Surface anomaly detection device, surface anomaly detection method, and surface anomaly detection program
The surface anomaly detection device with pressure sensors on a rolling unit and machine learning analysis addresses the limitations of existing methods by offering precise and cost-effective surface inspection, detecting anomalies like scratches.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- TOYOTA PRODN ENG CORP
- Filing Date
- 2025-01-16
- Publication Date
- 2026-07-29
AI Technical Summary
Existing surface inspection methods, such as those using cameras and lasers, are susceptible to external influences and lack precision, especially when inspecting non-roll or non-crankshaft shaped objects, and are often expensive.
A surface anomaly detection device with a roller portion equipped with pressure sensors that rolls over the object, analyzing pressure values to generate an image of the pressure distribution, utilizing machine learning and image recognition to detect surface abnormalities.
Provides high-precision, cost-effective surface inspection that is less affected by external factors, capable of accurately identifying surface anomalies like scratches.
Smart Images
Figure 2026122632000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a surface defect detection device, a surface defect detection method, and a surface defect detection program.
Background Art
[0002] When manufacturing and inspecting automotive parts and the like, the surface quality of aluminum products and the like is directly related to the reliability and appearance of the products, so high-precision inspection is required. Currently, inspection using a camera is common, but there are many factors that lead to misjudgment, such as light reflection, dirt on the surface of the inspection object, polishing scratches, etc., and the ambient light around lighting, factories, etc. significantly affects the camera image, so it often lacks reliability. Inspection using a laser is more expensive than inspection using a camera, and there is a risk of reduced accuracy for specific surface shapes.
[0003] As a method for inspecting surface quality, there is a physical inspection method using a pressure sensor (see, for example, Patent Documents 1 and 2). Patent Document 1 discloses a technique in which a plate-shaped contact member provided with a sensor is brought into contact with a roll-shaped inspection object, and a convex state is detected from a pressure change. Patent Document 2 discloses a technique in which a measuring element is brought into contact with a crankshaft to determine the presence or absence of scratches on the surface of the inspection location.
Prior Art Documents
Patent Documents
[0004]
Patent Document 1
Patent Document 2
Summary of the Invention
Problems to be Solved by the Invention
[0005] However, the technology disclosed in Patent Document 1 has problems such as being limited to objects with a roll shape, and the technology disclosed in Patent Document 2 has problems such as being limited to objects with a specific shape, such as crankshafts.
[0006] In view of the above problems, the present invention aims to provide a surface anomaly detection device, a surface anomaly detection method, and a surface anomaly detection program that are less susceptible to external influences, inexpensive, and capable of high-precision surface inspection. [Means for solving the problem]
[0007] A first aspect of the present invention is a surface abnormality detection device comprising: a plurality of pressure sensors arranged on the surface of a roller portion; a holding portion that holds the roller portion and causes it to roll in contact with an object at a predetermined pressure; and an analysis portion that analyzes the pressure values detected by the pressure sensors, wherein the analysis portion includes an acquisition portion that acquires the pressure values detected by the pressure sensors and an image generation portion that generates an image showing the pressure distribution from the surface of the object based on the fluctuations in the pressure values.
[0008] In the first embodiment of the present invention, the pressure sensor may be arranged linearly in the longitudinal direction of the roller on the body surface of the roller.
[0009] In a first embodiment of the present invention, the analysis unit may further include a detection unit that detects abnormalities on the surface of an object based on fluctuations in pressure values.
[0010] In a first embodiment of the present invention, the detection unit may determine that the pressure sensor has come into contact with the surface of the object when the pressure value acquired by the acquisition unit changes, and may start detecting an abnormality.
[0011] In a first embodiment of the present invention, the holding portion may roll the roller portion in different directions when it is brought into contact with the object and rolled.
[0012] In a first embodiment of the present invention, the detection unit may further include a training data generation unit that generates training data using a pressure distribution image to which the surface state of an object is tagged; a learning unit that performs deep learning using the training data to construct a surface state learning model; a classification unit that classifies the surface state of an object from the pressure distribution image by image recognition using the surface state learning model; and a determination unit that determines whether there is an abnormality based on the surface state of the object.
