Inspection device and inspection method

Abstract

To provide an inspection device and an inspection method capable of easily inspecting whether or not an inspection object is correctly arranged at an arrangement place.SOLUTION: An inspection device 1 includes imaging means 10 for capturing an image of a region including an arrangement place and a mark object to be a mark for detecting the arrangement place, mark detection means 20 for detecting the mark object from a captured image, arrangement place detection means 30 for detecting the arrangement place from the captured image based on a relative positional relationship of the arrangement place to the mark object, and determination means 40 for determining whether or not an inspection object is correctly arranged from a determination image of the arrangement place.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an inspection apparatus and an inspection method for inspecting whether an object to be inspected is correctly disposed at a disposed location.

Background Art

[0002] A cylinder block of an engine has a bearing portion that rotatably supports a journal portion of a crankshaft about an axis. The bearing portion that supports the main journal portion also has a function of receiving a thrust force acting on the crankshaft and positioning an axial position, and grooves are formed on both side surfaces of the bearing portion so that thrust metals are inserted between the bearing portion and the crank arm portion. Usually, a semi-circular recess corresponding to the diameter dimension of the journal portion is provided in the bearing portion of the cylinder block. In a state where the journal portion is inserted into the bearing portion, a cap provided with a semi-circular recess is combined, and the journal portion is configured to be rotatably supported.

[0003] Further, the thrust metal is also divided into two semi-circular arc shapes. In a state where the journal portion is inserted into the bearing portion of the cylinder block, one semi-circular arc-shaped thrust metal is inserted into a gap formed by a groove between the side surface of the bearing portion and the crank arm portion, and the other thrust metal is assembled when the cap is assembled. The thrust metal has different configurations on the front surface and the back surface. If the front and back surfaces of the thrust metal are assembled in the wrong direction or if the thrust metal is forgotten to be assembled, it may cause engine damage and lead to serious problems. Therefore, after the thrust metal is assembled, it is inspected whether the thrust metal is correctly disposed.

[0004] The inspection of thrust metals has conventionally been carried out, for example, visually by workers. However, the area where the thrust metal can be confirmed from the surface is small, and it is difficult to see, so it becomes a worrying task and there is a problem that the burden is large. In addition, a method of detecting whether the thrust metal is correctly arranged by pattern matching can be considered, but since there is variation in the image position of the thrust metal in the image of the inspection object taken, it is difficult to read at fixed coordinates. Moreover, due to the adhesion of oil, there is variation in the appearance, and there is a problem that it is difficult to correctly determine whether it is correctly arranged or not.

[0005] In Patent Document 1, a method for detecting misassembly of a seal ring having front and back directions of contact with the inner peripheral surface of a circular hole of one work piece, which is assembled to the other work piece fitted into a circular hole provided in one work piece, with respect to the other work piece is described. In this misassembly inspection method, after fitting the other work piece with the seal ring assembled to the detection circular hole of a jig having a detection circular hole set to have an inner diameter larger than the circular hole of one work piece by a predetermined amount into the detection circular hole of the jig, air is continuously supplied to the closed space formed between the detection circular hole of the jig and the other work piece to pressurize the closed space to a predetermined pressure. While maintaining the pressurized state, when the air flow rate leaking from between the inner peripheral surface of the detection circular hole and the seal ring exceeds the standard flow rate, it is determined as misassembly of the seal ring. According to this misassembly inspection method, although misassembly of the seal ring having front and back directions can be detected, since it is necessary to supply air and detect the flow rate, the device becomes complicated and it is difficult to apply.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] The present invention has been made based on such problems, and an object thereof is to provide an inspection apparatus and an inspection method capable of easily inspecting whether an object to be inspected is correctly disposed at a disposed position.

