A method for online detection of a malfunction of a fluidic spray head of a surface cleaning device

Abstract

The present application relates to the technical field of intelligent detection and automation, and discloses a kind of surface cleaning device's jet flow nozzle fault on-line detection method.Just determine the pixel cleaning area of each nozzle in image;Collect historical images before and after strip cleaning and construct difference map based on color difference, establish surface cleaning quality detection model in combination with the highest and lowest original score, and set qualified standard;Real-time acquisition of images before and after cleaning input model score, identify unqualified images according to qualified standard and mark unqualified area;Semantic segmentation and morphological processing are carried out on the marked area, the contour boundary is extracted, and the faulty nozzle is located in combination with the pixel cleaning area.The method solves the problem of difficult positioning of nozzle fault in existing high-pressure abrasive water jet surface cleaning device.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection and automation technology, and in particular to an image recognition and fault location nozzle for metal strip surface images. Background Technology

[0002] During the surface cleaning of metal strips using a high-pressure abrasive water jet surface cleaning device, the conveying equipment carries the strip through the cleaning station. The cleaning station is equipped with a set of nozzles, which consists of several nozzles positioned at different angles and staggered. This set of water jet nozzles sprays a mixture of high-pressure water and abrasive, driven by a high-pressure pump, to impact the surface of the strip, removing impurities and achieving a clean surface that meets the requirements of subsequent processing.

[0003] Due to factors such as corrosion, scale, and abrasive wear, the jet nozzle is one of the main points of failure in water jet surface cleaning devices. The main symptoms include poor abrasive passage, abrasive blockage, and reduced jet velocity. These can cause fluctuations or a decline in surface cleaning quality, requiring timely repair and replacement of the nozzle. Currently, there is no good online detection method for nozzle failure, and it mainly relies on manual periodic inspection. This requires workers to shut down the cleaning device and enter the cleaning station to check each nozzle one by one, which is time-consuming, labor-intensive, and has a low degree of automation. Summary of the Invention

[0004] This invention provides an online detection method for jet nozzle faults in surface cleaning devices, in order to solve the problem of difficulty in locating nozzle faults in existing high-pressure abrasive water jet surface cleaning devices.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] This invention provides an online detection method for jet nozzle malfunctions in a surface cleaning device, comprising the following steps:

[0007] Step 1: Determine the geometric cleaning area of ​​all nozzles of the surface cleaning device to be inspected, and obtain the pixel cleaning area of ​​all nozzles based on the geometric cleaning area of ​​all nozzles;

[0008] Step 2: Obtain multiple pairs of historical images before and after the strip enters the surface cleaning device to be inspected, obtain multiple original scores of the historical images, construct a surface cleaning quality detection model for scoring real-time images based on the highest and lowest original scores among the multiple original scores, and determine the pass standard based on the surface cleaning quality detection model combined with multiple pairs of historical images.

[0009] Step 3: Acquire multiple pairs of real-time images of the strip before and after cleaning, and input the multiple pairs of real-time images into the surface cleaning quality inspection model for scoring. Obtain multiple scores for the real-time images. Based on the multiple scores of the real-time images and the pass / fail criteria, identify unqualified images in the multiple pairs of real-time images, and mark the unqualified areas in the unqualified images with warnings.

[0010] Step 4: Perform semantic segmentation and morphological processing on the non-conforming areas marked by warnings in the non-conforming images to obtain the contour boundaries of the non-conforming areas. Based on the contour boundaries of the non-conforming areas and the pixel cleaning areas, locate the faulty nozzles.

[0011] Through the above design, a surface cleaning quality inspection model is constructed, enabling quantitative evaluation of cleaning quality and automatic identification of non-conforming areas. The scoring system, based on historical images, can dynamically adapt to changes in cleaning performance under different working conditions, ensuring the objectivity and adaptability of the inspection standards. Semantic segmentation and morphological processing can accurately locate the contour boundaries of non-conforming areas, and combined with preset pixel cleaning areas, rapid location of faulty nozzles is achieved. This method enables online monitoring and fault warning of cleaning quality, reducing reliance on human experience and improving the automation level of inspection.

[0012] Furthermore, it also includes step 5: obtaining the abrasive concentration and determining the fault type of the faulty nozzle based on the abrasive concentration;

[0013] The fault type of the nozzle determined based on abrasive concentration includes: if the abrasive concentration is greater than the normal value, the fault type is blockage; if the abrasive concentration is within the normal value, the fault type is low output speed.

