Image recognition method and device

Abstract

An image recognition method for recognizing and measuring the length of the outer shape of a transparent member (11) in an object (2) having the transparent member (11) disposed on the surface thereof, said image recognition method comprising: irradiating the transparent member (11) with illumination light (15) from an illumination device (5) capable of irradiating the illumination light (15) at each of a plurality of illumination light intensities; the illumination light (15) is irradiated to the object (2) at one illumination light intensity based on color information of the transparent member (11) among the plurality of illumination light intensities and within an illumination angle (theta) capable of total reflection within the transparent member (11), and the object (2) irradiated by the illumination light (15) is imaged. The outer shape of the transparent member (11) is recognized on the basis of the captured information.

Description

Technical Field

[0001] The present invention provides an image recognition method and apparatus that removes glare from the periphery of an object that occurs during recognition by photographing with a camera and realizes external shape recognition of the object regardless of color. Background Art

[0002] In recent years, image recognition technology, which identifies objects from images captured by cameras, has been effectively used in various fields. For example, in the field of product manufacturing processes, it is used for defect detection based on appearance inspection and measurement such as length measurement.

[0003] In order to reliably recognize an object to be recognized, it is necessary to use an illumination device to illuminate the object with illumination light to increase the sensitivity of the camera during imaging.

[0004] However, when a single surface of an object is covered with transparent glass, glare can easily occur on the surface of the object, depending on the color of the glass directly below, illumination light, or surrounding structures. This glare can reduce the accuracy of length measurement when identifying the object using captured images.

[0005] Patent Document 1 proposes a method that addresses these challenges. The background to this method stems from the fact that during the manufacturing process of battery plates, there are often both high-gloss areas made of metal, which is prone to glare, and low-gloss areas made of resin, which is less prone to glare. In this situation, the method in Patent Document 1 employs two separate light sources: diffuse light for the high-gloss areas and parallel light for the low-gloss areas. This allows for surface inspection based on recognition while avoiding glare.

[0006] exist Figure 10 In the present invention, when inspecting the outer casing 102 which is a low-gloss area and the electrode tabs 103a and 103b which are high-gloss areas of a rectangular thin secondary battery 101, parallel light illuminators L1 to L4 which generate parallel light and a diffuse light illuminator Lw which generates diffuse light are arranged around the thin secondary battery 101. The parallel light illuminators L1 to L4 and the diffuse light illuminator Lw are used in combination according to the degree of gloss, thereby enabling photography while suppressing glare.

[0007] Prior art literature

[0008] Patent Literature

[0009] Patent Document 1: Japanese Patent No. 6570977 Summary of the Invention

[0010] An image recognition method according to one embodiment of the present invention is an image recognition method for identifying and measuring the outer shape of an object having a transparent component arranged on its surface. In this image recognition method, an illumination device capable of irradiating illumination light at a plurality of illumination light intensities is used to irradiate the object with the illumination light at one illumination light intensity based on color information of the transparent component among the plurality of illumination light intensities, and within an illumination angle at which total reflection can be achieved within the transparent component. The object illuminated by the illumination light is photographed, and the outer shape of the transparent component is identified based on the photographed information.

[0011] Another embodiment of the present invention relates to an image recognition device for identifying and measuring the outer shape of an object having a transparent component arranged on its surface. The image recognition device comprises: an illumination device capable of irradiating illumination light at a plurality of illumination light intensities, irradiating the object with the illumination light at one illumination light intensity based on color information of the transparent component among the plurality of illumination light intensities, and within an illumination angle at which total reflection can be performed within the transparent component; a camera for photographing the object illuminated by the illumination light; and a calculation unit for identifying the outer shape of the transparent component based on the photographed information. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of an image recognition device in an embodiment of the present invention.

[0013] Figure 2 It is a schematic diagram of an object in an embodiment of the present invention.

[0014] Figure 3 This is a schematic diagram of a recognition system when edge portions are emphasized in an embodiment of the present invention.

