Shower process quality control system, and shower process quality control method

Abstract

To provide a shower process quality control system which can optimize a discharge condition of a shower and can drastically reduce a burden of an administrator.SOLUTION: A shower process quality control system according to the present invention has a detection model constructing means, an area detection means, and a prediction model constructing means. The detection model constructing means makes a drop detection model for detecting a drop on a coated object from an image of the coated object captured upon passage of a shower process. The area detection means carries our an image processing with using a drop detection model to detect a shower dispersed area A3 and a non-dispersed area A4 in a shower dispersion target area A2. The prediction model constructing means accumulates data of dispersion results obtained from the shower dispersed area A3 and the non-dispersed area A4 to thereby construct a quality prediction model for use in prediction of a quality of the shower dispersion target area A2 after passage of a predetermined period of time.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present invention relates to a shower process quality control system and shower process quality control method for controlling the quality of a shower process by controlling the discharge conditions of a shower that cleans a coated object. [Background technology]

[0002] Conventionally, the degreasing and chemical conversion treatment process before electrodeposition coating includes a shower process for washing the automobile body. In this shower process, the automobile body is carried into a washing line and is washed by spraying a high-pressure washing liquid onto the carried-in automobile body (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2009-112920 Summary of the Invention [Problem to be solved by the invention]

[0004] In conventional cleaning lines, on-site managers visually inspect discharge conditions such as the angle of the shower nozzles and the pressure of the pump that supplies the cleaning liquid as car bodies pass by. In this case, the manager visually determines the discharge conditions and decides how the liquid is distributed over the car bodies. However, since the distribution of the spray is largely determined by the manager's experience, there is a possibility that the quality will vary depending on the manager. In addition, the inspection area is large and has multiple floors, and the manager must look inside the cleaning line to check the condition of the showers, which can place a heavy burden on the manager.

[0005] The present invention has been made in consideration of the above-mentioned problems, and its purpose is to provide a shower process quality control system and shower process quality control method that can optimize shower discharge conditions and significantly reduce the burden on managers. [Means for solving the problem]

[0006] In order to solve the above problem, the invention described in claim 1 is a shower process quality control system that manages the quality of a shower process by controlling the spray conditions of a shower used to clean a coated object, and is characterized by comprising: a detection model construction means that constructs a droplet detection model that detects droplets on the coated object from an image of the coated object taken as the object passes through the shower process; an area detection means that detects areas where shower spraying has occurred and areas where no shower spraying has occurred within a predetermined shower spraying target area by performing image processing using the droplet detection model in the image; and a prediction model construction means that constructs a quality prediction model that predicts the quality of the shower spraying target area after a predetermined period of time by accumulating spraying history data obtained from the shower spraying areas and the unsprayed areas.

[0007] In the invention described in claim 1, the droplet detection model detects droplets within a shower spray area set in an image, and the quality prediction model predicts the quality of the shower spray area after a predetermined period of time. This automates the sensory evaluation (specifically, evaluation of the degree to which the cleaning liquid is applied to the workpiece), which was previously performed visually by a supervisor. This allows the spray distribution to be determined without being influenced by the supervisor's experience, thereby reducing quality variation. Therefore, the shower discharge conditions can be optimized based on the quality of the shower spray area predicted by the quality prediction model. Furthermore, automating the sensory evaluation eliminates the need for the supervisor to peer into the cleaning line where the workpiece is being washed, significantly reducing the supervisor's burden.

[0008] The invention described in claim 2 is based on claim 1 and is characterized in that it includes a data management means for managing data on the shower spray target area for each of the multiple parts that make up the workpiece, and the area detection means calculates the area of ​​overlap between the shower spray target area managed by the data management means and a bounding box that surrounds the droplets detected using the droplet detection model, and detects the overlapping area as the shower spray history area.

[0009] In the invention described in claim 2, the area detection means calculates the area of ​​the overlap between the shower spray target area managed by the data management means and the bounding box surrounding the droplets, thereby quantifying the area of ​​the overlapping area, the shower spray history area. This allows the quality prediction model to quickly predict the quality of the shower spray target area including the shower spray history area, making it possible to quickly optimize the shower discharge conditions based on the predicted quality.

[0010] The invention described in claim 3 is characterized in that, in claim 2, an imaging device is provided which is installed in a cleaning line that cleans the object to be coated and which captures an image of the object to be coated to obtain the image.

[0011] When a manager visually checks the shower discharge conditions, it is difficult for the manager to fully view the inside of the cleaning line where the coated objects are cleaned. In this case, if the coated objects are found to be defective due to insufficient cleaning, there is a problem that defective products will be released onto the market. Therefore, in the invention described in claim 3, an imaging device installed in the cleaning line is used to capture images of the coated objects. This makes it possible to reliably detect insufficiently cleaned coated objects.