[0014] The above design not only allows for the location of faulty nozzles but also enables further identification of the fault type (clogging or low output speed), providing maintenance personnel with clear diagnostic criteria, shortening troubleshooting time, and improving maintenance efficiency.

[0015] Furthermore, the pixel cleaning area of ​​a single nozzle is obtained through the following steps: acquiring the strip after the single nozzle of the surface cleaning device to be inspected performs the strip cleaning operation, determining the image of the geometric cleaning area formed on the surface of the strip, binding the actual width of the geometric cleaning area with the resolution width of the image in a proportional relationship, and obtaining the pixel cleaning area of ​​a single nozzle on the image.

[0016] Furthermore, in step 2, obtaining the original score for any pair of historical images includes the following steps:

[0017] Step 201: Denoise the pair of historical images to obtain a corresponding pair of denoised images;

[0018] Step 202: Convert the pair of denoised images from RGB color space images to LAB color space images, and calculate the LAB value for each pixel in the pair of denoised images;

[0019] Step 203: Based on the LAB value of each pixel in the pair of denoised images, calculate the color difference of each pixel before and after the denoised images are cleaned using the Euclidean distance calculation formula, and construct a difference map based on the color difference.

[0020] Step 204: Determine the effective clean area based on the difference map, calculate the pixel mean of the pixel values ​​in the effective clean area, and obtain the original score of the historical image based on the ratio of the pixel mean to the area of ​​the effective clean area.

[0021] Furthermore, the construction of the surface cleaning quality detection model includes the following steps: constructing a surface cleaning quality detection model based on the highest and lowest original scores among multiple original scores, combined with the pixel mean of the effective cleaning area in a pair of real-time images;

[0022] This can be expressed by the following formula:

[0023] ;

[0024] in, This represents the score given by the surface cleaning quality inspection model for a pair of real-time images; This represents the average pixel value of the effectively cleaned area in a pair of real-time images; Indicates the highest original score; This indicates the lowest original score.

[0025] Through the above design, through differences Figure 2 Value-based and effective cleaning area extraction, combined with pixel mean and area ratio for scoring, so that the scoring results can reflect both cleaning intensity and cleaning uniformity, and comprehensively evaluate the cleaning quality.

[0026] Furthermore, in step 204, determining the effective cleaning area based on the difference map includes: comparing the pixel value of each pixel in the difference map with a preset low threshold and performing binarization processing; defining pixels less than or equal to the low threshold as not having been effectively cleaned, defining pixels greater than the low threshold as having been effectively cleaned, and forming an effective cleaning area from the pixels that have been effectively cleaned.

[0027] Furthermore, in step 4, the acquisition of the contour boundary of the unqualified region includes the following steps: acquiring a historical difference map and confirming a lower limit value based on the historical difference map; comparing and assigning values ​​based on the lower limit value and the pixel values ​​of each pixel in the difference map corresponding to the unqualified image; acquiring the difference map after comparison and assignment; then using opening and closing operations to denoise the difference map after comparison and assignment to obtain an unqualified mask map; identifying connected regions based on the unqualified mask map; and extracting the pixel contours of the connected regions as the contour boundary of the unqualified region.

[0028] The step of comparing and assigning values ​​based on the lower limit value and the pixel values ​​of each pixel in the difference map corresponding to the unqualified image includes: if the pixel value of a pixel in the difference map corresponding to the unqualified image is less than the lower limit value, then the value is assigned as 1; if the pixel value of a pixel in the difference map corresponding to the unqualified image is greater than or equal to the lower limit value, then the value is assigned as 0.

[0029] Furthermore, the step of obtaining historical difference maps and determining the lower limit value based on historical difference maps includes: taking all difference maps in step 2 as historical difference maps, and statistically analyzing the pixel values ​​of pixels in all historical difference maps to determine the lowest pixel value as the lower limit value.

[0030] Through the above design, based on the statistical lower limit value of the historical difference map and combined with morphological opening and closing operations, noise can be effectively removed, the contour can be smoothed and the broken areas can be connected, and the contour of the unqualified area with practical significance can be extracted, providing clear and coherent boundary information for subsequent nozzle positioning.

