Quality control system and quality control methods

Abstract

This system automates the checking process for various items to be checked in the painting process. [Solution] In a quality control system 1 that includes a photographing unit 2 that photographs an object to be photographed, which includes at least one of a painting device and an object to be painted by the painting device, and a calculation unit that, when an image photographed by the photographing unit 2 is input, uses a classifier to output the state of the object to be photographed related to the image, the classifier is constructed based on images photographed when the temperature control of the painting device is normal and images photographed when the temperature control of the painting device is abnormal.

Description

Technical Field

[0001] The present invention relates to a quality management system and a quality management method.

Background Art

[0002] The quality of painting can be affected by various parameters in the painting process. Therefore, attempts have been made to clarify the relationship between the implementation status of the painting process and the painting quality of the object to be painted. However, among the items to be confirmed regarding the implementation status of the painting process, there are many items that are difficult to quantify and items that require expensive equipment even if quantification is possible. Therefore, at present, many manual confirmation operations are being carried out.

[0003] As an attempt to automate such confirmation operations, for example, Japanese Patent Application Laid-Open No. 2016-212044 (Patent Document 1) discloses a pointer reading device that reads the pointer value of an instrument from a photographed image of the instrument. Such a device can also be applied to the painting process.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, the matters for which automation of confirmation operations is desired in the painting process are not limited to reading pointer values and there are many of them. Since the function of the technique of Patent Document 1 is specialized in reading pointer values, it has been difficult to apply it generally to various confirmation matters in the painting process.

[0006] Therefore, there is a demand for realizing a quality management system and a quality management method that can generally automate the confirmation operations for various confirmation matters in the painting process. [Means for solving the problem]

[0007] The quality control system according to the present invention is a painting equipment installed in a painting booth Place The painted surface things The photography department Equipped with When an image captured by the aforementioned imaging unit is input, the image related to that image Condition of the object to be coated The state is output using a classifier. items In the quality control system, the classifier is constructed based on images taken when the temperature control of the paint booth is normal and images taken when the temperature control of the paint booth is abnormal.

[0008] The quality control method according to the present invention is a painting equipment installed in a painting booth Place The painted surface things When the shooting process and the image captured by the shooting process are input, the information relating to the image Condition of the object to be coated A quality control method comprising an output step of outputting a state using a classifier, wherein the classifier is constructed based on an image taken when the temperature control of the paint booth is normal and an image taken when the temperature control of the paint booth is abnormal.

[0009] With these configurations, various verification tasks in the painting process can be automated in a general-purpose manner.

[0010] Further features and advantages of the present invention will become clearer through the following description of exemplary and non-limiting embodiments, with reference to the drawings. [Brief explanation of the drawing]

[0011] [Figure 1] Schematic diagram of the painting line [Figure 2] Configuration diagram of a quality control system according to an embodiment of the present invention. [Figure 3] This figure shows a first example of use of a quality control system according to an embodiment of the present invention. [Figure 4] A figure showing a second example of use of a quality control system according to an embodiment of the present invention. [Figure 5] This figure shows a third example of the use of a quality control system according to an embodiment of the present invention. [Figure 6] This figure shows a fourth example of the use of a quality control system according to an embodiment of the present invention. [Figure 7] This figure shows a fifth example of use of the quality control system according to an embodiment of the present invention. [Figure 8] This figure shows a seventh example of the use of the quality control system according to an embodiment of the present invention. [Figure 9] A figure showing a variation of the seventh example of use of the quality control system according to an embodiment of the present invention. [Modes for carrying out the invention]

[0012] Embodiments of the quality control system and quality control method according to the present invention will be described with reference to the drawings. Below, an example will be described in which the quality control system according to the present invention is applied to a quality control system 1 capable of controlling the quality of a vehicle body B, which is a workpiece to be painted, in a painting line 100.

[0013] [Painting line configuration] First, the painting line 100 according to this embodiment will be described. The painting line 100 is a series of facilities capable of painting a vehicle body B (Figure 1). Each part of the painting line 100 is configured in a tunnel shape, and the vehicle body B is transported on a single axis (left-right direction in the plane of the paper in Figure 1) by a conveyor C while each painting process is performed on the vehicle body B. In this embodiment, a plurality of painting booths 101 and inspection booths 102 are provided along the transport path of the conveyor C. In addition, a quality control system 1 is provided capable of receiving signals from each painting booth 101, inspection booth 102, and conveyor C.