[0012] The invention described in claim 4 is characterized in that, in claim 3, the imaging device acquires a plurality of images by imaging the shower discharge range at predetermined time intervals as the object to be coated passes through the shower discharge range, and the area detection means detects the overlapping portion in each of the plurality of images.

[0013] In the invention described in claim 4, the imaging device captures images of the shower discharge range at predetermined intervals, thereby reducing the area of ​​overlap between the shower spray target area displayed in each image and the bounding box, thereby reducing the burden on the area detection means when calculating the area of ​​the overlap.

[0014] The invention described in claim 5 is based on claim 4, and its gist is that the area detection means sets one of the multiple images as a reference image, and combines the reference pre-image with the reference image so that the front end coordinate of the front end of the shower spray target area set in the reference image in the direction of travel of the workpiece matches the front end coordinate of the shower spray target area set in the reference pre-image acquired before the reference image, and combines the reference post-image with the reference image so that the rear end coordinate of the rear end of the shower spray target area set in the reference image in the direction of travel of the workpiece matches the rear end coordinate of the shower spray target area set in the reference image, and the rear end coordinate of the shower spray target area set in the reference post-image acquired after the reference image, thereby integrating the multiple images into a single image.

[0015] In the invention described in claim 5, the reference front image is combined with the reference image so that the front end coordinates of the shower spray target areas match, and the reference rear image is combined with the reference image so that the rear end coordinates of the shower spray target areas match, so that all overlapping areas are displayed in a single integrated image. In this case, the area detection means calculates the area of ​​all overlapping areas, thereby quantifying the area of ​​the previous shower spray area within the shower spray target area. This allows the quality prediction model to accurately predict the quality of the shower spray target area including the previous shower spray area, so that the shower spray conditions can be reliably optimized based on the predicted quality.

[0016] The invention described in claim 6 is based on claim 3, and the imaging device is a network camera capable of tracking the coated object being transported on the cleaning line, and the imaging points are registered for each part according to conveyor pulses and production instruction information to acquire the images.

[0017] In the invention described in claim 6, the imaging device registers imaging points for each part according to the conveyor pulse and production instruction information, so that the parts that make up the object to be coated can be accurately imaged based on the registered imaging points.

[0018] The invention of claim 7 is based on claim 1 and further comprises a display device that displays the shower sprayed areas and the unsprayed areas using colors and numerical values.

[0019] In the invention as recited in claim 7, by displaying shower sprayed areas and unsprayed areas in different colors or numerical values, it becomes easy to distinguish between shower sprayed areas and unsprayed areas.

[0020] The invention described in claim 8 is characterized in that, in claim 1, it is provided with an alarm means for notifying that cleaning is insufficient when the spraying history accumulated by the predictive model construction means falls below a specified value.

[0021] In the invention described in claim 8, the manager can quickly detect and take action if the object is not cleaned properly before it passes through the shower process, thereby preventing the outflow of defective objects.

[0022] Examples of notification means for notifying that cleaning is insufficient include light-emitting means such as a lamp that notifies that cleaning is insufficient by emitting light (lighting up, flashing, etc.), audio output means such as an alarm that notifies that cleaning is insufficient by sound (alert sound, etc.), and display means such as a liquid crystal display device that notifies that cleaning is insufficient by an alert image displaying letters, symbols, pictures, etc.

[0023] The invention described in claim 9 is based on claim 1, and its gist is that the optimal discharge conditions are derived based on the quality prediction model for each shower spray target area, and the shower is controlled based on the derived discharge conditions.

[0024] In the invention as recited in claim 9, the shower discharge conditions can be automatically switched for each shower spray target area.

[0025] The invention described in claim 10 is a method for managing the quality of a shower process by controlling the spray conditions of a shower used to clean a coated object, and is characterized by including: a detection model construction step for constructing a droplet detection model to detect droplets on the coated object from an image of the coated object taken as the object passes through the shower process; an area detection step for detecting areas where shower spraying has occurred and areas where no shower spraying has occurred within a predetermined shower spraying target area in the image by performing image processing using the droplet detection model; and a prediction model construction step for constructing a quality prediction model to predict the quality of the shower spraying target area after a predetermined period of time by accumulating spraying history data obtained from the shower spraying areas and the unsprayed areas.