[0015] Figure 4 1 and 2 are diagrams showing the results of length measurement and recognition based on the illumination angle in glass during edge-emphasized photography in the embodiment of the present invention.

[0016] Figure 5 This figure shows the recognition results differentiated by the intensity of illumination light irradiated on the center of the glass surface during edge-emphasized photography in the embodiment of the present invention.

[0017] Figure 6 This is a schematic diagram of a recognition system when a face is emphasized in an embodiment of the present invention.

[0018] Figure 7 This figure shows the results of recognition based on the intensity of illumination light irradiated on the center of the glass surface during surface emphasis photography in the embodiment of the present invention.

[0019] Figure 8 It is a diagram showing the design portion, various imaging methods, and length measurement and recognition results in the embodiment of the present invention.

[0020] Figure 9 This is a flowchart of the recognition system in the embodiment of the present invention.

[0021] Figure 10 This is an explanatory diagram schematically showing the layout of components of the surface inspection device described in Patent Document 1. DETAILED DESCRIPTION

[0022] In the invention described in Patent Document 1, diffused light, unlike parallel light, does not travel in a straight line. While this method is effective against glare near the lighting source, it is conceivably difficult to apply to large objects or situations where multiple objects must be identified simultaneously. For example, in recycling processes where a wide variety of objects are discharged, it would be impossible to develop a system tailored to each of these situations. Therefore, considering application to a wide variety of products, it is desirable to build a system capable of identifying objects regardless of the presence or absence of glare.

[0023] The present invention solves the above-mentioned conventional problems and aims to provide an image recognition method and apparatus that achieves recognition using the same lighting device by changing the illumination light intensity according to the color information of the object.

[0024] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0025] (Implementation Method)

[0026] Figure 1 An image recognition device 1 according to the illustrated embodiment recognizes and measures the outer shape of a transparent member 11 disposed on the surface of an object 2. The image recognition device 1 includes at least an illuminator 5, which is an example of an illumination device; a camera 4, which is an example of a video camera; and a length measurement data calculation unit 8, which is an example of a calculation unit.

[0027] The illuminator 5 is arranged obliquely above the object 2, and the illumination light intensity is variable, that is, it can irradiate the illumination light 15 at multiple illumination light intensities, and irradiate the object 2 with the illumination light intensity based on the color information of the transparent component 11 and within the illumination angle θ that can be totally reflected in the transparent component 11.

[0028] The camera 4 is disposed above the object 2 and captures an image of the surface of the object 2 illuminated by the illumination light 15 to acquire image data.

[0029] The length measurement data calculation unit 8 recognizes the outer shape of the transparent member 11 through calculation based on the captured information.

[0030] More specifically, the image recognition device 1 includes a recognition and length measurement unit 6 and a stage 3 . The recognition and length measurement unit 6 includes an image data storage unit 7 and an object processing control unit 9 in addition to a length measurement data calculation unit 8 .

[0031] The captured image data storage unit 7 stores image data captured by the camera 4 .

[0032] The length measurement data calculation unit 8 calculates the length measurement result using the image data stored in the image data storage unit 7 .

[0033] The object processing control unit 9 determines the illumination intensity of the illuminator 5 based on the color information of the transparent member 11 and can control the illumination intensity of the illuminator 5 to change to the determined illumination intensity. In addition, the object processing control unit 9 can also perform color determination processing such as color recognition processing as described later.

[0034] Here, the color information of the transparent part 11 refers to the model information of the object 2, the color of the design part 12 or the base material 13 through the transparent part 11, the color of the bottom surface of the transparent part through the transparent part 11, the color of the transparent part itself, or a color based on any combination of these.

[0035] The object 2 can be placed on the stage 3 .

[0036] Figure 2 Schematic diagram showing, for example, a rectangular object 2. A transparent member 11, for example, a rectangular plate, is adhered to at least one surface of a design portion 12, and a substrate 13, for example, a rectangular plate, is adhered to the surface where the transparent member 11 is not present.