[0031] Furthermore, the step of using opening and closing operations to denoise the difference map after comparison and assignment to obtain the unqualified mask map includes: using opening operations to perform erosion and then dilation on the difference map after comparison and assignment, and then using closing operations to perform dilation and then erosion on the difference map after comparison and assignment to obtain the unqualified mask map.

[0032] Furthermore, in step 4, locating the faulty nozzle based on the contour boundary of the defective area and the pixel cleaning area includes: determining the area of ​​the defective area to be cleaned based on the contour boundary of the defective area; if the area of ​​the defective area to be cleaned is greater than a preset area threshold, calculating the intersection area of ​​the contour boundary and the pixel cleaning areas of all nozzles, sorting the intersection areas from largest to smallest, and locating the faulty nozzle based on the sorting.

[0033] Beneficial effects:

[0034] This invention provides an online fault detection method for a surface cleaning device's jet nozzle. By constructing a surface cleaning quality detection model, it achieves real-time quantitative evaluation and fluctuation monitoring of the cleaning effect, facilitating dynamic tracking of cleaning quality and timely identification of abnormalities. This method enables online fault detection and intelligent early warning for water jet nozzles, effectively improving the automation level and status awareness of equipment operation. Simultaneously, it significantly reduces downtime caused by manual troubleshooting, contributing to improved production efficiency and capacity utilization, and lower maintenance costs. Attached Figure Description

[0035] Figure 1 This is a flowchart of an online detection method for jet nozzle failure in a surface cleaning device according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the structure of the water jet surface cleaning device according to an embodiment of the present invention;

[0037] exist Figure 2 The components are: 1. Front-end cleaning camera at the inlet; 2. Rear-end cleaning camera at the outlet; 3. Surface cleaning quality inspection model; 4. Conveyor roller; 5. Strip to be cleaned; 6. Cleaning station; 7. Sand extraction pipe; 8. Abrasive concentration detection device; 9. Nozzle. Detailed Implementation

[0038] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "connected" or "linked" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up," "down," "left," "right," etc., are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship also changes accordingly.

[0040] Please see Figure 1 This application provides an online detection method for jet nozzle failure in a surface cleaning device, comprising the following steps:

[0041] Step 1: Determine the geometric cleaning area of ​​all nozzles of the surface cleaning device to be inspected, and obtain the pixel cleaning area of ​​all nozzles based on the geometric cleaning area of ​​all nozzles;

[0042] The pixel cleaning area of ​​a single nozzle is obtained through the following steps: acquiring the strip after the single nozzle of the surface cleaning device to be inspected performs the strip cleaning work, determining the image of the geometric cleaning area formed on the surface of the strip, binding the actual width of the geometric cleaning area with the resolution width of the image in a proportional relationship, and obtaining the pixel cleaning area of ​​a single nozzle on the image.

[0043] By repeating the above steps, the pixel cleaning area of ​​all nozzles of the surface cleaning device to be tested can be determined.

[0044] Step 2: Obtain multiple pairs of historical images before and after the strip enters the surface cleaning device to be inspected, obtain multiple original scores of the historical images, construct a surface cleaning quality detection model for scoring real-time images based on the highest and lowest original scores among the multiple original scores, and determine the pass standard based on the surface cleaning quality detection model combined with multiple pairs of historical images.

[0045] The process of obtaining the original score for any pair of historical images includes the following steps:

[0046] Step 201: Denoise the pair of historical images to obtain a corresponding pair of denoised images;

[0047] The denoising here uses Gaussian filtering or median filtering to denoise the pair of historical images.

[0048] Step 202: Convert the pair of denoised images from RGB color space images to LAB color space images, and calculate the LAB value for each pixel in the pair of denoised images;

[0049] Step 203: Based on the LAB value of each pixel in the pair of denoised images, calculate the color difference of each pixel before and after the denoised images are cleaned using the Euclidean distance calculation formula, and construct a difference map based on the color difference.

[0050] This is represented by the following formula:

[0051] ;

[0052] in, This represents a color difference diagram; Represents the coordinates of a pixel; This indicates the LAB value of the pixel in the LAB color space in a historical image before it entered the surface cleaning device;

[0053] This indicates the LAB value of the pixel in the LAB color space in a historical image after the surface cleaning device has been used.

[0054] Step 204: Determine the effective clean area based on the difference map, calculate the pixel mean of the pixel values ​​in the effective clean area, and obtain the original score of the historical image based on the ratio of the pixel mean to the area of ​​the effective clean area.