[0026] In the invention described in claim 10, the droplet detection model constructed in the detection model construction step detects droplets within the shower spray target area set in the image, and the quality prediction model constructed in the prediction model construction step predicts the quality of the shower spray target area after a predetermined period of time. This automates the sensory evaluation (specifically, evaluation of the degree to which the cleaning liquid is applied to the workpiece), which was previously performed visually by a supervisor. This allows the spray distribution to be determined without being influenced by the supervisor's experience, thereby reducing quality variation. Therefore, the shower discharge conditions can be optimized based on the quality of the shower spray target area predicted by the quality prediction model. Furthermore, automating the sensory evaluation eliminates the need for the supervisor to peer into the cleaning line where the workpiece is being washed, significantly reducing the supervisor's burden. [Effects of the Invention]

[0027] As described above in detail, according to the inventions set forth in claims 1 to 10, shower discharge conditions can be optimized and the burden on the manager can be significantly reduced. [Brief explanation of the drawings]

[0028] [Figure 1] FIG. 2 is a schematic cross-sectional view showing a shower process quality control system according to the present embodiment. [Figure 2] FIG. [Figure 3] FIG. [Figure 4] FIG. 10 is a front view showing an image in which a shower spray target area is set. [Figure 5] (a) is a front view showing the first image, (b) is a front view showing the tenth image, and (c) is a front view showing the twentieth image. [Figure 6] (a) is a front view of the first image with a bounding box displayed, (b) is a front view of the tenth image with a bounding box displayed, and (c) is a front view of the twentieth image with a bounding box displayed. [Figure 7] (a) is a front view of the first image showing areas where shower spraying has been performed and areas where it has not been performed, (b) is a front view of the tenth image showing areas where shower spraying has been performed and areas where it has not been performed, and (c) is a front view of the 20th image showing areas where shower spraying has been performed and areas where it has not been performed. [Figure 8] (a) is a front view showing the pre-reference image (first image), (b) is a front view showing the reference image (tenth image), and (c) is a front view showing the post-reference image (twentieth image). [Figure 9] FIG. 10 is a front view showing an image obtained by combining a reference image with a pre-reference image and a post-reference image. [Figure 10] FIG. 10 is a front view showing an image obtained by combining all pre-reference images and all post-reference images with a reference image. [Figure 11] FIG. 10 is a front view showing an image in which shower sprayed areas and unsprayed areas are displayed in color and numerical values. DETAILED DESCRIPTION OF THE INVENTION

[0029] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] As shown in Figure 1, the shower process quality control system 10 of this embodiment is a system that controls the quality of the shower process by controlling the discharge conditions of a shower that cleans an automobile body W1 (subject to be coated) before painting. The shower process quality control system 10 is equipped with a degreasing tank 11 that stores a degreasing agent P1 and a chemical conversion tank (not shown) that stores a chemical conversion agent. Note that the chemical conversion tank has a configuration similar to that of the degreasing tank 11, so in this embodiment, only the degreasing tank 11 will be described, and a description of the chemical conversion tank will be omitted.

[0031] The degreasing tank 11 has a ceiling 12, a floor 13, and a pair of side walls 14. The degreasing tank 11 also has an inlet 15 for carrying the automobile body W1 into the degreasing tank 11, and an outlet 16 for carrying the automobile body W1 out of the degreasing tank 11. Inside the degreasing tank 11, the automobile body W1 is transported while immersed in the degreaser P1.

[0032] As shown in FIGS. 1 and 2, the shower process quality control system 10 includes a cleaning line 20 for cleaning the automobile body W1 downstream (on the left side in FIG. 1) of the degreasing tank 11 and the chemical conversion tank. The cleaning line 20 includes a chamber 21 to which the automobile body W1 removed from the degreasing tank 11 is transported. The chamber 21 includes a ceiling 22, a floor 23, and a pair of side walls 24. The chamber 21 also has an inlet 25 for loading the automobile body W1 into the chamber 21 and an outlet 26 for removing the automobile body W1 from the chamber 21. The chamber 21 also has a number of shower nozzles 27 for spraying a cleaning liquid onto the surface of the automobile body W1. By spraying the cleaning liquid from each nozzle 27, excess degreaser P1 adhering to the surface of the automobile body W1 is washed away.

[0033] As shown in FIG. 1, the shower process quality control system 10 includes a conveyor 31 that transports an automobile body W1 in a transport direction (leftward in FIG. 1). The conveyor 31 lowers the automobile body W1 to load it into the degreasing tank 11 through the inlet 15, and then lifts the automobile body W1 to unload it from the degreasing tank 11 through the outlet 16. The conveyor 31 also loads the automobile body W1 into the chamber 21 through the inlet 25 and unloads it from the chamber 21 through the outlet 26. The conveyor 31 includes a rail 32 extending in the transport direction and a plurality of hanger rails 33 attached to the rail 32 for suspending and transporting the automobile body W1. Each hanger rail 33 is also provided with a pulse transmitter 34 that periodically emits a pulse signal (conveyor pulse).

