Article detection device, article detection method, and program

Abstract

The item detection device includes: an image acquisition unit that acquires multiple images of the item by changing its relative position to the item; an image processing unit that generates a composite image using the multiple images acquired by the image acquisition unit; and an item region inference unit that infers the region where the item exists from the composite image generated by the image processing unit.

Description

Technical Field

[0001] This invention relates to an article detection device, article detection method, and procedure. Background Technology

[0002] Japanese Patent JP2023-049479A discloses an adhesive device that detects articles being conveyed along a conveying direction and affixes labels to designated positions on the articles while they are being conveyed.

[0003] The label pasting system disclosed in Japanese Patent JP2023-049479A acquires image data of an article in a detection unit that can photograph an article that serves as a cover, and can paste a label on a specified position of the article based on the acquired image data.

[0004] In the label pasting system disclosed in Japanese Patent JP2023-049479A, when it is difficult to detect the boundary between the item and the conveyor unit that serves as the background of the item in the acquired image data, the pasting position of the label on the item sometimes cannot be determined, resulting in label pasting errors. Therefore, there is still room for further improvement in the label pasting system. Summary of the Invention

[0005] Therefore, the purpose of this invention is to improve the detection accuracy of items that are the objects to which labels are affixed.

[0006] According to a certain aspect of the present invention, an item detection apparatus is provided, comprising: an image acquisition unit that acquires multiple images of the item by changing its relative position to the item; an image processing unit that generates a composite image using the multiple images acquired by the image acquisition unit; and an item region inference unit that infers the region where the item exists using the composite image generated by the image processing unit.

[0007] According to a certain aspect of the present invention, since the area where an object exists is inferred from a composite image generated by using multiple images obtained by changing relative positions, the detection accuracy of the object can be improved. Attached Figure Description

[0008] Figure 1 This is a schematic diagram illustrating an example of the configuration of the article detection device according to this embodiment.

[0009] Figure 2 This is a block diagram illustrating the article detection device according to this embodiment.

[0010] Figure 3 This is a flowchart illustrating the article detection method described in this embodiment.

[0011] Figure 4 This is a schematic diagram used to explain the processing of extracting feature regions in the article detection method according to this embodiment.

[0012] Figure 5 This is a concept diagram illustrating the concept of generating a synthetic image based on four feature regions extracted from an image set.

[0013] Figure 6 This is a schematic diagram illustrating the composite image obtained through compositing.

[0014] Figure 7 This is a schematic diagram illustrating an example of a reference shape used to determine the area where an object exists.

[0015] Figure 8 This is a schematic diagram illustrating the label pasting system according to this embodiment.

[0016] Figure 9 This is a diagram illustrating the label that is affixed to an item using a label-attaching system. Detailed Implementation

[0017] The following description is not limited to the accompanying drawings, which are explained in a simple manner.

[0018] A first aspect of the present invention is an article detection apparatus comprising: an image acquisition unit that acquires multiple images of the article by changing its relative position to the article; an image processing unit that generates a composite image using the multiple images acquired by the image acquisition unit; and an article region inference unit that infers the region where the article exists using the composite image generated by the image processing unit.

[0019] According to a first aspect of a certain method of the present invention, in the image acquisition unit, the relative position with respect to the item is changed to acquire multiple images of the item; in the image processing unit, the multiple images acquired by the image acquisition unit are used to generate a composite image; and in the item region inference unit, the composite image generated by the image processing unit is used to infer the region where the item exists.

[0020] Therefore, according to a first aspect of the present invention, since a composite image generated by using multiple images obtained by changing their relative positions is used to infer the area where the object exists, the detection accuracy of the object can be improved.

[0021] Furthermore, in a second aspect of the present invention, based on the article detection device described in the first aspect, when the composite image containing regions of pixel information that are characteristic in each of the plurality of images is inscribed in a predetermined reference shape, the article region inference unit infers the composite image as the region where the article exists.

[0022] According to a second aspect of a certain method of the present invention, when a composite image containing regions of pixel information that are characteristic in each of a plurality of images is inscribed within a predetermined reference shape, the composite image is inferred as the region where an object exists by the object region inference unit, thereby improving the detection accuracy of the object.