[0055] For the effective cleaning area, the pixel value of each pixel in the difference map is compared with the preset low threshold and binarized. Pixels less than or equal to the low threshold are defined as not effectively cleaned, and pixels greater than the low threshold are defined as effectively cleaned. The pixels that have been effectively cleaned are then grouped into an effective cleaning area.

[0056] In this embodiment, the low threshold is set to 5. If a pixel value in the difference map is greater than 5, then the pixel is defined as having been effectively cleaned. If a pixel value is less than or equal to 5, then the pixel is defined as not having been effectively cleaned.

[0057] Repeat steps 201 to 204 above for multiple pairs of historical images to obtain multiple original scores corresponding to the multiple pairs of historical images. Determine the highest and lowest original scores among the multiple original scores, and then construct a surface cleaning quality detection model based on the highest and lowest original scores and the pixel mean of the effective clean area in a pair of real-time images, expressed by the following formula:

[0058] ;

[0059] in, This represents the score given by the surface cleaning quality inspection model for a pair of real-time images; This represents the average pixel value of the effectively cleaned area in a pair of real-time images; Indicates the highest original score; This indicates the lowest original score.

[0060] The surface cleaning quality detection model constructed here scores a pair of input real-time images from 0 to 100, with a higher score indicating better cleanliness.

[0061] Based on this score, and using human experience, the critical image group between unsatisfactory and satisfactory cleanup and its corresponding score are found in multiple pairs of historical images. This score threshold is then used as the standard for acceptance.

[0062] Step 3: Acquire multiple pairs of real-time images of the strip before and after cleaning, and input the multiple pairs of real-time images into the surface cleaning quality inspection model for scoring. Obtain multiple scores for the real-time images. Based on the multiple scores of the real-time images and the pass / fail criteria, identify unqualified images in the multiple pairs of real-time images, and mark the unqualified areas in the unqualified images with warnings.

[0063] Specifically, if the score of a pair of real-time images is greater than the pass standard, then the pair of real-time images is defined as pass images, which means that the nozzle of the surface cleaning device to be tested is not malfunctioning at this time. If the score of a pair of real-time images is less than or equal to the pass standard, then the pair of real-time images is defined as fail images, and the fail areas in the fail images are marked with warnings.

[0064] Step 4: Perform semantic segmentation and morphological processing on the non-conforming areas marked by warnings in the non-conforming images to obtain the contour boundaries of the non-conforming areas. Based on the contour boundaries of the non-conforming areas and the pixel cleaning area, locate the faulty nozzle.

[0065] Specifically, all the difference maps in step 2 are used as historical difference maps, and the pixel values ​​of the pixels in all historical difference maps are statistically analyzed to determine the lowest pixel value as the lower limit. If the pixel value of the pixel in the difference map corresponding to the unqualified image is less than the lower limit, it is assigned a value of 1; if the pixel value of the pixel in the difference map corresponding to the unqualified image is greater than or equal to the lower limit, it is assigned a value of 0. The difference map after comparison and assignment is obtained.

[0066] First, an opening operation is used to perform erosion and then dilation on the difference map after comparison assignment to remove small isolated points, burrs and background noise, and smooth the outer contour. Then, a closing operation is used to perform dilation and then erosion on the difference map after comparison assignment to fill small holes or cracks, connect broken areas, and smooth the inner contour, resulting in an unqualified mask map containing multiple independent connected regions. The pixel contour of each connected region is extracted as the contour boundary of the unqualified region.

[0067] The area of ​​non-conforming areas is determined based on the contour boundary of the non-conforming area. If the area of ​​non-conforming areas is greater than the preset area threshold, the intersection area of ​​the cleanup contour boundary and the pixel cleanup area of ​​all nozzles is calculated, and the intersection areas are sorted from largest to smallest. The nozzle with the largest intersection area in the sorted location is identified as the faulty nozzle.

[0068] Step 5: Obtain the abrasive concentration and determine the fault type of the faulty nozzle based on the abrasive concentration.

[0069] Regarding the fault type, if the abrasive concentration is greater than the preset normal value, the nozzle fault is blockage; if the abrasive concentration is within the preset normal value, the nozzle fault is low output speed, i.e., poor flow.