[0034] As shown in FIG. 2, the cleaning line 20 is equipped with IP cameras 40 (imaging devices) that capture images 41 (see FIG. 3) of an automobile body W1 passing through the chamber 21. A pair of IP cameras 40 are provided on either side of the automobile body W1 passing through the chamber 21, and are respectively disposed near a pair of side walls 24 that constitute the chamber 21. Another IP camera 40 is provided above the automobile body W1 passing through the chamber 21, and is disposed near the ceiling 22 that constitutes the chamber 21. Each IP camera 40 stores image data of the captured image 41. Note that the IP cameras 40 of this embodiment are network cameras that can track the automobile body W1 transported on the cleaning line 20. The IP cameras 40 are corrosion-resistant PTZ (Pan-Tilt-Zoom) cameras that can be controlled by pan, tilt, and zoom.

[0035] Next, the electrical configuration of the shower process quality control system 10 will be described.

[0036] As shown in FIG. 1, the shower process quality control system 10 includes a personal computer 50, which includes a control device 51 for overall control of the entire system. The control device 51 is composed of a CPU 52, a ROM 53, a RAM 54, an input / output circuit, etc. The CPU 52 is electrically connected to the conveyor 31, the pulse generator 34, the IP camera 40, etc., and controls them using various drive signals. The CPU 52 is also electrically connected to the pump 28 for supplying cleaning liquid to the shower nozzles 27, and controls it using drive signals. The CPU 52 is also electrically connected to a display 55 (display device) and a keyboard 56. The display screen of the display 55 displays an image 41 captured by the IP camera 40. The CPU 52 is also periodically input with conveyor pulses output from each pulse generator 34. The ROM 53 stores a program for controlling the shower process quality control system 10.

[0037] Next, a shower process quality control method using the shower process quality control system 10 will be described.

[0038] First, the CPU 52 outputs a drive signal to the conveyor 31, for example, to continuously transport the automobile bodies W1 (hanger rails 33) into the degreasing tank 11 and continuously transport the automobile bodies W1 out of the degreasing tank 11. When the automobile bodies W1 transported into the degreasing tank 11 are immersed in the degreasing agent P1, oil adhering to the surface of the automobile bodies W1 is removed.

[0039] The CPU 52 can also continuously load the automobile bodies W1 into the chemical conversion tank and continuously load the automobile bodies W1 out of the chemical conversion tank. In this case, when the automobile bodies W1 loaded into the chemical conversion tank are immersed in the chemical conversion agent, a chemical conversion coating is formed on the surface of the automobile bodies W1.

[0040] Then, after being transported out of the degreasing tank 11 or the chemical conversion tank, the automobile body W1 is transported into the chamber 21 of the cleaning line 20 through the transport inlet 25. At this time, the CPU 52 outputs a drive signal to the pump 28, causing the shower nozzles 27 to spray cleaning liquid onto the automobile body W1. This performs a shower process in which the automobile body W1 is cleaned, and excess degreaser P1 and chemical conversion agent adhering to the surface of the automobile body W1 are washed away.

[0041] The CPU 52 also outputs a drive signal to the IP camera 40, controlling the IP camera 40 to capture an image 41 of the automobile body W1 as it passes through the shower process. Specifically, the IP camera 40 references a pre-registered imaging point in response to conveyor pulses and production instruction information to capture the image 41. Note that the part W2 (see FIG. 3) that constitutes the automobile body W1 is the right front door of the automobile body W1, but similar work is also performed on the right rear door, left front door, left rear door, roof, fenders, etc. In this embodiment, only the case where the right front door is used as the part W2 will be described, and a description of the case where the right rear door, left front door, left rear door, roof, fenders, etc. is omitted.

[0042] More specifically, first, the CPU 52 counts the number of conveyor pulses (pulse count) output from the pulse transmitter 34 corresponding to the automobile body W1 while the automobile body W1 moves from a reference position near the entrance 25 of the chamber 21 to its current position. The pulse count is performed for all automobile bodies W1 passing through the shower process. Next, the CPU 52 calculates the relative distance between the IP camera 40 and the automobile body W1 at its current position based on the counted pulse count. Then, based on the calculated relative distance, the CPU 52 detects whether the automobile body W1 has arrived at a pre-registered imaging point, and upon detecting the arrival of the automobile body W1, issues an imaging command to the IP camera 40.

[0043] Next, the CPU 52 controls (PTZ control) the operation of the IP camera 40 in accordance with the progress of the automobile body W1 passing through the shower process. At the same time, the CPU 52 outputs a drive signal to the IP camera 40, and controls the IP camera 40 to capture an image 41 by capturing an image of the part W2.