[0013] A second aspect of the present invention relates to a surface anomaly detection device comprising a plurality of pressure sensors arranged on the surface of a roller portion, a holding portion that holds the roller portion and causes it to roll in contact with an object at a predetermined pressure, and an analysis portion that analyzes the pressure values detected by the pressure sensors, wherein the device comprises an acquisition step of acquiring the pressure values detected by the pressure sensors, and an image generation step of generating an image showing the pressure distribution from the surface of an object based on the fluctuations in the pressure values.
[0014] A third aspect of the present invention is a surface anomaly detection program, comprising a surface anomaly detection device comprising a plurality of pressure sensors arranged on the surface of a roller portion, a holding portion that holds the roller portion and causes it to roll in contact with an object at a predetermined pressure, and an analysis portion that analyzes the pressure values detected by the pressure sensors, wherein the computer implements an acquisition function that acquires the pressure values detected by the pressure sensors, and an image generation function that generates an image showing the pressure distribution from the surface of the object based on the fluctuations in the pressure values.
[0015] According to the present invention, a surface anomaly detection device, a surface anomaly detection method, and a surface anomaly detection program are provided, which are less susceptible to external influences, inexpensive, and capable of accurate surface inspection. The device comprises a plurality of pressure sensors arranged on the surface of a roller, a holding unit that holds the roller and causes it to roll against an object at a predetermined pressure, and an analysis unit that analyzes the pressure values detected by the pressure sensors. The analysis unit comprises an acquisition unit that acquires the pressure values detected by the pressure sensors and an image generation unit that generates an image showing the pressure distribution from the surface of the object based on the fluctuations in the pressure values. [Brief explanation of the drawing]
[0016] [Figure 1] This is a schematic diagram illustrating an example of a surface anomaly detection device according to an embodiment of the present invention. [Figure 2] This figure shows the roll portion of the surface abnormality detection device according to the embodiment being moved in the rotatable direction. [Figure 3] This figure illustrates the pressure data obtained when the roll portion of the surface abnormality detection device according to the embodiment is rolled in the rotatable direction. [Figure 4] This block diagram shows an example of the configuration and function of the analysis unit of the surface anomaly detection device according to an embodiment. [Figure 5] This is a schematic diagram of an example in which the holding part of the surface anomaly detection device according to the embodiment is fixed to the tip of a robot arm. [Figure 6] This is a first flowchart illustrating the surface anomaly detection method according to the embodiment. [Figure 7] This is a second flowchart illustrating the surface anomaly detection method according to the embodiment. [Modes for carrying out the invention]
[0017] Next, embodiments of the present invention will be described with reference to the drawings. In the drawings of the embodiments, identical or similar parts are denoted by the same or similar reference numerals. However, it should be noted that the drawings are schematic and the relationships with planar dimensions etc. may differ from those in reality. Therefore, specific dimensions should be determined by referring to the following explanation. Furthermore, it goes without saying that there are parts in the drawings where the relationships and ratios of dimensions differ from those of other parts.
[0018] Furthermore, the embodiments illustrate devices and methods for embodying the technical idea of the present invention, and the technical idea of the present invention does not specify the configuration, arrangement, layout, etc. of each component as follows. The technical idea of the present invention can be variously modified within the technical scope defined by the claims described in the claims.
[0019] (Embodiment) An example of a surface abnormality detection device according to an embodiment of the present invention is shown in FIG. 1. The surface abnormality detection device 10 shown in FIG. 1 is composed of a roller unit 11 and an analysis unit 13.
[0020] The roller unit 11 is composed of a holding unit 12, a cylindrical roll unit 111, and a shaft unit 112. A plurality of pressure sensors 113 are arranged on the surface of the roll unit 111. The holding unit 12 is connected to the shaft unit 112 of the roller unit 11 so that the roll unit 111 can rotate.
[0021] As shown in FIG. 1, the holding unit 12 abuts the roll unit 111 against the object 14 with a predetermined pressure P and rolls the roll unit 111 in a rotatable direction. The roll unit 111 moves while rotating from position A to position B in the right direction in the plane of the paper in FIG. 1.