[0023] Furthermore, a third aspect of the present invention is based on the article detection device described in the first or second aspect, wherein at least a portion of the article has the same color as the background of the image of the article acquired by the image acquisition unit.

[0024] According to a third aspect of the present invention, a composite image is generated using multiple images obtained by changing their relative positions. In the generated composite image, the area where the object exists is inferred. Therefore, even if at least a part of the object is the same color as the background of the object, the detection accuracy of the object can be improved.

[0025] Furthermore, a fourth aspect of the present invention is based on an article detection device described in any one of the first to third aspects, wherein at least a portion of the article is transparent.

[0026] According to a fourth aspect of the present invention, since a composite image is generated by using multiple images obtained by changing their relative positions, and the area where the object exists is inferred from the generated composite image, the detection accuracy of the object can be improved even if at least a part of the object is transparent.

[0027] Furthermore, a fifth aspect of the present invention is based on an article detection device described in any one of the first to fourth aspects, wherein the image acquisition unit acquires an image of the article being transported along a predetermined transport direction.

[0028] According to a fifth aspect of the present invention, since the image acquisition unit acquires an image of the article being transported along a predetermined transport direction, even if the image acquisition unit is fixed relative to the article detection device, it is still possible to easily acquire multiple images whose relative position to the article has changed.

[0029] Furthermore, a sixth aspect of the present invention is based on an article detection device described in any of the first to fifth aspects. The image acquisition unit includes a first image acquisition unit that acquires a planar image of the article and a second image acquisition unit that acquires height information of the article. The article region inference unit infers the region where the article exists based on the region inferred from the composite image generated using the plurality of images acquired by the first image acquisition unit and the height information acquired by the second image acquisition unit.

[0030] According to a sixth aspect of a certain method of the present invention, the image acquisition unit acquires a planar image of an object through a first image acquisition unit and acquires height information of the object through a second image acquisition unit. The object region inference unit infers the region where the object exists based on the region inferred from a composite image generated using multiple images acquired by the first image acquisition unit and the height information acquired by the second image acquisition unit.

[0031] Therefore, according to the sixth aspect of a certain method of the present invention, the detection accuracy of the article can be improved.

[0032] Furthermore, a seventh aspect of the present invention is based on an article detection device described in any of the first to sixth aspects. The image processing unit uses the plurality of images acquired from the image acquisition unit to generate a first composite image. When the article region inference unit cannot infer the region where the article exists from the first composite image, the image processing unit uses the next plurality of images acquired after the plurality of images to generate a second composite image.

[0033] According to a seventh aspect of a certain embodiment of the present invention, when the item region inference unit is unable to infer the area where the item exists based on the first composite image generated by the image processing unit, the area where the item exists is inferred from the second composite image generated by the image processing unit.

[0034] Therefore, according to the seventh aspect of a certain method of the present invention, the detection accuracy of the article can be improved.

[0035] In addition, an eighth aspect of the present invention is an item detection method, wherein the relative position of the item is changed to acquire multiple images of the item, the acquired multiple images are used to generate a composite image, and the area where the item exists is inferred from the generated composite image.

[0036] According to an eighth aspect of a certain method of the present invention, the relative position of the object is changed to acquire multiple images of the object, and the acquired multiple images are used to generate a composite image, inferring the area where the object exists in the generated composite image.

[0037] Therefore, according to the eighth aspect of a certain method of the present invention, since the area where the object exists is inferred from the synthetic image generated by using multiple images obtained by changing relative positions, the detection accuracy of the object can be improved.

[0038] [Item Detection Device]

[0039] The article detection device 1 according to the embodiments of the present invention will be described.

[0040] Figure 1This is a schematic diagram illustrating the article detection device 1 according to this embodiment. Figure 2 This is a block diagram illustrating the item detection device 1.

[0041] In this embodiment, the article M, which is the object of inspection by the article inspection device 1, is formed by encapsulating food or components using packaging. Therefore, the shape of article M is the shape of the packaging covering the food or components.

[0042] The article detection device 1 described in this embodiment is particularly suitable for situations where at least a portion of the package is transparent or at least a portion of the package is the same color as the conveying surface.