Means for Solving the Problems

[0008] The inspection apparatus of the present invention inspects whether an object to be inspected is correctly disposed at a disposed position, and includes a photographing means for photographing an area including the disposed position and a target object located around the disposed position and serving as a mark when detecting the disposed position, a mark detecting means for detecting the target object from the photographed image obtained by the photographing means, a disposed position detecting means for detecting the disposed position from the photographed image based on the position of the target object detected by the mark detecting means and the relative positional relationship between the disposed position and the target object, and a discriminating means for discriminating whether the object to be inspected is correctly disposed or not from the discrimination image of the disposed position detected by the disposed position detecting means.

[0009] The inspection method of the present invention inspects whether an object to be inspected is correctly disposed at a disposed position, and includes a photographing procedure for photographing an area including the disposed position and a target object located around the disposed position and serving as a mark when detecting the disposed position, a mark detecting procedure for detecting the target object from the photographed image obtained by the photographing procedure, a disposed position detecting procedure for detecting the disposed position from the photographed image based on the position of the target object detected by the mark detecting procedure and the relative positional relationship between the disposed position and the target object, and a discriminating procedure for discriminating whether the object to be inspected is correctly disposed or not from the discrimination image of the disposed position detected by the disposed position detecting procedure.

Effects of the Invention

[0010] According to the present invention, a target object with a mark is detected from a captured image, and a placement location is detected from the captured image based on the position of the target object with the mark and the relative positional relationship between the placement location with respect to the target object with the mark. Therefore, even if the size of the placement location is small, there is variation in the position of the placement location in the captured image, or there is variation in the way it appears due to dirt or the like, the placement location can be easily detected. Thus, it is possible to easily determine whether the object to be inspected is correctly placed from the discrimination image of the placement location.

[0011] Further, based on an image learning model that has learned the characteristics when the object to be inspected is correctly placed and when it is not correctly placed, if it is determined whether the object to be inspected is correctly placed, it can be determined more easily and with high accuracy.

[0012] Furthermore, having an image learning model generation means for generating an image learning model, and in the image learning model generation means, if a processed image obtained by processing a learning image when the object to be inspected is correctly placed is included in the learning image when the object to be inspected is not correctly placed, a learning image when the object to be inspected is not correctly placed can be easily prepared, and the discrimination accuracy can be easily improved.

Brief Description of the Drawings

[0013]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Embodiments for Carrying Out the Invention

[0014] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0015] FIG. 1 shows the configuration of an inspection apparatus 1 according to an embodiment of the present invention. This inspection apparatus 1 inspects whether an inspection object is correctly installed at the installation location. This inspection apparatus 1 can be preferably used, for example, when inspecting whether a thrust metal M1 is correctly installed at the installation location as an inspection object. In this embodiment, in the manufacture of an engine, the case of inspecting whether the thrust metal M1 is correctly installed at the installation location will be described as an example.

[0016] The inspection apparatus 1 includes, for example, imaging means 10 for imaging a region including the installation location of the thrust metal M1 and a mark object M2 that is located around the installation location and serves as a mark when detecting the installation location, and from the captured image obtained by the imaging means 10, a mark detection means 20 for detecting the mark object M2, based on the position of the mark object M2 detected by the mark detection means 20 and the relative positional relationship of the installation location with respect to the mark object M2, an installation location detection means 30 for detecting the installation location from the captured image, and a discrimination means 40 for discriminating whether the inspection object is correctly installed from the discrimination image of the installation location detected by the installation location detection means 30. Further, the inspection apparatus 1 preferably includes, for example, a display means 50 for displaying the discrimination result of the discrimination means 40.

[0017] The imaging means 10 is constituted by, for example, a camera such as a CCD camera, and is arranged to image a region including the location where the thrust metal M1 is disposed and the target object M2 located around the location where the thrust metal M1 is disposed in the engine manufacturing line. The target object M2 is located around the location where the thrust metal M1 is disposed, and may be anything as long as it can be easily detected. For example, a fastening hole is preferable. Further, it is preferable that the imaging means 10 images the locations where a pair of thrust metals M1 provided on both side surfaces of the bearing portion are included in one region. That is, it is preferable that one captured image includes the locations where the pair of thrust metals M1 are disposed and the target object M2. This is because the pair of thrust metals M1 can be inspected simultaneously.