[0022] The plurality of pressure sensors 113 arranged on the surface of the roll unit 111 are connected to the analysis unit 13, and the pressure data detected by the plurality of pressure sensors 113 is transmitted to the analysis unit 13. The analysis unit 13 generates an image showing the pressure distribution on the surface of the object 14 based on the pressure data detected by the plurality of pressure sensors 113.
[0023] When the holding unit 12 abuts the roll unit 111 against the object 14 with a predetermined pressure, the plurality of pressure sensors 113 arranged on the surface of the roll unit 111 detect the pressure received from the object 14.
[0024] The multiple pressure sensors 113 arranged on the surface of the roll section 111 may be, for example, sheet-shaped pressure sensors. Alternatively, for example, the multiple pressure sensors 113 may be arranged linearly along the longitudinal direction of the roll section 111 on the body surface of the roll section 111.
[0025] The multiple pressure sensors 113 arranged on the surface of the roll section 111 may be pressure sensors that detect the pressure applied to the pressure sensor 113 based on the change in the electrical resistance value between multiple conductors arranged inside the pressure sensor 113 caused by the pressure applied to the pressure sensor 113.
[0026] Figure 2 shows the roll section 111 being rolled in the rotatable direction while in contact with the object 14 at a predetermined pressure, and Figure 3 shows the pressure data obtained when the roll section 111 is rolled as shown in Figure 2. In Figure 2, the analysis unit 13 is omitted. As an example, multiple pressure sensors 113 are assumed to be arranged linearly in the longitudinal direction of the roll section 111 on the body surface of the roll section 111. In Figure 3, multiple linear pressure sensors 22a, 22b, 22c, ... are arranged on the roll section 111. Figure 3 shows a cross-section 22 of the roll section 111, showing how multiple linear pressure sensors 22a, 22b, 22c, ... are arranged in the longitudinal direction of the surface of the roll section 111.
[0027] When the roll section 111 is at position C, the linear pressure sensor 22a is in contact with the object 14, and linear pressure data 21a is obtained by the pressure sensor 22a. When the roll section 111 is rolled to position D, the linear pressure sensor 22b is in contact with the object 14, and linear pressure data 21b is obtained by the pressure sensor 22b. When the roll section 111 is rolled to position E, the linear pressure sensor 22c is in contact with the object 14, and linear pressure data 21c is obtained by the pressure sensor 22c. When the roll section 111 is rolled to position F, linear pressure data 21a, 21b, 21c, ..., 21n are obtained by the pressure sensors 22a, 22b, 22c, ..., 22n. The linear pressure data 21a, 21b, 21c, ..., 21n represent the pressure data obtained by the pressure sensors 22a, 22b, 22c, ..., 22n as viewed from the z-axis direction in Figure 2.
[0028] Here, a scratch 15 is present on the surface of the object 14. When the roll section 111 is in contact with the object 14 at a predetermined pressure, the pressure value from the object 14 detected by the pressure sensor 113 when the scratch 15 is present changes compared to the pressure value when the scratch 15 is absent. For example, when the scratch 15 causes the portion of the object 14's surface to be concave, the pressure value from the object 14 detected by the pressure sensor 113 is expected to be smaller than the pressure value when the scratch 15 is absent. Also, when the scratch 15 causes the portion of the object 14's surface to be convex, the pressure value from the object 14 detected by the pressure sensor 113 is expected to be larger than the pressure value when the scratch 15 is absent. Figure 3 shows the pressure data 23a and 24a that have changed due to the scratch 15.
[0029] Pressure data 21a, 21b, 21c, ..., 21n detected by multiple pressure sensors 113 are transmitted to the analysis unit 13. Based on the pressure data 21a, 21b, 21c, ..., 21n, the analysis unit 13 generates an image 25 showing the pressure distribution on the surface of the object 14. In the image 25, changes in the pressure distribution 26 can be seen at the locations corresponding to the scratches 15. As shown in the figure, the presence or absence of scratches 15 on the surface of the object 14, as well as the size and shape of the scratches 15, can be identified from the image 25 showing the pressure distribution. Note that, in order to improve measurement accuracy, the roll unit 111 may be rotated multiple times in the same direction, or it may be rolled in a different direction.