[0014] Each painting booth 101 is provided with a painting device according to the content of the painting work performed in each painting booth 101. Here, the "painting device" in the present embodiment includes devices that directly perform the work of the painting process on the vehicle body B, such as a spray painting device, an electrocoating device, a dip treatment tank, a drying device, etc., and also includes utility devices such as a storage tank, a compressed air supply facility, an air conditioning facility, a supply and exhaust facility, etc. In addition, each painting booth 101 is provided with measuring instruments such as a thermometer, a hygrometer, a wind speed meter, an ammeter, a tachometer, etc. for observing the operating status of the painting device and the environment inside the painting booth 101, and instruments such as a liquid level gauge and a pressure gauge.

[0015] The inspection booth 102 is a booth where quality inspection related to the painting applied to the vehicle body B is performed. In the inspection booth 102, the hue, film thickness, and finish appearance are detected by inspections automatically performed using inspection devices, and the presence or absence of dust, bumps, thread-like bumps, peeling, hue, and the degree of the finish texture are determined by manual inspections. The results of the manual inspections are input into a computer and sent to the quality management system 1 as electronic data together with the inspection results by the inspection devices.

[0016] The conveyor C is a device capable of conveying the vehicle body B along the conveyance path. The conveyor C is provided with a vehicle body identification device D capable of reading the vehicle body identification number (an example of individual identification information) attached to the vehicle body B and an encoder E capable of detecting the movement amount of the vehicle body B.

[0017] 〔Configuration of Quality Management System〕 Next, the configuration of the quality management system 1 according to the present embodiment will be described. The quality management system 1 is provided with a photographing unit 2, a storage unit 3, a learning unit 4, an input unit 5, and a display 6 (FIG. 2). Briefly summarizing the functions of the quality management system 1, the quality management system 1 constructs a classifier that takes the image of the photographing object as an input and outputs the state of the photographing object based on the image of the painting device, the object to be painted, or both (hereinafter referred to as the photographing object).

[0018] The imaging unit 2 is a device configured to photograph an object including at least one of a painting apparatus and an object to be painted, and is specifically implemented as a digital camera module, a digital thermal camera module, etc. As is clear from the example of a digital thermal camera module, the images captured by the imaging unit 2 are not limited to ordinary digital images, but may be images with special functions such as thermal images. Furthermore, the images may be still images or moving images (i.e., a collection of still images). The captured images are stored in the storage unit 3.

[0019] Furthermore, if the shooting unit 2 is implemented as a digital camera module, it is preferable that the digital camera module is capable of operating in both normal mode and night vision mode. Here, normal mode refers to a mode in which an image (first image) is captured using a sensor capable of detecting visible light. Night vision mode refers to a mode in which an image (second image) is captured using, for example, a sensor capable of detecting infrared light. With night vision mode, a clear image can be obtained even when the illumination is insufficient and a clear image cannot be captured in normal mode. It is also preferable that the shooting unit 2 (digital camera module) has an illuminance meter capable of detecting ambient illumination and is configured to automatically select either normal mode or night vision mode based on the illumination detected by the illuminance meter. Furthermore, it is preferable that the data related to the captured image includes information that can identify the shooting mode (normal mode or night vision mode).

[0020] The storage unit 3 is a device configured to store images and other data captured by the shooting unit 2, and is specifically implemented as a known storage device such as a hard disk or semiconductor memory. If the data related to the captured image includes information that can identify the shooting mode (normal mode or night vision mode), the storage unit 3 stores the image together with that information.

[0021] The learning unit 4 is configured to construct a classifier that outputs the state of the object being photographed when a new image is input. This classifier is constructed based on multiple images of the object being photographed stored in the memory unit 3. The learning unit 4 is implemented as a known processing unit such as a CPU. Examples of algorithms used to construct the classifier include linear regression, random forest, Bayesian regression, logistic regression, and support vector machines. The constructed classifier is stored in the memory unit 3. If there are images taken in normal mode (first image) and images taken in night vision mode (second image), the learning unit 4 separately constructs a first classifier (classifier for normal mode) based on multiple first images and a second classifier (classifier for night vision mode) based on multiple second images taken in night vision mode.

[0022] The input unit 5 is configured to accept input of inspection results sent from the inspection booth 102. Specifically, it is implemented as an input interface provided on a computer. The connection between the input unit 5 and the inspection booth 102 may be wired or wireless. The input inspection results are stored in the storage unit 3.