[0018] The target detection means 20 can be constituted by, for example, a computer, and is configured to function as the discrimination means 20 by executing a program. The target detection means 20 is preferably connected to the imaging means 10, and is configured to detect the target object M2 from the captured image input from the imaging means 10 and specify the position (for example, the center position) of the target object M2 in coordinates. The coordinates can be represented, for example, by two-dimensional positions (x, y) with a certain point in the captured image as the origin. The detection of the target object M2 in the captured image can use known object detection techniques, and is preferably performed by image processing such as circle detection or pattern matching.

[0019] In the captured image, for example, when the imaging means 10 and the target object M2 cannot be directly opposed under constraint conditions such as a process layout, and when the target object M2 appears as an ellipse in the captured image, it is preferable to perform a processing operation to stretch the captured image in the minor axis direction so that the target object M2 becomes a circle. This is to enable easy detection. FIG. 2 shows an example of a captured image captured by the imaging means 11, FIG. 3 shows an example of a captured image stretched so that the target object M2 becomes a circle, and FIG. 4 conceptually shows the position of the target object M2 detected by the mark detection means 20 in the stretched captured image. In FIG. 4, the detected circle is indicated by a thick broken line, and its center position is indicated by a white circle.

[0020] The installation location detection means 30 can be constituted by, for example, a computer, and is configured to function as the installation location detection means 30 by executing a program. The installation location detection means 30 is connected to the imaging means 10 and the mark detection means 20, and is preferably configured to specify the position of the installation location in coordinates from the captured image based on the position of the target object M2 detected by the mark detection means 20 and the relative positional relationship of the installation location with respect to the target object M2. The relative positional relationship of the installation location with respect to the target object M2 is preferably obtained, for example, in advance from a preparatory image obtained by capturing an area including the installation location and the target object M2 by the imaging means 10. FIG. 5 conceptually shows the position of the installation location detected by the installation location detection means 30 in the stretched captured image. In FIG. 5, the detected areas of the pair of installation locations are indicated by thick dotted lines, and as in FIG. 4, the circle detected by the mark detection means 20 is indicated by a thick broken line, and its center position is indicated by a white circle.

[0021] The discrimination means 40 can be configured by, for example, a computer and is configured to function as the discrimination means 40 by executing a program. The discrimination means 40 is connected to, for example, the imaging means 10 and the installation location detection means 30, and preferably configured to use the image of the installation location detected by the installation location detection means 30 as a discrimination image and discriminate from this discrimination image whether the inspection object is correctly installed. Specifically, for example, the discrimination means 40 preferably discriminates by an image learning model 41 that executes machine learning using a learning image that is a positive example of the discrimination image and learns the characteristics when the inspection object is correctly installed and when it is not correctly installed.

[0022] For example, the discrimination means 40 preferably includes an image learning model 41, a positive / negative determination means 42 that inputs the discrimination image to the image learning model 41 and determines whether the inspection object is correctly installed based on the output value obtained thereby, and an image learning model generation means 43 that generates the image learning model 41. The image learning model generation means 43 preferably prepares, for example, a plurality of learning images for each of the cases where the inspection object is correctly installed and where the inspection object is not correctly installed as positive examples of the discrimination image, extracts the characteristics of each learning image by deep learning (Deep learning), and is configured to learn in association with the installation state of the inspection object. It is preferable to use a convolutional neural network (CNN) for deep learning.