[0044] Furthermore, the CPU 52 constructs a semantic segmentation model for each part W2 from the image 41 of the automobile body W1 stored in the IP camera 40. The CPU 52 then stores the constructed semantic segmentation model in the RAM 54. The CPU 52 then performs image processing on the image 41 using the semantic segmentation model to detect a shower spray target area A2 (see FIG. 4) within the part W2. The shower spray target area A2 is the area surrounded by a dashed line in FIG. 4 and other figures, and indicates approximately the same range as the part W2. Semantic segmentation refers to object detection at the pixel level of the image 41. However, constructing a semantic segmentation model and performing image processing in real time involves problems such as long calculation times and high calculation costs.

[0045] Therefore, in this embodiment, when the part W2 passes through the shower discharge area A1, the IP camera 40 captures images of the shower discharge area A1 every predetermined time (1 second in this embodiment) to obtain multiple images 42 (20 images in this embodiment) (see FIG. 5). At this time, the IP camera 40 captures an area wider than the shower discharge area A1. The CPU 52 then performs image processing on each image 42 using a semantic segmentation model to detect the shower spray target area A2. This reduces calculation costs. The CPU 52 stores (manages) the shower spray data A2 for each of the multiple parts W2 in the RAM 54. In other words, the CPU 52 functions as a "data management means."

[0046] Next, the CPU 52 performs processing of a detection model construction step to construct a droplet detection model that detects droplets S1 (see FIG. 6) on the part W2 from each image 42 captured during the shower process. That is, the CPU 52 functions as a "detection model construction means." The droplet S1 may be a single droplet of cleaning liquid sprayed from the shower nozzle 27, or may be a collection of multiple particles. The CPU 52 then stores the constructed droplet detection model in the RAM 54.

[0047] In the subsequent area detection step, the CPU 52 detects multiple droplets S1 on the part W2 by performing image processing using the droplet detection model in the shower spray target area A2 set in the image 41. Specifically, the CPU 52 detects droplets S1 by performing image processing using the droplet detection model on each of the images 42 (see FIG. 6) obtained by capturing images of the shower discharge area A1 every predetermined time (one second in this embodiment). This makes it possible to reduce calculation costs.

[0048] Each detected droplet S1 is surrounded by a rectangular bounding box B1 (see FIG. 6). For ease of explanation, the bounding box B1 is shown relatively large in FIG. 6, but the actual bounding box B1 is shown to be much smaller and more numerous than in FIG. 6. The CPU 52 then acquires the coordinates of the area on the surface of the part W2 with which the droplet S1 is in contact. Specifically, the CPU 52 acquires, for example, the X coordinate of a first end (the right end in FIG. 6) of the long side of the bounding box B1 that surrounds the detected droplet S1, and the X coordinate of a second end (the left end in FIG. 6) of the long side. The acquired X coordinates are then stored in the RAM 54. The CPU 52 may also acquire the Y coordinate of a first end (the upper end in FIG. 6) of the short side of the bounding box B1, and the Y coordinate of a second end (the lower end in FIG. 6) of the short side.

[0049] Next, the CPU 52 detects a previous shower spraying area A3 (see FIG. 7) and an unsprayed area A4 (see FIG. 7) within the shower spraying target area A2. That is, the CPU 52 functions as an "area detection unit." Specifically, the CPU 52 calculates the area of ​​overlap between the shower spraying target area A2 stored in RAM 54 and the bounding box B1, and detects the overlapping area as the previous shower spraying area A3. The CPU 52 detects the overlapping area in each of the images 42 obtained by capturing images of the shower discharge range A1 every predetermined time (one second in this embodiment). The CPU 52 also detects the portion of the shower spraying target area A2 excluding the overlapping area as the unsprayed area A4. Although the previous shower spraying area A3 shown in FIG. 7 has an angular shape, the actual bounding box B1 is quite small, so the actual shower spraying area A3 has a rounded shape (see FIG. 10).

[0050] Next, the CPU 52 sets the tenth image 42 of the 20 images 42 as the reference image 43 (see FIG. 8(b)). Next, the CPU 52 superimposes the reference image 44 onto the reference image 43 by matching the front end coordinate F1 (see FIG. 8(b)) of the shower spray target area A2 set in the reference image 43 (the front end in the traveling direction of the part W2, which is the right end in FIG. 8(b)) with the front end coordinate F2 (see FIG. 8(a)) of the shower spray target area A2 set in the reference image 44 (see FIG. 8(a)) acquired before the reference image 43. In other words, the CPU 52 superimposes the reference image 44 onto the reference image 43 by moving the front end coordinate F2 of the reference image 44 forward by a first offset distance D1 (to the right in FIG. 8(a)). In this embodiment, the first to ninth images 42 are set as the reference images 44. At the same time, the CPU 52 superimposes the reference image 43 onto the reference image 45 by matching the rear end coordinate R1 (see FIG. 8(b)) of the shower spray target area A2 set in the reference image 43, which is the rear end (left end in FIG. 8(b)) in the traveling direction of the part W2, with the rear end coordinate R2 (see FIG. 8(c)) of the shower spray target area A2 set in the reference image 45 (see FIG. 8(c)) acquired after the reference image 43. In other words, the CPU 52 superimposes the reference image 45 onto the reference image 43 by moving the rear end coordinate R2 of the reference image 45 backward (leftward in FIG. 8(c)) by the second offset distance D2. In this embodiment, the eleventh to twentieth images 42 are set as the reference image 45. As a result, all of the reference images 44 and all of the reference image 45 are integrated into the reference image 43. For ease of explanation, Figure 9 shows one pre-reference image 44 and one post-reference image 45 integrated into the reference image 43, but in reality, nine pre-reference images 44 and ten post-reference images 45 are integrated into the reference image 43 (see Figure 10).