[0023] The display 6 is configured to display images captured by the imaging unit 2 and information related to the state of the object being photographed, output by the classifier, and is implemented as a known liquid crystal display or the like.

[0024] Furthermore, the quality control system 1 is equipped with an individual identification unit (not shown). The individual identification unit first identifies which booth a vehicle B, identified by the vehicle identification number identified by the vehicle identification device D, was in at what time, based on the amount of movement detected by the encoder E. Next, based on the relationship between the time and booth identified as described above, it identifies which individual vehicle B the images taken by the imaging unit 2 and the inspection results input to the input unit 5 are related to. As a result, the various information stored in the storage unit 3 is associated with the vehicle identification number of vehicle B.

[0025] [Examples of using a quality control system] The following sections will describe, with specific examples, how to actually use the quality management system 1 according to this embodiment. In particular, the following descriptions will focus on the type of classifier constructed in each specific example. Note that the following examples do not limit the methods of using the quality management system according to the present invention.

[0026] (Specific example 1: Determining the quality of the spray pattern in the spraying process) In the first specific example, a classifier is constructed that takes an image P1 (Figure 3) of a spray painting apparatus 103 (an example of a painting apparatus) and a vehicle body B during painting as input, and outputs the quality of the spray pattern (an example of the condition of the object being photographed). Here, the spray pattern in this specific example refers to the discharge direction α, discharge angle β, and reach distance L of the spray liquid discharged from the spray painting apparatus 103. In this specific example, the imaging unit 2 is installed in a position and orientation that allows it to photograph the spray painting apparatus 103.

[0027] In this specific example, the classifier constructed by the learning unit 4 first identifies the contour lines 104 of the spray liquid discharged from the spray painting device 103 and applied to the vehicle body B in the input image P1. Such a classifier may be constructed by supervised learning using a group of images P1 in which the contour lines 104 have been artificially identified, or it may be constructed by unsupervised learning using a group of images P1 in which the contour lines 104 have not been identified.

[0028] Next, the classifier determines the discharge direction α, discharge angle β, and reach distance L of the spray liquid based on the identified contour line 104. Subsequently, the quality of the spray pattern is determined based on the discharge direction α, discharge angle β, and reach distance L, and the values ​​of discharge direction α, discharge angle β, and reach distance L, as well as the determination result, are displayed on the display 6.

[0029] (Specific example 2: Determining the liquid level in the storage tank) In the second specific example, a classifier is constructed that takes an image P2 (Figure 4) of a storage tank 105 (an example of a painting device) installed on the painting line 100 as input and outputs the liquid level in the storage tank 105 (an example of the state of the object being photographed). In this specific example, the imaging unit 2 is installed in a position and orientation that allows it to photograph the liquid level (paint, etc.) stored in the storage tank 105.

[0030] In this specific example, the classifier constructed by the learning unit 4 first identifies the contour line 106 in the input image P2 where the liquid surface of the liquid stored in the storage tank 105 touches the inner wall surface of the storage tank 105. As with the first specific example, the classifier may be constructed using supervised learning or unsupervised learning.

[0031] Next, the classifier calculates the distance H between the identified contour line 106 and the upper end 107 of the storage tank 105. Since the position of the upper end 107 is constant, there is a one-to-one correspondence between the distance H and the liquid level. Therefore, the liquid level can be determined based on the distance H. The liquid level determined in this way is displayed on the display 6.

[0032] (Specific example 3: Determination of liquid surface velocity and foam generation status in storage tanks) In the third specific example, a classifier is constructed that takes a video P3 (Figure 5) of a dipping tank 108 (an example of a painting apparatus) installed in the painting line 100 as input, and outputs a determination result regarding the flow velocity and bubble generation status at the liquid surface of the paint stored in the dipping tank 108 (an example of the state of the object being filmed). In this specific example, the filming unit 2 is installed in a position and orientation that allows it to film the liquid surface of the paint stored in the dipping tank 108. More precisely, in order to reduce the processing load on the learning unit 4, a group of still images extracted at a constant sampling rate from the video P3 filmed by the filming unit 2 is used as input to the classifier. However, in the explanation of this specific example, the video P3 is simply treated as input.

[0033] In this specific example, the classifier constructed by the learning unit 4 first identifies the bubbles 109 that have formed on the surface of the paint stored in the dipping tank 108 in the input video P3. As with the first specific example, the classifier may be constructed using supervised learning or unsupervised learning.