[0037] In addition, although the transparent member 11 alone is not a problem, the rectangular frame-shaped edge portion 10 surrounds the edge of the transparent member 11, so that the outer shape of the transparent member 11 can be made clearer when recognizing and measuring the length.

[0038] When the color information of the transparent member 11 is the color information of the design portion 12 that is transmitted through the transparent member 11, the color information of the design portion 12 includes color information of a high-reflection design 12a, color information of a low-reflection design 12b, and color information of a mirror-finished design 12c. Regarding each design, assuming a reference reflectance of 50% for the wavelength of the light emitted from the illuminator 5, i.e., the illumination light 15, the design portions 12 can be classified into three categories: a high-reflection design 12a is defined as a design portion exhibiting a reflectance of 50% or more and 100% or less and not subjected to mirror finishing; a low-reflection design 12b is defined as a design portion exhibiting a reflectance of 0% or more and less than 50%; and a mirror-finished design portion is defined as a mirror-finished design 12c regardless of reflectance.

[0039] The color information of the high-reflection design 12 a is a color such as white, which is a color showing a reflectivity of 50% or higher with respect to light having a visible wavelength, for example.

[0040] The color information of the low-reflection design 12b is a color that exhibits a reflectivity of less than 50% for light having a visible wavelength, such as black or gray. The reference reflectivity is not limited to 50%.

[0041] The color information of the mirror surface design 12 c is, for example, a color having a property of glaring objects existing around the object 2 and at positions symmetrical to the design portion 12 , like a mirror.

[0042] In the design portion 12 , a design is formed by various printing methods such as offset printing, gravure printing, or screen printing, or by painting, or is colored.

[0043] Transparent member 11 is made of a glass material such as soda-lime glass or borosilicate glass, or a transparent resin material such as PMMA (acrylic), PET, PC (polycarbonate), or PVC (polyvinyl chloride). Transparent member 11 can be colorless and transparent, but can also be colored as long as it allows most of illumination light 15 to pass through.

[0044] In the image recognition device 1 , the imaging method differs depending on the type of the design portion 12 described above.

[0045] Regarding the principles of various imaging methods, Figure 3 as well as Figure 6 Provide explanation.

[0046] exist Figure 3, a schematic diagram of the recognition system for edge-emphasized photography 14 for low-reflection design 12b or mirror-finished design 12c is shown. Edge-emphasized photography 14 is an imaging method that makes edge 10 present around object 2 appear. Camera 4 images object 2. Illuminator 5 generates illumination light 15. Illumination light that is totally reflected within transparent member 11 is total reflection light 16. Total reflection light 16 emitted from the end of transparent member 11 is emitted at edge 10, resulting in edge glow 17. The light imaged by camera 4 within edge glow 17 is edge reflection.

[0047] When performing edge-emphasized imaging 14, the illumination angle of the illuminator 5 must be determined based on the refractive index of the transparent member 11. The illumination angle refers to the angle θ formed between the illumination light 15 from the illuminator 5 and the surface of the transparent member 11. Illumination of the transparent member 11 by the illumination light 15 generates total reflection light 16 within the transparent member 11. The illumination light 15 then propagates through the transparent member 11 and leaks toward the edge 17. The leaked illumination light 15 emphasizes the edge 10, resulting in an image of edge reflection 18 formed by the camera 4.

[0048] The refractive index of the transparent member 11 is 1.5 in the case of glass, for example, and total internal reflection can occur at an illumination angle θ exceeding 0 degrees and within 45 degrees.

[0049] As conditions required for length measurement and recognition, the following evaluation was performed based on the absence of glare around the object 2 and whether an edge outline could be generated.

[0050] exist Figure 4 The length measurement and recognition results based on this evaluation and the changes in the illumination angle θ in the glass are summarized in FIG.