[0023] Each learning image can be obtained by photographing, for example, the case where the thrust metal M1 is correctly arranged, the case where the front and back of the thrust metal M1 are arranged in reverse, and the case where the thrust metal M1 is not arranged, respectively, by the photographing means 10. Among these, the images in the case where the front and back of the thrust metal M1 are arranged in reverse and the case where the thrust metal M1 is not arranged correspond to the case where the thrust metal M1 is not correctly arranged. Further, it is preferable that the learning image in the case where the inspection object is not correctly arranged includes a processed image obtained by processing the learning image in the case where the inspection object is correctly arranged. Examples of the processed image include an inverted image obtained by inverting the learning image in the case where the inspection object is correctly arranged. This is because it is difficult to prepare by photographing with the photographing means 10 in the case where the inspection object is not correctly arranged, whereas it can be easily prepared by inverting the learning image in the case where the inspection object is correctly arranged.

[0024] FIG. 6 shows an example of the hardware configuration of the mark detection means 20, the arrangement position detection means 30, and the discrimination means 40. The mark detection means 20, the arrangement position detection means 30, and the discrimination means 40 include, for example, a CPU (Central Processing Unit) 61, a ROM (Read Only Memory) 62, a RAM (Random Access Memory) 63, an HDD (hard disk drive) 64, and an operation interface (operation I / F) 65. The CPU 61 executes various processes according to various programs recorded in the ROM 62 or various programs loaded from the HDD 64 to the RAM 63. The RAM 63 appropriately stores data and the like necessary for the CPU 61 to execute various processes. The HDD 64 stores various data.

[0025] This inspection apparatus 1 is used, for example, as follows. FIG. 7 shows the flow of the inspection method using the inspection apparatus 1. First, as a preparation procedure, an image learning model 41 is generated by the image learning model generation means 43 (preparation procedure; step S110).

[0026] In the preparation procedure, first, for example, when the object to be inspected is correctly arranged and when the object to be inspected is not correctly arranged, the imaging means 10 captures images respectively to prepare learning images that are positive examples. Specifically, for example, when the thrust metal M1 is correctly arranged, when the front and back of the thrust metal M1 are arranged in reverse, and when the thrust metal M1 is not arranged, the imaging means 10 captures images respectively to prepare each learning image that is a positive example. Further, it is preferable to prepare, as learning images when the object to be inspected is not correctly arranged, processed images obtained by processing the learning images when the object to be inspected is correctly arranged, for example, inverted images obtained by inversion. Next, for example, features of the learning images are extracted by deep learning using the learning images, and learning is performed in association with the arrangement state of the object to be inspected.

[0027] After the preparation procedure (step S110), for example, in the manufacture of the engine, it is inspected whether the thrust metal M1 is correctly arranged at the installation location. First, for example, the imaging means 10 captures an area including the installation location of the thrust metal M1 and a target object M2 that is located around the installation location and serves as a mark for detecting the installation location (imaging procedure; step S120). Next, the mark detection means 20 detects the target object M2 from the captured image obtained by the imaging means 10 (mark detection procedure; step S130). For example, the target object M2 is detected by pattern matching, and the position of the target object M2 is specified by coordinates.

[0028] Subsequently, the installation location detection means 30 detects the installation location from the captured image based on the position of the target object M2 detected by the mark detection means 20 and the relative positional relationship between the installation location and the target object M2 (installation location detection procedure; S140). For example, based on the relative positional relationship between the installation location and the target object M2 obtained in advance, the coordinates of the installation location are specified from the coordinates of the target object M2.

[0029] Next, the discrimination means 40 discriminates whether or not the inspection object is correctly arranged from the discrimination image of the arrangement location detected by the arrangement location detection means 30 (discrimination procedure; S150). Specifically, for example, an image learning model 41 is used, the discrimination image is input to the image learning model 41, and based on the output value obtained thereby, it is determined whether the inspection object is correctly arranged or not. Thereafter, the display means 50 displays the result discriminated by the discrimination means 40 (display procedure; step S160).