[0034] Next, the classifier compares multiple still images that make up the video P3 and tracks the movement of each individual bubble 109. Based on this tracking, it calculates the movement speed of each bubble 109. Furthermore, based on this movement speed, it calculates the flow velocity at the liquid surface of the paint stored in the dipping tank 108. In addition, based on the number of identified bubbles 109, it determines whether the bubble generation situation at the liquid surface of the paint stored in the dipping tank 108 is normal or abnormal. The judgment results regarding the flow velocity and bubble generation situation derived in this way are displayed on the display 6.

[0035] Furthermore, the second and third specific examples described above may be applied simultaneously to the same image taken of the same tank. That is, based on an image of the liquid surface stored in a tank, the liquid level in the tank, the flow velocity at the liquid surface, and the state of bubble generation can be determined simultaneously.

[0036] (Specific example 4: Determining the contamination level of the treated liquid) In the fourth specific example, a classifier is constructed that takes an image P4 (Figure 6) of a processing liquid pipe 110 (an example of a painting apparatus) installed in the painting line 100 as input, and outputs the contamination level of the processing liquid flowing through the processing liquid pipe 110 (an example of the state of the object being photographed). In this specific example, the imaging unit 2 is installed in a position and orientation that allows it to photograph the sight glass 110a installed in the middle of the processing liquid pipe 110.

[0037] In this specific example, the classifier constructed by the learning unit 4 first identifies the viewing window portion 110b of the sight glass 110a in the input image P4. As with the first specific example, the classifier may be constructed using supervised learning or unsupervised learning.

[0038] Next, the classifier acquires the brightness value of the viewing window portion 110b of image P4. Since this brightness value reflects the color of the processing liquid flowing through the processing liquid piping 110, the contamination level of the processing liquid can be determined based on this brightness value. More specifically, the acquired brightness value is converted to a contamination level based on a pre-set correspondence table between brightness values ​​and contamination levels, and this contamination level is displayed on the display 6.

[0039] (Specific example 5: Determining the amount of leakage at the booth entrance / exit) In the fifth specific example, a classifier is constructed that takes an image P5 (Figure 7) of the inlet 111 of the painting line 100 (an example of a painting device) as input and outputs the amount of steam 112 leaking from the inlet 111 (an example of the state of the object being photographed). In this specific example, the imaging unit 2 is installed in a position and orientation that allows it to photograph the inlet 111.

[0040] In this specific example, the classifier constructed by the learning unit 4 first identifies the outline 112a of steam leaking from the inlet 111 to the outside of the painting line 100 in the input image P5. As with the first specific example, the classifier may be constructed using supervised learning or unsupervised learning.

[0041] Next, the classifier calculates the area of ​​the region enclosed by the contour line 112a. This area increases as the amount of steam leaking from the inlet 111 increases, so the amount of steam 112 leaking can be calculated based on this area. The calculated amount of leakage is displayed on the display 6. In addition, the amount of leakage of water, powder, paint, etc. can also be determined in the same manner as in this example.

[0042] (Specific example 6: Determining vehicle body temperature) In the sixth specific example, a classifier is constructed that takes a thermal image of a vehicle body B (an example of an object to be painted) taken in a paint booth 101 as input and outputs a judgment result regarding the quality of the temperature state of the vehicle body B (an example of the state of the object being photographed). In this specific example, the shooting unit 2 is implemented as a thermal camera capable of taking thermal images and is installed in the paint booth 101 in a position and orientation that allows it to photograph the entire vehicle body B.

[0043] In this specific example, the classifier constructed by the learning unit 4 is constructed using multiple thermal images of the vehicle body B taken when the temperature control of the paint booth 101 is normal, and multiple thermal images of the vehicle body B taken when the temperature control of the paint booth 101 is abnormal. At this time, each thermal image may be labeled with the temperature control status of the paint booth 101 (normal or abnormal) and supervised learning may be performed, or unsupervised learning may be performed without such labels. The constructed classifier determines whether the input thermal image corresponds to the case where the temperature control of the paint booth 101 is normal or abnormal, based on the temperature distribution pattern (or color distribution pattern) of the input thermal image. The result of this determination is displayed on the display 6.