[0070] Please see Figure 2 To verify the technical effect of this application, a front-end camera 1 for the inlet and a rear-end camera 2 for the outlet of a strip equipped with a water jet surface cleaning device were set up for synchronous shooting. The two cameras were fixed in position and equipped with the same focal length lens and the same resolution. The pixel resolution of the captured image remained constant. When shooting with a 2-megapixel camera (1920x1080), the ratio of the geometric dimension of the strip in the width direction to the pixel in the width direction was 1080 / 400 (pixels / mm).

[0071] The water jet surface cleaning device is equipped with 15 nozzles 9 at different angles. First, nozzle 1 9 is turned on individually for test cleaning to obtain the pixel cleaning range of nozzle 1 9. The cleaning range of nozzle 1 9 is 0-27mm, which corresponds to a cleaning range of 0-73 pixels on the image. Similarly, the cleaning ranges of nozzles 2-15 9 on the image can be measured.

[0072] Collect 200 sets of images of the strip before cleaning. and cleaned images Combining synchronous historical images This yields a set composed of image combinations. A surface cleaning quality inspection model 3 was constructed for scoring, and some data are shown in Table 1.

[0073] Table 1: Quantitative scoring of surface cleaning quality inspection model.

[0074]

[0075] go through Statistical and empirical judgment, The cleaning was substandard. The cleaning is satisfactory, and the scores are close. Therefore, the passing standard is set between 62.9 and 65.1. In this example, the midpoint 64 is taken as the judgment score for whether the cleaning is satisfactory. Note: The experience judgment here is a manual judgment based on experience.

[0076] A simulated fault experiment was conducted. No. 1 nozzle 9 was turned off, and nozzles 2-15 9 were turned on to perform a test wash on the strip. Multiple experiments could be conducted. The images before and after cleaning were combined and input into the surface cleaning quality detection model 3. The results such as scores and area ranking are shown in Table 2.

[0077] Table 2: Location and early warning of faulty nozzles.

[0078]

[0079] In Table 2, This represents the area (in pixels) of the largest outline of the uncleaned area in the image. The area represents the intersection (in pixels) of the maximum outline of the uncleaned area and the area of ​​the pixel-cleaned area corresponding to each nozzle. The abrasive concentration corresponds to the abrasive concentration detection result of the sand extraction pipe. It can be seen that when simulating the blockage of nozzle 1 9, the corresponding surface cleaning quality detection result is unqualified, and the overlap area between the unqualified area and the corresponding cleaning area of ​​nozzle 1 9 is also the largest. Under the detection result of abrasive concentration higher than the normal value, we can draw a warning conclusion that nozzle 1 9 is likely to be blocked.

[0080] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for detecting a failure of a jet nozzle of a surface cleaning apparatus online, characterized in that, Includes the following steps: Step 1: Determine the geometric cleaning area of ​​all nozzles of the surface cleaning device to be inspected, and obtain the pixel cleaning area of ​​all nozzles based on the geometric cleaning area of ​​all nozzles; Step 2: Obtain multiple pairs of historical images before and after the strip enters the surface cleaning device to be inspected, obtain multiple original scores of the historical images, construct a surface cleaning quality detection model for scoring real-time images based on the highest and lowest original scores among the multiple original scores, and determine the pass standard based on the surface cleaning quality detection model combined with multiple pairs of historical images. Obtaining the original score for any pair of historical images involves the following steps: Step 201: Denoise the pair of historical images to obtain a corresponding pair of denoised images; Step 202: Convert the pair of denoised images from RGB color space images to LAB color space images, and calculate the LAB value for each pixel in the pair of denoised images; Step 203: Based on the LAB value of each pixel in the pair of denoised images, calculate the color difference of each pixel before and after the denoised images are cleaned using the Euclidean distance calculation formula, and construct a difference map based on the color difference. Step 204: Determine the effective cleaning area based on the difference map, calculate the pixel mean of the pixel values ​​in the effective cleaning area, and obtain the original score of the pair of historical images based on the ratio of the pixel mean to the area of ​​the effective cleaning area; The construction of the surface cleaning quality detection model includes the following steps: constructing a surface cleaning quality detection model based on the highest and lowest original scores among multiple original scores, combined with the pixel mean of the effective cleaning area in a pair of real-time images; This can be expressed by the following formula: ； in, This represents the score given by the surface cleaning quality inspection model for a pair of real-time images; This represents the average pixel value of the effectively cleaned area in a pair of real-time images; Indicates the highest original score; Indicates the lowest original score; Step 3: Acquire multiple pairs of real-time images of the strip before and after cleaning, and input the multiple pairs of real-time images into the surface cleaning quality inspection model for scoring. Obtain multiple scores for the real-time images. Based on the multiple scores of the real-time images and the pass / fail criteria, identify unqualified images in the multiple pairs of real-time images, and mark the unqualified areas in the unqualified images with warnings. Step 4: Perform semantic segmentation and morphological processing on the non-conforming areas marked by warnings in the non-conforming images to obtain the contour boundaries of the non-conforming areas. Based on the contour boundaries of the non-conforming areas and the pixel cleaning areas, locate the faulty nozzles.