[0051] Based on these results, the illumination angle θ of the illuminator 5 needs to be set to 15 to 45 degrees relative to the surface of the object 2. When the illumination angle θ is 0 degrees, the illumination light does not illuminate the object 2, making it impossible to recognize. When the illumination angle θ is 60 degrees or higher, total reflection does not occur within the glass, resulting in glare from surrounding structures and making it impossible to recognize.

[0052] In addition, Figure 5 2 shows the length measurement and recognition result based on the above evaluation and the change in the illumination light intensity of the illumination light 15 irradiated on the surface of the object 2. The illumination angle θ is fixed at 30 degrees.

[0053] This result shows that it is necessary to provide an illuminator 5 that can set the illumination intensity of the surface of the object 2 to 1000 to 15000 (lux). At an illumination intensity of 800 (lux), the illumination intensity is insufficient to form an image, and recognition is impossible.

[0054] exist Figure 6 , a schematic diagram of the recognition system for the surface-emphasized image 19 of the high-reflection design 12a is shown. While having the same structure as the edge-emphasized image 14, when irradiated with illumination light 15, there is no glare, and the design surface reflection 20 of the color of the high-reflection design 12a is directly imaged.

[0055] If the illumination angle θ at which the illumination light 15 can illuminate the object 2 is greater than or equal to 0 degrees and less than or equal to 90 degrees, imaging can be performed without problems. When the transparent member 11 is irradiated with the illumination light 15, the highly reflective design 12a reflects the illumination light 15, and the surface itself emerges, thereby forming an image as the design surface reflection 20.

[0056] exist Figure 7 , the length measurement and recognition result based on the above evaluation and the illumination light intensity of the illumination light 15 irradiated on the surface of the object 2 is shown. The illumination angle θ is fixed at 30 degrees.

[0057] These results indicate that it is necessary to install an illuminator 5 that provides an illumination intensity of 50 to 500 lux on the surface of object 2. At an illumination intensity of 40 lux, the illumination intensity is insufficient, resulting in no image formation and no recognition. Furthermore, at an illumination intensity of 1000 lux, the surface of object 2 is too bright, causing surface expansion and making it impossible to recognize.

[0058] The illumination intensity must be 1000 (lux) or higher for edge emphasis imaging 14, and 500 (lux) or lower for surface emphasis imaging 19. Therefore, in order to achieve both length measurement and recognition using the same configuration, it is desirable to have an illumination device 5 with a light intensity switching function based on the illumination intensity.

[0059] According to the present embodiment, it is understood that different illumination light intensities are required for edge emphasis imaging 14 and surface emphasis imaging 19. Therefore, as a function of the present embodiment, it is desirable that the image recognition device 1 be equipped with an illumination light intensity switching function.

[0060] exist Figure 8 , it is shown that in this embodiment, the imaging method varies depending on the type of design of the design portion 12. Edge emphasis photography 14 is used to photograph the low-reflection design 12b and the mirror surface design 12c. Surface emphasis photography 19 is used to photograph the high-reflection design 12a.

[0061] exist Figure 9 , which is a flowchart involved in the implementation method.

[0062] First, in step S01 , the object 2 is placed at a predetermined position on the stage 3 .

[0063] Next, in step S02, as an example, if the color information of transparent member 11 is the color of design portion 12 that is transparent to transparent member 11, object processing control unit 9 determines whether design portion 12 is a high-reflection design 12a, a low-reflection design 12b, or a mirror-finished design 12c based on the color of design portion 12. At this time, object 2 is imaged by camera 4, and color recognition is performed by object processing control unit 9, thereby acquiring information on the color of design portion 12.