[0030] Figure 4 shows an example of the configuration and functions of the analysis unit 13 according to this embodiment. The analysis unit 13 shown in Figure 4 includes a CPU 41 for executing various calculations, a ROM 42 for storing processing programs, a RAM 43 for storing data, a storage unit 44 for storing various data and calculation results, an I / O (input / output interface) 45, a display unit 46, an input unit 47, and the like.
[0031] I / O45 is an interface, buffer, etc., for communication (transmitting and receiving).
[0032] The analysis unit 13 according to this embodiment may also be connected to an input keyboard, mouse, or the like.
[0033] The analysis unit 13 is a so-called computer, and can be various types of electronic computing devices (computational resources) such as mobile terminals, personal computers (PCs), mainframes, workstations, and cloud computing systems.
[0034] Furthermore, the block diagram in Figure 4 shows the functional units within the CPU 41. When each functional unit of the CPU 41 is implemented by software, the CPU 41 implements these functions by executing instructions from the program, which is the software that realizes each function. Specifically, it includes an acquisition unit 411, an image generation unit 412, and so on.
[0035] The acquisition unit 411 acquires pressure data detected by multiple pressure sensors 113.
[0036] The image generation unit 412 generates an image showing the pressure distribution on the surface of the object 14 based on the pressure data detected by the multiple pressure sensors 113.
[0037] The multiple pressure sensors 113 may be configured such that only the pressure sensor in contact with the surface of the object 14 measures the pressure value and transmits the pressure data to the acquisition unit 411.
[0038] The multiple pressure sensors 113 may be configured to measure pressure values at both the pressure sensor in contact with the surface of the object 14 and the pressure sensor not in contact with the surface of the object 14, and transmit the pressure data to the acquisition unit 411. In this case, the image generation unit 412 may determine that only the pressure values from the received pressure data that have changed from the pressure values when the roll unit 111 is not in contact with the surface of the object 14 are the pressure values from the surface of the object 14, and generate an image based only on the pressure values from the surface of the object 14.
[0039] As an example, as shown in Figure 5, the holding part 12 may be fixed to the tip 52 of the robot arm 51. In this case, the analysis unit 13 may further include a control unit, and the object 14 may be fixed, and the control unit may control the robot arm 51 so that the roll part 111 rolls from one predetermined position to another predetermined position on the surface of the object 14 while contacting the object 14 with a predetermined pressure P.
[0040] Alternatively, as shown in Figure 5, the holding unit 12 may be fixed to the tip 52 of the robot arm 51, and the analysis unit 13 may further include a control unit, the object 14 may not be fixed, and the robot arm 51 may be controlled so that the roll unit 111 contacts the object 14 with a predetermined pressure P, and furthermore, the object 14 may be moved so that the roll unit 111 rolls from one predetermined position on the surface of the object 14 to another predetermined position.
[0041] The CPU 41 may further include a detection unit 413. The detection unit 413 detects changes in pressure values based on pressure data acquired by the acquisition unit 411 or images generated by the image generation unit 412, and determines whether or not there is an abnormality on the surface of the object 14 based on the detected change in pressure values.
[0042] In the examples shown in Figures 2 and 3, the detection unit 413 detects the pressure data 23a and 24a that have been altered by the scratch 15 as changes in pressure values, and determines that there is an abnormality on the surface of the object 14 based on the detected pressure data 23a and 24a.
[0043] When the detection unit 413 detects pressure data 23a and 24a that have changed due to the damage 15, it may detect pressure data 23a and 24a based on the difference in pressure values between pressure data 23b and 24b, which are the parts of pressure data 23 and 24 measured at the same time as pressure data 23a and 24a, excluding pressure data 23a and 24a. The detection unit 413 may also detect pressure data 23a and 24a when the difference in pressure values between pressure data 23a and pressure data 23b, and the difference in pressure values between pressure data 24a and pressure data 24b, each exceed a predetermined first threshold.