2. The online detection method for jet nozzle faults in the surface cleaning device according to claim 1, characterized in that, It also includes step 5: obtaining the abrasive concentration and determining the fault type of the faulty nozzle based on the abrasive concentration; The fault type of the nozzle determined based on abrasive concentration includes: if the abrasive concentration is greater than the normal value, the fault type is blockage; if the abrasive concentration is within the normal value, the fault type is low output speed.

3. The online detection method for jet nozzle failure in the surface cleaning device according to claim 1, characterized in that, The pixel cleaning area of ​​a single nozzle is obtained through the following steps: acquire the strip after the single nozzle of the surface cleaning device to be inspected performs the strip cleaning operation, determine the image of the geometric cleaning area formed on the surface of the strip, bind the actual width of the geometric cleaning area with the resolution width of the image in a proportional relationship, and obtain the pixel cleaning area of ​​the single nozzle on the image.

4. The online detection method for jet nozzle failure in the surface cleaning device according to claim 1, characterized in that, In step 204, determining the effective cleaning area based on the difference map includes: comparing the pixel value of each pixel in the difference map with a preset low threshold and performing binarization processing; defining pixels less than or equal to the low threshold as not having been effectively cleaned, defining pixels greater than the low threshold as having been effectively cleaned, and forming an effective cleaning area from the pixels that have been effectively cleaned.

5. The online detection method for jet nozzle failure in the surface cleaning device according to claim 1, characterized in that, In step 4, obtaining the contour boundary of the unqualified region includes the following steps: obtaining a historical difference map and confirming a lower limit value based on the historical difference map; comparing and assigning values ​​based on the lower limit value and the pixel values ​​of each pixel in the difference map corresponding to the unqualified image; obtaining the difference map after comparison and assignment; then using opening and closing operations to denoise the difference map after comparison and assignment to obtain an unqualified mask map; identifying connected regions based on the unqualified mask map; and extracting the pixel contours of the connected regions as the contour boundary of the unqualified region. The step of comparing and assigning values ​​based on the lower limit value and the pixel values ​​of each pixel in the difference map corresponding to the unqualified image includes: if the pixel value of a pixel in the difference map corresponding to the unqualified image is less than the lower limit value, then the value is assigned as 1; if the pixel value of a pixel in the difference map corresponding to the unqualified image is greater than or equal to the lower limit value, then the value is assigned as 0.

6. The online detection method for jet nozzle failure in the surface cleaning device according to claim 5, characterized in that, The step of obtaining historical difference maps and determining the lower limit value based on historical difference maps includes: taking all difference maps in step 2 as historical difference maps, and statistically analyzing the pixel values ​​of pixels in all historical difference maps to determine the lowest pixel value as the lower limit value.

7. The online detection method for jet nozzle faults in the surface cleaning device according to claim 6, characterized in that, The step of using opening and closing operations to denoise the difference map after comparison and assignment to obtain an unqualified mask map includes: using opening operations to perform erosion and then dilation on the difference map after comparison and assignment, and then using closing operations to perform dilation and then erosion on the difference map after comparison and assignment to obtain an unqualified mask map.

8. The online detection method for jet nozzle failure in the surface cleaning device according to claim 1, characterized in that, In step 4, locating the faulty nozzle based on the contour boundary of the defective area and the pixel cleaning area includes: determining the area of ​​the defective area to be cleaned based on the contour boundary of the defective area; if the area of ​​the defective area to be cleaned is greater than a preset area threshold, calculating the intersection area of ​​the contour boundary and the pixel cleaning area of ​​all nozzles, sorting the intersection areas from largest to smallest, and locating the faulty nozzle based on the sorting.

Images