[0043] like Figure 1 As shown, the item detection device 1 includes an image acquisition unit 10 that acquires an image of an item M, and a controller 15 that performs specific image processing on the image data.

[0044] In this embodiment, an image refers not only to a visually observable image displayed on a display device such as a monitor, but also to data used for display and storage in digital devices such as computers and monitors.

[0045] Furthermore, in this embodiment, at least a portion of the packaging is transparent because, in the image acquired by the image acquisition unit 10, it is difficult to distinguish the packaging covering the item M from the background image, whether viewed visually or as pixel information. The same applies when at least a portion of the packaging is the same color as the conveying surface.

[0046] In this embodiment, the conveying devices 20 are arranged in parallel, and the item M is conveyed by the conveying devices 20 along... Figure 1 The conveying direction F is used for conveying. Figure 1 The conveying destination of the conveying device 20 (not shown) may be, for example, a sorting device that can be configured to sort items M, or a label pasting system that can paste labels onto items M.

[0047] In this embodiment, the image acquisition unit 10 is a camera capable of continuously acquiring images of the article M at predetermined intervals. The image acquisition unit 10 is provided vertically above the mounting surface (conveyor surface 20a) on which the article M is mounted in the conveyor device 20, facing the conveyor surface 20a.

[0048] In this embodiment, the image acquisition unit 10 is configured such that the shooting range S, including the conveying surface 20a and the article M placed on the conveying surface 20a for conveying, falls within the viewing angle AF (refer to...). Figure 4 )Inside.

[0049] like Figure 2As shown, the image acquisition unit 10 includes a 2D camera 11 as a first image acquisition unit and a 3D camera 12 as a second image acquisition unit.

[0050] like Figure 1 As shown, the 2D camera 11 acquires images on a plane extending in the transport direction F (X direction) of the item M and in the direction intersecting F (Y direction).

[0051] In addition to the XY plane, the 3D camera 12 can also acquire the depth information (height information) of the object M in the Z direction.

[0052] In this embodiment, the article M is conveyed by the conveying device 20 at a predetermined speed. Therefore, by continuously acquiring images of the article M being conveyed at a predetermined speed at predetermined intervals, the image acquisition unit 10 can acquire multiple images of the article M with different relative positions to the image acquisition unit 10 without changing the position of the image acquisition unit 10.

[0053] The controller 15, not shown, is a computer consisting of a microprocessor, ROM (Read Only Memory), RAM (Random Access Memory) and other storage units, input / output interfaces and buses that connect them.

[0054] like Figure 2 As shown, the controller 15 includes, as a functional configuration, an image processing unit 16 for generating a composite image and an item region inference unit 17 for inferring the area where the item M exists based on the composite image.

[0055] The controller 15 causes the image processing unit 16 and the item area inference unit 17 to operate by executing a control program stored in a storage unit such as ROM in the microprocessor.

[0056] The image processing unit 16 is capable of performing processes such as extracting regions with characteristic pixel information from images obtained by the image acquisition unit 10, and generating composite images using multiple images obtained by the image acquisition unit 10. In this embodiment, the regions with characteristic pixel information will be referred to as feature regions.

[0057] In this embodiment, as an example of the process for extracting feature regions, background subtraction processing can be applied.

[0058] When background differential processing is applied, the image of the conveying surface of the conveying device 20 (hereinafter referred to as the background image) acquired in advance is stored in the storage unit in the item detection device 1.

[0059] By comparing the background image with the image acquired by the image acquisition unit 10, the controller 15 can extract images that are not present in the background image as feature regions.

[0060] The image acquired by the image acquisition unit 10 is a collection of pixels with RGB digital information. In this embodiment, as an example, the brightness value and the grayscale value of the brightness obtained from the information of the pixels can be used for background difference processing.

[0061] Image processing unit 16 can take multiple images (P1~P4) acquired by image acquisition unit 10 as a set image group G (refer to) Figure 4 This is used to process and apply it to image processing.

[0062] As an example, the image processing unit 16 uses multiple images (P1~P4) contained in an image group G to generate a composite image VA (see reference). Figures 4-6 When the item region inference unit 17 is unable to infer the region where item M exists based on the composite image VA, the image processing unit 16 uses other image sets to create a second composite image, wherein the other image sets consist of multiple images captured consecutively at a time later than the time when image set G was acquired.