[0044] Furthermore, in this specific example, each thermal image used in constructing the classifier is associated with the vehicle identification number of the vehicle body B corresponding to each thermal image, and the inspection results related to the vehicle body B are associated with that vehicle identification number. Therefore, the thermal image and the inspection results are associated with the vehicle identification number. Thus, the classifier in this specific example is also a classifier that takes a thermal image as input and outputs predicted inspection results. Inspection items that have a strong correlation with the temperature state of vehicle body B include, for example, the finished appearance and the degree of the finished surface, and according to this specific example, the inspection results related to these inspection items can be predicted with high accuracy. The predicted inspection results are displayed on the display 6 along with the judgment result (normal or abnormal) regarding the temperature control state of the paint booth 101.

[0045] In this specific example, the quality control system 1 may be equipped with an alarm unit (not shown) to issue an alarm if the predicted inspection results regarding the finished appearance and the degree of the finished skin meet predetermined warning criteria.

[0046] Furthermore, the quality control system 1 may be equipped with an operation correction unit (not shown) that can propose changes to the operating conditions of the paint booth 101 necessary to meet a predetermined quality objective if the predicted quality control values ​​do not meet that quality objective. In addition, it may be configured to actually change the operating conditions of the paint booth 101 according to the proposed operating conditions. In the example above, if it is predicted that the quality control values ​​related to the finished appearance and the degree of the finished surface will not meet the quality objective, the temperature conditions and processing time of the paint booth 101 can be changed to aim for meeting the quality objective.

[0047] In addition to the above example, process control methods that can be implemented based on the quality control values ​​predicted by the prediction unit include proposing repairs or replacements of painting equipment related to the painting booth 101, or proposing a temporary suspension of the entire painting line 100. Utilizing the predicted quality control values ​​in these examples can reduce product loss and the associated time and cost losses.

[0048] (Specific example 7: Meter reading of instrument gauges) In the seventh specific example, a classifier is constructed that takes an image P6 (Figure 8) of an instrument gauge 113 (an example of a painting device) installed in a painting booth 101 as input, and outputs the indicated value of the instrument gauge 113 (an example of the state of the object being photographed). In this specific example, the imaging unit 2 is installed in a position and orientation that allows it to photograph the instrument gauge 113. Note that Figure 8 shows an example where the instrument gauge 113 is a pressure gauge.

[0049] In this specific example, the classifier constructed by the learning unit 4 first identifies the dial portion 113a and the pointer portion 113b of the instrument gauge 113 in the input image P6. As with the first specific example, the construction of the classifier may be performed by supervised learning or by unsupervised learning. For example, in the case of supervised learning, master data consisting of image data of the instrument gauge 113, information identifying the dial portion 113a and the pointer portion 113b of the instrument gauge 113 in the image data, and the indicated value of the instrument gauge 113 in the image data is used as training data.

[0050] Next, the classifier determines the reading of the instrument gauge 113 based on the positional relationship between the identified dial portion 113a and the pointer portion 113b. For example, the rotation angle θ of the pointer portion 113b can be used as this positional relationship. In the example shown in Figure 8, for instance, the reading of the instrument gauge 113 is determined to be 0.4 MPa. The determined reading is then displayed on the display 6.

[0051] In this specific example, an image P7 (Figure 9) showing multiple instrument gauges 113 simultaneously may be used as input. The example shown in Figure 9 shows an example where a compressed air pressure gauge 113A, a steam temperature gauge 113B, and a steam pressure gauge 113C are photographed simultaneously. In this case, for each instrument gauge 113A to 113C, the dial portion and the pointer portion are identified, and the compressed air pressure and the steam temperature and pressure are identified.

[0052] Furthermore, as a variation of this specific example, a preferred embodiment will be described in which the imaging unit 2 is implemented as a digital camera module that can switch between normal mode and night vision mode. In this variation, a first classifier constructed based on images captured by the imaging unit 2 in normal mode and a second classifier constructed based on images captured by the imaging unit 2 in night vision mode are constructed separately. For example, in the case of supervised learning, the first master data based on images captured by the imaging unit 2 in normal mode and the second master data based on images captured by the imaging unit 2 in night vision mode are used as master data to construct the first and second classifiers.

[0053] When reading a meter, the first step is to identify the shooting mode of the images taken for meter reading. The mode in which each image was taken can be determined, for example, by including information that identifies the shooting mode (normal mode or night vision mode) in the data related to the captured image. Then, using a classifier (first classifier or second classifier) ​​corresponding to the mode in which the input image was taken, the indicated value of the instrument gauge 113 is determined by the method described above.