[0064] An example of a design determination method by the object processing control unit 9 is to pre-create a database that classifies the object 2 into three designs 12a, 12b, and 12c by color. Using this database, as a result of color recognition, a white object is determined to be a high-reflection design 12a, while a black or gray object is determined to be a low-reflection design 12b. Alternatively, a method can be used to determine the design by combining the color information of the design 12 included in the model information of the object 2 with the aforementioned database. Alternatively, the color recognition can be performed based on the RGB color of the design 12. Alternatively, the reflectivity of the object 2 can be measured by imaging the object 2 with the camera 4, and the determination can be made based on a database that shows the relationship between the reflectivity and the three designs. Furthermore, the color information of the transparent member 11 can be replaced with information on the color of the base material 13 that transmits the transparent member 11, information on the color of the bottom surface of the transparent member that transmits the transparent member 11, information on the color of the transparent member itself, or any combination thereof.

[0065] Next, in step S03 , the object processing control unit 9 specifies the illumination light intensity that is predetermined based on the color or design of the object 2 determined by the object processing control unit 9 in step S02 .

[0066] Next, in step S04 , the surface of the object 2 is irradiated with illumination light 15 from the illuminator 5 at the illumination light intensity specified in step S03 .

[0067] Next, in step S05 , the camera 4 captures an image of the object 2 illuminated by the illumination light 15 at the aforementioned illumination light intensity.

[0068] Next, in step S06 , the image data captured by the camera 4 in step S05 is stored in the captured image data storage unit 7 .

[0069] Next, in step S07, the length measurement data calculation unit 8 calculates the length measurement results using the image data stored in the camera data storage unit 7, thereby identifying the outer shape of the transparent component 11. Regarding the calculation of the length measurement results, the following recognition process is performed as an example. First, after obtaining the color image captured by the camera 4, grayscale imaging processing is performed to convert it into a grayscale image. Next, contrast processing is performed to sharpen the image. Next, differential processing is performed to display the outline. Next, two-level binarization processing is performed to remove noise. Next, the outer shape of the transparent component 11 is extracted to detect the size and position of the outer shape. At this time, the deviation is removed from the detected size and position of the outer shape, and the outer shape is finally identified.

[0070] As described above, according to the image recognition method and apparatus of the embodiment, by varying the illumination light intensity of the illumination light 15 in accordance with the color information of the object 2, recognition can be achieved using the same illuminator 5. According to this embodiment, by varying the illumination light intensity of the illumination light 15 in accordance with the color information of the object 2, diffuse reflection on the object surface, which is a major cause of glare, can be suppressed. The low-reflection design 12b or the mirror-finish design 12c can be used to optically emphasize the edge 10 of the object 2. Furthermore, by reducing the illumination light intensity of the high-reflection design 12a, which has a high light reflectivity at the illumination light wavelength, compared to the other designs 12b and 12c, the outline of the object 2 can be accurately recognized while suppressing optical expansion. This makes it possible to achieve length measurement that allows for outline recognition of the object 2 regardless of its various colors.

[0071] Furthermore, by appropriately combining any of the various embodiments or modifications described above, the effects of each embodiment or modification can be achieved. Furthermore, it is possible to combine the embodiments with each other, or to combine the embodiments with each other, or to combine the embodiments with the embodiments, and it is also possible to combine the features of different embodiments or embodiments with each other.

[0072] As described above, according to the image recognition method and apparatus according to the aforementioned aspects of the present invention, recognition using the same lighting device can be achieved by changing the illumination light intensity of the illumination light in accordance with the color information of the object.

[0073] Industrial applicability

[0074] The image recognition method and device involved in the above-mentioned embodiment of the present invention can achieve recognition using the same lighting device by changing the illumination light intensity of the illumination light in accordance with the color information of the object. This is useful in situations where recognition is to be carried out in a situation where an increasing number of diverse objects are required, such as in a recycling process.