[0051] The CPU 52 then controls the display 55 to display the shower sprayed area A3 and the unsprayed area A4 in color and numerical values ​​(see FIG. 11). Specifically, the CPU 52 displays the analysis results of the integrated final analysis image 46 as a heat map on the display 55. For example, if the number of pixel coordinate points (number of pixels) of the shower sprayed area A2 detected by semantic segmentation is "100" per image 42, the total number of pixels for 20 images 42 will be "2000." In contrast, the sum (number of pixels) of the analysis results of the overlap of the shower sprayed area A3 in each image 42 on a pixel-by-pixel basis can be displayed as a percentage of the total, for example, "1325." The CPU 52 calculates the ratio (i.e., the sprayed area ratio; 85% in FIG. 11) of the number of pixels in the shower spraying history area A3 ("1325" in FIG. 11) to the number of pixels in the shower spraying target area A2 ("2000" in FIG. 11), and displays the calculation result 47 on the display 55. When displaying 20 images 42 as a heat map of blue to red, the numbers range from "0" to "20" and are displayed in blue to red (for example, if there is overlap between 10 images 42, the number range becomes "10" and is displayed in an intermediate color between blue and red), and this is performed for each pixel in which the part W2 is detected. In this embodiment, the shower spraying history area A3 is displayed in dark blue (dense hatching) or light blue (medium-coarse hatching), and the unsprayed area A4 is displayed in light red (coarse hatching).

[0052] In the subsequent prediction model construction step, the CPU 52 accumulates (stores in RAM 54) spray performance data obtained from the shower spray performance area A3 and the unsprayed area A4 to construct a quality prediction model that predicts the quality of the shower spray target area A2 after a predetermined period of time. In other words, the CPU 52 functions as a "prediction model construction means." Specifically, the CPU 52 utilizes the spray performance data as traceability and constructs a quality prediction model based on the history of combinations of shower discharge conditions (such as the angle of the nozzle 27, the pressure of the pump 28, and the temperature of the cleaning liquid) and the quality of the shower spray target area A2 obtained by those discharge conditions. The CPU 52 then stores the constructed quality prediction model in RAM 54.

[0053] Thereafter, the CPU 52 records (stores in the RAM 54) the final analysis image 46 in association with the body ID associated with each automobile body W1. Furthermore, the CPU 52 registers the final analysis image 46 associated with the body ID in a database of a system server (not shown). This allows the registered final analysis image 46 to be used for quality analysis of the shower spraying target area A2 and for detecting alerts.

[0054] The CPU 52 then derives optimal discharge conditions (optimal nozzle 27 angle, optimal pump 28 pressure, optimal cleaning liquid temperature, etc.) based on the quality prediction model for each showering target area A2 (each part W2). This calculates showering discharge conditions that will allow the quality of the showering target area A2 to be achieved after a predetermined period of time. The CPU 52 then controls the shower based on the derived discharge conditions.

[0055] If the spraying record stored (accumulated) in RAM 54 falls below a specified value (specifically, if the spraying area ratio is less than, for example, 70%), CPU 52 outputs a drive signal to display 55 and performs control to notify (display on display 55) that cleaning is insufficient (for example, the text "Insufficient cleaning"). In other words, display 55 functions as a "notification means." On the other hand, if the spraying record stored in RAM 54 achieves the specified value, CPU 52 performs control to terminate (turn off) shower discharge control (specifically, drive control of pump 28).

[0056] Therefore, according to this embodiment, the following effects can be obtained.