[0044] Alternatively, when the detection unit 413 detects pressure data 23a and 24a that have changed due to the damage 15, it may detect pressure data 23a and 24a based on the difference in pressure values between pressure data measured at a different time than when pressure data 23a and 24a were measured, for example, pressure data 21a, 21b, and 21c. The detection unit 413 may also detect pressure data 23a and 24a when the difference in pressure values between pressure data 23a and 24a and pressure data measured at a different time than when pressure data 23a and 24a were measured exceeds a predetermined second threshold.
[0045] After detecting the pressure data 23a and 24a as described above, the detection unit 413 may determine that an abnormality exists on the surface of the object 14 when it exceeds a predetermined third threshold.
[0046] The detection unit 413 may determine whether or not there is an abnormality on the surface of the object 14 by machine learning and image recognition. The detection unit 413 may include a training data generation unit, a learning unit, a classification unit, and a determination unit.
[0047] The training data generation unit may generate training data using a pressure distribution image to which the surface condition of the object is tagged. The pressure distribution image is a pressure distribution image of the surface of the object generated using the surface anomaly detection device according to this embodiment. The surface condition of the object tagged to the pressure distribution image may be the presence or absence of an anomaly, or it may be the type of anomaly, such as a scratch.
[0048] The learning unit may perform deep learning using training data to construct a surface state learning model.
[0049] The classification unit may classify the surface state of an object from a pressure distribution image using image recognition based on a surface state learning model.
[0050] The determination unit may determine whether there is an abnormality based on the surface condition of the object.
[0051] Furthermore, the detection unit 413 may determine that the pressure sensor 113 has come into contact with the surface of the object 14 when the pressure value acquired by the acquisition unit 411 changes, and may begin detecting an abnormality.
[0052] Depending on the shape and type of abnormality on the surface of the object 14, such as the shape of a scratch, the detected pressure value may differ depending on whether the roll part 111 is rolled in a predetermined direction on the surface of the object 14 or in a direction other than the predetermined direction on the surface of the object 14, when the roll part 111 is rolled in the rotatable direction while in contact with the object 14 at a predetermined pressure. As a result, for example, even if there is an abnormality on the surface of the object 14, depending on the shape and type of the abnormality, the detection unit 413 may not detect the abnormality if the roll part 111 is rolled in only one direction while in contact with the object 14 at a predetermined pressure, but may detect the abnormality when the roll part 111 is rolled in a different direction.
[0053] Therefore, when the holding unit 12 brings the roll unit 111 into contact with the object 14 and rolls it, it may roll the roll unit 111 in a predetermined direction, and then roll the roll unit 111 in a different direction from the predetermined direction. The image generation unit 412 may generate multiple pressure distribution images based on the pressure values obtained by rolling the roll unit 111 in different directions. The detection unit 413 may detect abnormalities on the surface of the object 14 based on the multiple pressure distribution images generated based on the pressure values obtained by rolling the roll unit 111 in different directions.
[0054] If the holding unit 12 is fixed to the tip of the robot arm 51, the control unit may control the robot arm 51 so that when the holding unit 12 brings the roll unit 111 into contact with the object 14 and rolls it, the roll unit 111 is first rolled in a predetermined direction, and then rolled in a different direction from the predetermined direction.
[0055] Next, the surface anomaly detection method according to the embodiment will be described with reference to Figure 6. Figure 6 is a flowchart for explaining the surface anomaly detection method according to the embodiment. The surface anomaly detection method according to the embodiment is performed in a surface anomaly detection device that includes a plurality of pressure sensors arranged on the surface of a roller, a holding unit that holds the roller and rolls it in contact with an object at a predetermined pressure, and an analysis unit that analyzes the pressure values detected by the pressure sensors.
[0056] In step ST601, the pressure value detected by the pressure sensor is acquired (acquisition step).
[0057] In step ST602, an image showing the pressure distribution from the surface of the object is generated based on the fluctuation of the pressure value (image generation step).
[0058] The procedure for determining whether or not there is an abnormality on the surface of the object 14 by the detection unit 413 of the surface abnormality detection device according to the embodiment, using machine learning and image recognition, will be explained with reference to Figure 7. Figure 7 is a flowchart for explaining the operation of the detection unit 413 of the surface abnormality detection device according to the embodiment. Here, the detection unit 413 is assumed to include a training data generation unit, a learning unit, a classification unit, and a determination unit.