[0075] Explanation of symbols

[0076] 1 Image recognition device

[0077] 2 Objects

[0078] 3. Carrier

[0079] 4 cameras

[0080] 5 illuminators

[0081] 6 Identify the length measuring part

[0082] 7. Camera data storage unit

[0083] 8. Length measurement data calculation unit

[0084] 9 Object processing control unit

[0085] 10 Edge

[0086] 11 Transparent parts

[0087] 12 Design Department

[0088] 12a High reflective design

[0089] 12b low reflection design

[0090] 12c mirror design

[0091] 13 Base material

[0092] 14. Edge-emphasized shots

[0093] 15 illumination light

[0094] 16 Totally reflected light

[0095] 17 Edge Glow

[0096] 18 Edge Reflection

[0097] 19 highlighted shots

[0098] 20 Design surface reflection.

Claims

1. An image recognition method for identifying and measuring the shape of an object having a transparent component disposed on its surface. In the image recognition method, irradiating the object with the illumination light from an illumination device capable of irradiating illumination light at a plurality of illumination light intensities, at one illumination light intensity based on color information of the transparent member among the plurality of illumination light intensities, and within an illumination angle at which total reflection can be achieved within the transparent member; photographing the object illuminated by the illumination light, The shape of the transparent component is identified based on the photographed information.

2. The image recognition method according to claim 1, wherein: When the illumination light is irradiated, the illumination angle of the illumination light incident on the surface of the object is greater than or equal to 15 degrees and less than or equal to 45 degrees.

3. The image recognition method according to claim 1 or 2, wherein: The color information of the transparent member is color information of a design portion disposed between the transparent member and a substrate of the object. The color information of the design portion includes color information of a high-reflection design that exhibits a reflectivity of not less than a reference reflectivity and not more than 100% with respect to the wavelength of the illumination light emitted from the illumination device and is not subjected to mirror finishing, color information of a low-reflection design that exhibits a reflectivity of not less than 0% and less than the reference reflectivity, and color information of a mirror-finished design that is subjected to mirror finishing regardless of the reflectivity. The illumination light intensity of each of the color information of the low-reflection design and the color information of the mirror design is greater than the illumination light intensity of the color information of the high-reflection design.

4. The image recognition method according to claim 3, wherein: When irradiating the illumination light, the illumination light intensity of the low-reflection design color information and the mirror design color information is respectively greater than or equal to 1000 and less than or equal to 15000, and the illumination light intensity of the high-reflection design color information is 50 to 500, wherein the unit of illumination light intensity is lux.

5. An image recognition device for recognizing and measuring the shape of an object having a transparent member disposed on its surface. The image recognition device comprises: an illumination device capable of irradiating illumination light at a plurality of illumination light intensities, and irradiating the illumination light toward the object at one illumination light intensity among the plurality of illumination light intensities that is based on color information of the transparent member and within an illumination angle that allows total reflection within the transparent member; a camera for photographing the object illuminated by the illumination light; and The computing unit recognizes the outer shape of the transparent component based on the captured information. The image recognition device according to claim 5 , wherein: In the lighting device, the illumination angle of the illumination light incident on the surface of the object is greater than or equal to 15 degrees and less than or equal to 45 degrees.

7. The image recognition device according to claim 5 or 6, wherein: The color information of the transparent member is color information of a design portion disposed between the transparent member and a substrate of the object. The color information of the design portion includes color information of a high-reflection design that exhibits a reflectivity of not less than a reference reflectivity and not more than 100% with respect to the wavelength of the illumination light emitted from the illumination device and is not subjected to mirror finishing, color information of a low-reflection design that exhibits a reflectivity of not less than 0% and less than the reference reflectivity, and color information of a mirror-finished design that is subjected to mirror finishing regardless of the reflectivity. In the lighting device, each of the illumination light intensity of the color information of the low-reflection design and the color information of the mirror design is greater than the illumination light intensity of the color information of the high-reflection design.

8. The image recognition device according to claim 7, wherein: In the lighting device, the illumination light intensity of the low-reflection design color information and the mirror design color information is 1000 to 15000, and the illumination light intensity of the high-reflection design color information is 50 to 500, wherein the unit of illumination light intensity is lux.

Images