[0030] As described above, according to the present embodiment, the target object M2 is detected from the captured image, and the arrangement location is detected from the captured image based on the position of the target object M2 and the relative positional relationship of the arrangement location with respect to the target object M2. Therefore, even if the size of the arrangement location is small, there is variation in the position of the arrangement location in the captured image, or there is variation in the appearance due to dirt or the like, the arrangement location can be easily detected. Thus, it is possible to easily discriminate whether or not the inspection object is correctly arranged from the discrimination image of the arrangement location.

[0031] In addition, if it is determined whether or not the inspection object is correctly arranged based on an image learning model that has learned the characteristics of the case where the inspection object is correctly arranged and the case where it is not correctly arranged, it can be determined more easily and with higher accuracy.

[0032] Furthermore, if an image learning model generation means for generating an image learning model is provided, and in the image learning model generation means, a processed image obtained by processing a learning image when the inspection object is correctly arranged is included in the learning image when the inspection object is not correctly arranged, a learning image when the inspection object is not correctly arranged can be easily prepared, and the discrimination accuracy can be easily improved.

[0033] The present invention has been described above by way of embodiments, but the present invention is not limited to the above embodiments and can be variously modified. For example, in the above embodiments, each component has been specifically described, but the specific structure and shape of each component may be different, and all of the above-described components may not be provided, or other components may be provided.

[0034] Furthermore, in the above embodiments, the case where the inspection object is the thrust metal M1 has been specifically described as an example, but the present invention can be similarly applied to the case of inspecting whether other inspection objects are correctly arranged at the arranged positions.

[0035] In addition, in the above embodiments, as the processed image obtained by processing the learning image when the inspection object is correctly arranged, an inverted image obtained by inverting the learning image when it is correctly arranged has been described as an example. However, depending on the inspection object, other processing such as rotating the learning image when it is correctly arranged may be performed.

Explanation of Reference Numerals

[0036] 1... inspection device, 10... imaging means, 20... mark detection means, 30... arrangement position detection means, 40... discrimination means, 41... image learning model, 42... correctness determination means, 43... image learning model generation means, 50... display means, 61... CPU, 62... ROM, 63... RAM, 64... HDD, 65... operation interface

Claims

1. An inspection device for inspecting whether an object to be inspected is correctly disposed at a disposal location, comprising: imaging means for imaging an area including the disposal location and a target object that is located around the disposal location and serves as a mark when detecting the disposal location; mark detection means for detecting the target object from the captured image obtained by the imaging means; disposal location detection means for detecting the disposal location from the captured image based on the position of the target object detected by the mark detection means and the relative positional relationship between the disposal location and the target object; discrimination means for discriminating whether the object to be inspected is correctly disposed from the discrimination image of the disposal location detected by the disposal location detection means An inspection device characterized by comprising the above.

2. The inspection device according to claim 1, wherein the discrimination means executes machine learning using a learning image that is a positive example of the discrimination image, and discriminates using an image learning model that has learned the characteristics when the object to be inspected is correctly disposed and when it is not correctly disposed.

3. The discrimination means has image learning model generation means for generating an image learning model, The inspection device according to claim 2, wherein the image learning model generation means includes a processed image obtained by processing a learning image when the object to be inspected is correctly disposed in a learning image when the object to be inspected is not correctly disposed for learning.

4. The inspection device according to any one of claims 1 to 3, wherein the inspection is for whether a thrust metal is correctly disposed at a disposal location.

5. An inspection method for inspecting whether an object to be inspected is correctly disposed at a disposal location, comprising: an imaging procedure for imaging an area including the disposal location and a target object that is located around the disposal location and serves as a mark when detecting the disposal location; a mark detection procedure for detecting the target object from the captured image obtained by the imaging procedure; a disposal location detection procedure for detecting the disposal location from the captured image based on the position of the target object detected by the mark detection procedure and the relative positional relationship between the disposal location and the target object; a discrimination procedure for discriminating whether the object to be inspected is correctly disposed from the discrimination image of the disposal location detected by the disposal location detection procedure An inspection method characterized by including the above.

Images