[0059] In step ST701, the training data generation unit generates training data using a pressure distribution image to which the surface state of the object is tagged (training data generation step).
[0060] In step ST702, the learning unit performs deep learning using training data to construct a surface state learning model (learning step).
[0061] In step ST703, the classification unit classifies the surface state of the object from the pressure distribution image using image recognition with a surface state learning model (classification step).
[0062] In step ST704, the determination unit determines whether there is an abnormality based on the surface condition of the object (determination step).
[0063] As described above, the surface anomaly detection device according to this embodiment makes it possible to provide a surface anomaly detection device that is less susceptible to external influences, inexpensive, and capable of accurate surface inspection.
[0064] As stated above, the present invention naturally includes various embodiments and the like that are not described herein. Therefore, the technical scope of the present invention is determined solely by the inventive features relating to the claims that are reasonable based on the above description. [Explanation of symbols]
[0065] 10. Surface anomaly detection device 11 Roller section 12 Holding part 13 Analysis Department 14. Object 15 wounds 111 Roll section 112 Shaft 113, 22a, 22b, 22c Pressure Sensors 21a, 21b, 21c, ..., 21n pressure data 41 CPU 42 ROM 43 RAM 44 Storage section 45 I / O (Input / Output Interface) 46 Display section 47 Input section 411 Acquisition Department 412 Image generation unit 413 Detection Unit 51 Robot Arm 52 Tip
Claims
1. Multiple pressure sensors are placed on the surface of the roller section, A holding part that holds the roller part and causes it to roll when brought into contact with an object at a predetermined pressure, A surface anomaly detection device comprising an analysis unit that analyzes the pressure value detected by the pressure sensor, The aforementioned analysis unit, An acquisition unit that acquires the pressure value detected by the pressure sensor, An image generation unit generates a pressure distribution image showing the pressure distribution from the surface of the object based on the fluctuation of the pressure value. A surface anomaly detection device characterized by including [a certain component].
2. The surface abnormality detection device according to claim 1, characterized in that the pressure sensor is arranged linearly in the longitudinal direction of the roller on the body surface of the roller.
3. The surface abnormality detection device according to claim 1, characterized in that the analysis unit further comprises a detection unit that detects abnormalities on the surface of the object based on fluctuations in the pressure value.
4. The surface abnormality detection device according to claim 3, characterized in that the detection unit determines that the pressure sensor has come into contact with the surface of the object when the pressure value acquired by the acquisition unit changes, and starts detecting an abnormality.
5. The surface abnormality detection device according to claim 1, wherein the holding portion rolls the roller portion in different directions when the roller portion is brought into contact with the object and rolled.
6. The detection unit is A training data generation unit that generates training data using the pressure distribution image to which the surface state of the object is assigned as a tag, A learning unit that performs deep learning using the aforementioned training data and constructs a surface state learning model, A classification unit that classifies the surface state of the object from the pressure distribution image using the surface state learning model described above, A determination unit that determines whether there is an abnormality based on the surface condition of the object. The surface abnormality detection device according to claim 3, further comprising:
7. Multiple pressure sensors are placed on the surface of the roller section, A holding part that holds the roller part and causes it to roll when brought into contact with an object at a predetermined pressure, A surface abnormality detection device comprising an analysis unit that analyzes the pressure value detected by the pressure sensor, An acquisition step of acquiring the pressure value detected by the pressure sensor, An image generation step of generating an image showing the pressure distribution from the surface of the object based on the fluctuation of the pressure value. A surface anomaly detection method characterized by comprising the following:
8. Multiple pressure sensors are placed on the surface of the roller section, A holding part that holds the roller part and causes it to roll when brought into contact with an object at a predetermined pressure, A surface anomaly detection device comprising an analysis unit that analyzes the pressure value detected by the pressure sensor, wherein a computer, An acquisition function that acquires the pressure value detected by the pressure sensor, An image generation function that generates an image showing the pressure distribution from the surface of the object based on the fluctuation of the aforementioned pressure value. A surface anomaly detection program characterized by achieving this.