[0057] (1) In the shower process quality control system 10 of this embodiment, the droplet detection model detects droplets S1 within the shower spray target area A2 set in the image 42, and the quality prediction model predicts the quality of the shower spray target area A2 after a predetermined period of time. This automates the sensory evaluation (specifically, evaluation of the degree to which the cleaning liquid is applied to the automobile body W1), which was previously performed visually by a supervisor. This allows the spray distribution to be determined without being influenced by the supervisor's experience, thereby reducing quality variation. Therefore, shower discharge conditions (such as the angle of the nozzle 27, the pressure of the pump 28, and the temperature of the cleaning liquid) can be optimized based on the quality of the shower spray target area A2 predicted by the quality prediction model. Furthermore, automating the sensory evaluation eliminates the need for the supervisor to peer into the cleaning line 20 where the automobile body W1 is washed, thereby significantly reducing the supervisor's burden.

[0058] (2) For example, when the automobile body W1 has finished passing through the shower process, if it is detected that the pressure or flow rate of the pump 28 is below a specified value, it is possible to notify of insufficient cleaning. However, in this case, if the angle of the nozzle 27 is improper or the operation switch of the shower process quality control system 10 is forgotten to be turned on, the automobile body W1 may become a defective product and flow into a subsequent process (such as an electrodeposition coating process) due to the degreasing agent P1 or chemical agent remaining on the body.

[0059] On the other hand, in this embodiment, when the accumulated spraying performance (percentage of sprayed area) falls below a specified value, the display 55 notifies the user that the cleaning is insufficient. In other words, the manager can quickly detect and deal with the insufficient cleaning before the automobile body W1 passes through the shower process, thereby preventing the outflow of defective automobile bodies W1.

[0060] (3) For example, when a manager checks the condition of the shower by peering into the cleaning line 20, the manager may not be able to see one side of the automobile body W1 due to the location of the inspection window (not shown) or access door (not shown) in the chamber 21. In this case, there is a risk that the automobile body W1, which has been deemed defective due to insufficient cleaning, will be released. Therefore, in this embodiment, the automobile body W1 is imaged using a pair of IP cameras 40 installed on either side of the automobile body W1 passing through the chamber 21, and one IP camera 40 installed above the automobile body W1 passing through the chamber 21. This allows the manager to reliably check the condition of the entire automobile body W1, thereby preventing the release of defective automobile bodies W1.

[0061] (4) In this embodiment, the CPU 52 combines the reference image 43 with the reference front image 44 by aligning the front end coordinates F1 and F2 of the shower spray target area A2, and combines the reference rear image 45 with the reference image 43 by aligning the rear end coordinates R1 and R2 of the shower spray target area A2. This eliminates the need for markers, which are attached to the front end or rear end of the part W2 in the traveling direction, for example, and serve as positional references when combining images. However, if markers are used, there is a risk that the markers may become invisible in a chemical atmosphere, for example. Another problem is the cost of maintaining the markers.

[0062] (5) In this embodiment, when the CPU 52 determines that the spray performance of the shower has reached a predetermined value, it terminates the shower discharge control (specifically, the drive control of the pump 28). This reduces the power required to drive the pump 28, which can lead to energy savings.

[0063] (6) In this embodiment, the IP camera 40 registers an imaging point for each part W2 according to the conveyor pulse and production instruction information, and can accurately image the part W2 based on the registered imaging point. Furthermore, the IP camera 40 is a camera that has a computer, is assigned an IP address, and is connected to the control device 51 via a network. This eliminates the need for a process of extracting captured data and sending it to the CPU 52 of the control device 51.

[0064] The above embodiment may be modified as follows.

[0065] The shower process quality control system 10 in the above embodiment is a system that controls the shower discharge conditions for washing the automobile body W1 immersed in the degreaser P1 in the degreasing tank 11 and the shower discharge conditions for washing the automobile body W1 immersed in the chemical conversion agent in the chemical conversion tank. However, the shower process quality control system 10 may also be a system that controls the shower discharge conditions for washing the automobile body W1 immersed in the electrodeposition paint in the electrodeposition tank.

[0066] In the above embodiment, the droplet S1 detected using the droplet detection model is surrounded by a rectangular bounding box B1. However, the droplet S1 may be surrounded by a bounding box of another shape, such as a triangle, square, diamond, trapezoid, circle, or ellipse.

[0067] In the above embodiment, the IP camera 40 with an integrated computer is used as the imaging device, but a camera without a computer may also be used as the imaging device.

[0068] In the above embodiment, a pair of IP cameras 40 are provided on either side of the automobile body W1 passing through the chamber 21, and one IP camera 40 is provided above the automobile body W1 passing through the chamber 21. However, four or more IP cameras 40 may be provided, or only one IP camera 40 may be provided.

[0069] In the above embodiment, an automobile body W1 is used as an example of the object to be washed with a shower, but this is not a limitation. For example, the object may be an automobile interior part such as an instrument panel, console box, or armrest, or an automobile exterior part such as a bumper or an aerodynamic part (spoiler, etc.). Furthermore, the object does not necessarily have to be an automobile part.

