axis region according to the background axis image. Using the coordinates of the upper left corner of the background axis region as the origin, and the coordinates corresponding to the point 50 pixels to the right of the origin as the starting point, the background axis region image is cropped according to the origin, starting point, width, and height to obtain the background axis region image corresponding to each surface image. The background axis region image is smaller than the projected area of ​​the background axis region.

4. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 3, characterized in that, The method for obtaining the corrected surface image includes: Sub-pixel edge points of the left and right generatrices of the background axis cylinder and the left and right boundary points of the inner wall of the glass are extracted from the surface image. The surface image is vertically aligned, and the sampling index is obtained through row-by-row correspondence. The pixel coordinates of the edge point set and boundary point set are remapped using the Brownian-Conraddy distortion model to obtain the corrected abscissa. Based on the corrected abscissa, the center abscissa of the background axis and the center abscissa of the inner wall of the glass in each surface image are calculated, and an optimization objective function for the distortion parameter vector is constructed. The calculation formula of the optimization objective function is as follows: ; Among them, Let be the distortion parameter vector, and be , For the first and second order radial distortion coefficients, These are the first-order and second-order tangential distortion coefficients. The coordinates of the principal point are given, and the initial values ​​are the coordinates of the image center. The ideal busbar parameters for the background axis are... , The x-coordinate of the left generatrix of the background axis in the corrected image. The x-coordinate of the right generatrix of the background axis in the corrected image. Image sequence number The left and right vertical boundaries of the background axis in the image. For vertical sampling index, For the first Frame background axis middle Side Line edge point Corrected x-coordinate For the first The x-coordinate of the background axis center in the m-th line of frame. For the first Frame number The x-coordinate of the center of the inner wall of the glass. To correct the weighting coefficients; The objective function is minimized using the Levenberg-Marquardt algorithm to obtain the optimal distortion parameter vector. Based on the optimal distortion parameter vector, the surface image is remapped pixel by pixel and resampled bilinearly to obtain the corrected surface image.

5. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 1, characterized in that, The method for obtaining the optimized image includes: The width of the annular slit is calculated based on the concentric circle fit between the diameter of the upper cylinder of the background axis and the inner diameter of the washing sample. The imageable arc surface through the side wall of the washing sample in a single imaging is determined according to the width of the annular slit, and the field of view of a single imaging is obtained. The rotation angle of the electric rotary stage is used as the stepping shooting angle. The corresponding overlapping field of view is determined according to the difference between the imaging field of view and the stepping shooting angle. The overlapping field of view between adjacent side wall images is used as a specific overlapping region, and the width of the specific overlapping region is obtained. The overlapping region image is cropped from the edge of the adjacent background axis region image according to the width of the specific overlapping region, and the overlapping region image pair is obtained. Set the non-zero pixels of the overlapping region image pair to 255 to obtain a binarized image pair. Extract the 8 extreme points of the connected region in the binarized image. Subtract the adjacent point pairs between the extreme points in turn to obtain a two-dimensional difference vector. Use the two-dimensional difference vector as a contour feature. Perform cyclic shift on the contour feature to generate a sequence variant. Calculate the cosine similarity between the two-dimensional difference vector and the sequence variant. Take the maximum value as the similarity score. The number of sequence variants is the same as the length of the two-dimensional difference vector sequence. Based on the contour features of the binarized image, the template contour with the highest similarity score is extracted from another binarized image to obtain contour pairs between the binarized image pairs. Candidate contour pairs with a vertical coordinate deviation of less than 200 pixels are selected. If the similarity score between the candidate contour pairs is greater than or equal to a preset similarity threshold, the vertical coordinates of the vertices of the two connected regions are obtained. If the absolute value of the difference between the vertical coordinates of the vertices is greater than or equal to 50 pixels, it is determined to be non-repeating. If the absolute value of the difference between the vertical coordinates of the vertices is less than 50 pixels, it is determined to be a repeated stain residue area. The pixels of the repeated stain residue area are set to 0 to complete the removal and obtain the optimized image.

6. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 5, characterized in that, The method for obtaining the contour pairs includes: If the cosine similarity of one of the first-to-last connected vectors in the two-dimensional difference vector is greater than or equal to the preset similarity threshold, then the two connected regions in the binarized image pair overlap with each other. The two connected regions are used as subset similarity matching regions. The number of sequence variants is set to 1, and the maximum similarity value is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair. The similarity score is calculated using the following formula: ； in Contour feature vector and eigenvectors Similarity score between Image of overlapping regions The first connected region in the middle A two-dimensional difference vector, Image of overlapping regions The first one aligned with it in the shift variant A two-dimensional difference vector, The length of the two-dimensional difference vector sequence; If the two connected regions in the binarized image pair overlap completely, the number of sequence variants is set to 8, the two connected regions are used as the whole sequence similarity matching regions, and the maximum similarity is calculated based on the shift position as the similarity score of the contour features of the two connected regions to obtain the contour pair.

7. The method for evaluating the washing score of a glass used in dishwasher performance testing according to claim 1, characterized in that, The method for obtaining the cleaning performance score includes: The bottom image of the washed sample is converted to the HSV color space, and a mask is created based on a threshold for binarization to extract the physical area of ​​the stain residue at the bottom of the washed sample. The threshold has H channel values ​​of [0.448, 0.586], S channel values ​​of [0.392, 0.944], and V channel values ​​of [0.904, 1.000]. Based on the optimized image, the connected regions are taken as the side wall stain residue area. The number of connected regions is taken as the number of residue points. According to the pinhole camera model, the side wall stain residue area is converted from pixel area to physical area, and the physical area unit is square millimeter. The cleaning performance score is output by the cleaning score evaluation system based on the physical area of ​​stain residue on the bottom and the area of ​​stain residue on the side walls.

8. A glass cleaning score evaluation system for dishwasher performance testing, used to perform the glass cleaning score evaluation method for dishwasher performance testing according to any one of claims 1 to 7, characterized in that, The system includes:

1. Glass cup; 2. Mobile phone; 3. Mobile phone holder; 4. LED ultraviolet ring light source; 5. Background axis; 6. Central fixed axis; 7. Rotating tray; 8. O-ring; 9. Electric rotating worktable; and 10. Optical breadboard. The glass cup 1 is placed on the upper end of the rotating tray 7. An O-ring 8 is provided between the glass cup 1 and the rotating tray 7. There is no relative sliding between the glass cup 1 and the rotating tray 7. The background shaft 5 is located inside the glass cup 1. The inner wall of the glass cup 1 does not contact the surface of the background shaft 5. One end of the central fixing shaft 6 is fixed to the bottom optical breadboard 10, and the other end is fixedly connected to the background shaft 5. An LED ultraviolet ring light source 4 is provided above the rotating tray 7. The bottom of the rotating tray 7 is fixedly connected to the electric rotating worktable 9. An optical breadboard 10 is installed at the bottom of the electric rotating worktable 9. A positioning hole is provided on the optical breadboard 10. The positioning hole is aligned with the central axis of the background axis 5. The background axis 5 coincides with the geometric center of the glass cup 1 cavity in the vertical direction.

9. A glass cleaning score evaluation system for dishwasher performance testing according to claim 8, characterized in that, include: The electric rotary worktable 9 is a two-phase stepper motor. The background shaft 5 is a matte black opaque cylinder with a surface roughness of less than or equal to 3.2 micrometers. The background shaft 5 has a light absorption rate of greater than or equal to 95% and a specular reflectivity of less than or equal to 5% for the LED ultraviolet ring light source 4. The background shaft 5 includes upper and lower cylinders. The upper cylinder has a diameter of 47 mm and is adapted to a glass cup with an inner diameter of 54 mm. The lower cylinder has a diameter of 18 mm and is smaller than the diameter of the central hole of the electric rotary worktable 9. The outer diameter of the glass cup 1 is less than or equal to 65.5 mm. The outer diameter of the rotating tray 7 is 100 mm and the diameter of the central hole is 30 mm. The outer diameter and the diameter of the central hole are the same as the outer diameter and the diameter of the central hole of the electric rotary table 9. The surface of the rotating tray 7 is provided with an annular groove, and a through hole is machined in the annular groove. The rotating tray 7 is fixedly connected to the mounting surface of the electric rotary table 9 through the through hole and the thread. The diameter of the annular groove is 68 mm. The outer diameter of the O-ring 8 is the same as the diameter of the annular groove and the wire diameter is 4 mm. The O-ring 8 is fixedly installed in the annular groove.

Images