[0054] In other words, the modified quality control system further includes a shooting mode determination unit capable of determining the shooting mode in which the input image was taken. Based on the shooting mode determined by the shooting mode determination unit, a first classifier or a second classifier is selected, and when the image is input to the selected classifier, it outputs the state of the object to be photographed related to the image and the predicted inspection result for the coated object associated with the image.

[0055] [Other Embodiments] Finally, other embodiments of the quality control system, quality control method, and quality control program according to the present invention will be described. Note that the configurations disclosed in each of the following embodiments can be applied in combination with configurations disclosed in other embodiments, provided that no inconsistencies arise.

[0056] In the above embodiment, a quality control system 1 comprising a shooting unit 2, a storage unit 3, a learning unit 4, an input unit 5, and a display 6 was described as an embodiment of the present invention. However, the present invention may be a quality control program that causes a computer to perform the same functions as the above quality control system 1.

[0057] In the above embodiment, a configuration was described as one in which the image of the object to be photographed by the imaging unit 2 is used directly as input to the classifier. However, the system is not limited to such a configuration, and a preprocessing unit capable of performing preprocessing to remove the effects of disturbances from the image taken by the imaging unit may also be provided. Examples of such disturbances include effects related to the object to be photographed itself, such as dirt and scratches, and effects related to the lighting conditions of the shooting location, such as reflected light, color cast, and overexposure.

[0058] In the above embodiment, a configuration was described as an example in which a classifier is constructed in the quality control system 1 that takes an image of the object to be photographed as input and outputs the state of the object to be photographed based on images of the painting apparatus, the object to be painted, or both. However, the configuration is not limited to such an example, and the classifier may be configured to use electrical signals acquired from the painting apparatus, inspection apparatus, etc., in addition to images as input. Examples of such electrical signals include those emitted from thermometers, hygrometers, water meters, pressure gauges, tachometers, ammeters, etc., installed in the painting equipment.

[0059] In the above embodiment, a configuration was described in which the classifier constructed by the learning unit 4 is stored in the storage unit 3. However, the configuration is not limited to this, and the constructed classifier may be stored, for example, on an external server. In this case, the classifier can be used by quality control systems of individuals other than the quality control system that constructed the classifier.

[0060] In the above embodiment, a configuration in which the object to be painted is a vehicle body B was described as an example. However, the quality control system according to the present invention is not limited to such a configuration and can be applied to a painting line that paints any object to be painted. Furthermore, although the above embodiment exemplifies a configuration in which individual vehicle bodies B are identified using a vehicle body identification number, the quality control system according to the present invention can use any individual identification information corresponding to the object to be painted.

[0061] With regard to other configurations, the embodiments disclosed herein are illustrative in all respects, and it should be understood that the scope of the present invention is not limited thereto. Those skilled in the art will readily understand that modifications can be made as appropriate without departing from the spirit of the invention. Therefore, other embodiments modified without departing from the spirit of the invention are naturally included within the scope of the present invention. [Industrial applicability]

[0062] This invention can be used, for example, in a quality control system in a paint line for painting automobile bodies. [Explanation of Symbols]

[0063] 1: Quality Management System 2: Photography Department 3: Storage section 4: Learning Department 5: Input section 6: Display 100: Painting line 101: Paint Booth 102: Inspection Booth C: Conveyor D: Vehicle identification device E: Encoder B: Vehicle body

Claims

1. It is equipped with a camera unit that photographs the object to be painted by the painting equipment installed in the painting booth, When an image captured by the aforementioned imaging unit is input, the quality control system outputs the state of the object to be painted related to the image using a classifier, The quality management system is characterized in that the classifier is constructed based on images taken when the temperature control of the paint booth is normal and images taken when the temperature control of the paint booth is abnormal.

2. The quality control system according to claim 1, characterized in that the imaging unit is a thermal camera capable of capturing thermal images.

3. The quality control system according to claim 1 or 2, characterized in that it issues an alarm if the output inspection result falls under a predetermined warning criterion.

4. The quality control system according to any one of claims 1 to 3, characterized in that if the output inspection results do not meet a predetermined quality standard, the system proposes changing the operating conditions of the paint booth.

5. A photographic process for photographing the object to be painted by the painting equipment installed in the paint booth, A quality control method comprising: an output step in which, upon input of an image captured by the aforementioned shooting step, a classifier outputs the state of the object to be coated related to the image; The quality control method is characterized in that the classifier is constructed based on images taken when the temperature control of the paint booth is normal and images taken when the temperature control of the paint booth is abnormal.

Images