[0070] Next, in addition to the technical ideas set forth in the claims, the technical ideas grasped by the above-described embodiments will be listed below.

[0071] (1) In claim 3, the shower process quality control system is characterized in that the imaging devices are arranged in pairs to sandwich the object passing through the cleaning line, and are arranged above the object passing through the cleaning line.

[0072] (2) In claim 5, the area detection means is characterized in that it composites the reference pre-image onto the reference image by moving the front end coordinate of the reference pre-image forward by a first offset distance, and composites the reference post-image onto the reference image by moving the rear end coordinate of the reference post-image backward by a second offset distance.

[0073] (3) In claim 8, the shower process quality control system is characterized in that, when the spraying performance accumulated by the predictive model construction means achieves the specified value, control is performed to terminate shower discharge control. [Explanation of symbols]

[0074] 10...Shower process quality control system 20...Washing line 40...IP camera as an imaging device 41, 42...Image of the object to be painted 43...Reference image 44...Before reference image 45...Baseline image 52...CPU as detection model construction means, area detection means, prediction model construction means, and data management means 55...Display as a display device and notification means A1...Shower discharge range A2: Shower spray target area A3: Shower spraying area A4: Unsprayed area B1...Bounding box F1,F2…front end coordinates R1,R2…Rear end coordinates S1…Droplet W1: Automobile body as the object to be painted W2…Parts

Claims

1. A system for managing the quality of a shower process by controlling the discharge conditions of a shower that cleans a workpiece, a detection model constructing means for constructing a droplet detection model for detecting droplets on the object to be coated from an image of the object to be coated taken while the object is passing through the showering process; an area detection means for detecting areas where shower spraying has been performed and areas where shower spraying has not been performed within a predetermined shower spraying target area set in the image by performing image processing using the droplet detection model; a prediction model construction means for constructing a quality prediction model for predicting the quality of the shower spray target area after a predetermined period of time by accumulating data on the spray performance obtained from the shower spray performance area and the unsprayed area; A shower process quality control system comprising:

2. a data management means for managing data on the shower spray target area for each of a plurality of parts constituting the object to be coated, The area detection means calculates an area of ​​an overlap between the shower spray target area managed by the data management means and a bounding box surrounding the droplets detected using the droplet detection model, and detects the overlap as the shower spray history area. The shower process quality control system according to claim 1 .

3. 3. The shower process quality control system according to claim 2, further comprising an imaging device that is installed in a cleaning line that cleans the object to be coated and captures the image of the object to be coated.

4. the imaging device captures an image of the shower discharge area at predetermined time intervals when the object passes through the shower discharge area, thereby obtaining a plurality of images; The area detection means detects the overlapping portion in each of the plurality of images. The shower process quality control system according to claim 3 .

5. The area detection means setting one of the plurality of images as a reference image; The reference image is synthesized with the reference image so that the front end coordinates of the shower spray target area set in the reference image, which is at the front end in the traveling direction of the object to be coated, match the front end coordinates of the shower spray target area set in a reference image acquired before the reference image; and The rear end coordinates of the shower spray target area set in the reference image, which is at the rear end in the traveling direction of the object to be coated, are matched with the rear end coordinates of the shower spray target area set in a reference rear image acquired after the reference image, and the reference rear image is synthesized with the reference image; Merging the images into one image The shower process quality control system according to claim 4 .

6. The shower process quality control system described in claim 3, characterized in that the imaging device is a network camera capable of tracking the coated object being transported on the cleaning line, and the imaging point is registered for each part according to conveyor pulses and production instruction information to acquire the image.

7. 2. The shower process quality control system according to claim 1, further comprising a display device that displays the shower sprayed areas and the unsprayed areas using colors and numerical values.

8. 2. The shower process quality control system according to claim 1, further comprising a notification means for notifying that cleaning is insufficient when the spraying performance accumulated by the prediction model construction means falls below a specified value.

9. 2. The shower process quality control system according to claim 1, wherein the optimal discharge conditions are derived based on the quality prediction model for each shower spray target area, and the shower is controlled based on the derived discharge conditions.

10. A method for managing the quality of a shower process by controlling the discharge conditions of a shower for cleaning a coated object, comprising: a detection model construction step of constructing a droplet detection model for detecting droplets on the object to be coated from an image of the object to be coated taken while the object is passing through the showering process; an area detection step of detecting areas where shower spraying has been performed and areas where shower spraying has not been performed within a predetermined shower spraying target area set in the image by performing image processing using the droplet detection model; a prediction model construction step of constructing a quality prediction model for predicting the quality of the shower spray target area after a predetermined period of time by accumulating data on the spray performance obtained from the shower spray performance area and the unsprayed area; A shower process quality control method comprising:

Images