[0063] As an example, if the 2D camera 11 can capture 10 images in 0.3 seconds, multiple images can be acquired during the transport of the item M across the shooting range S. Furthermore, the item M moving during transport can be captured from various angles. Thus, in order to acquire enough images for the image processing performed by the image processing unit 16, the shooting interval can be adjusted according to the transport speed of the transport device 20.

[0064] The item region inference unit 17 performs the process of inferring the region where item M exists from the composite image generated by the image processing unit 16.

[0065] When the composite image of the feature region generated by the image processing unit 16 is inscribed within a preset reference shape, the item region inference unit 17 infers the composite image as the region where item M exists.

[0066] In this embodiment, the reference shape refers to a shape corresponding to the outline of the article M. The reference shape is a rectangle (see reference 1) pre-defined in the article detection device 1 based on the shape of the article M, which is the object of detection. Figure 7 ).

[0067] When the area where the item M exists cannot be inferred from the composite image generated by the image processing unit 16, that is, when the generated composite image does not match the reference shape, the item area inference unit 17 can use the second composite image produced by the image processing unit 16 to infer the area where the item M exists.

[0068] In addition, the item region inference unit 17 can perform the following processing: inferring the area where item M exists based on the region inferred from the composite image generated using multiple images acquired by 2D camera 11 and the height information acquired by 3D camera 12.

[0069] [Item Inspection Methods]

[0070] Next, the method for detecting item M by item detection device 1 will be explained.

[0071] Figure 3 This is a flowchart illustrating the article detection method performed by article detection device 1. Figure 3 The image processing performed on the image acquired by the 2D camera 11 in the image acquisition unit 10 is described below.

[0072] In step S1, the controller 15 acquires multiple images of the item M transported by the conveying device 20 from the 2D camera 11 in the image acquisition unit 10.

[0073] In step S2, the controller 15 performs processing on each of the multiple images acquired from the 2D camera 11 by the image processing unit 16 to extract regions with pixel information that become features.

[0074] In this embodiment, as an example, background subtraction processing is applied using information about the brightness value and the grayscale value of the brightness obtained from the information of the pixels in the acquired image.

[0075] Figure 4 This is a schematic diagram used to illustrate the process of extracting feature regions in an article detection method.

[0076] exist Figure 4 The image shows an image group G consisting of multiple images P1, P2, P3, and P4 captured sequentially by a 2D camera 11. Image P1 is the image captured at time t1, image P2 is the image captured at time t2, image P3 is the image captured at time t3, and image P4 is the image captured at time t2.

[0077] For each of the acquired images P1, P2, P3, and P4, the controller 15 extracts the region consisting of pixels with values ​​exceeding a specified brightness value as a feature region.

[0078] Rectangular representation of image P1 Figure 1 The image shows the field of view AF of the shooting range S. The diagonal lines in image P1 represent a portion of the conveyor surface 20a and the object M seen through the conveyor surface 20a. Alternatively, they represent a portion of the conveyor surface 20a and the object M of the same color as the conveyor surface 20a.

[0079] The blank areas in image P1 represent pixels with values ​​exceeding a predetermined brightness value. In this embodiment, the set of blank areas in image P1 is represented as feature region V1.

[0080] Similarly, controller 15 extracts feature regions V2, V3, and V4 from each of the acquired images P2, P3, and P4.

[0081] Since item M is transported by conveyor 20, therefore... Figure 4 As shown, in each of the images P1 to P4 taken at specified intervals, the positions of the feature regions V1 to V4 under the AF view are different.

[0082] Next, in step S3, the controller 15 generates a composite image of the feature regions V1 to V4 extracted from the image group G.

[0083] Figure 5 This is a concept diagram illustrating the concept of generating a synthetic image based on four feature regions V1~V4 extracted from image group G. Figure 6 This is a schematic diagram illustrating the composite image obtained through compositing.

[0084] like Figure 5 As shown, the controller 15 selects a brightness value with a strong feature in the feature region V1 and sets the reference coordinates. Figure 5 In the figure, (X11, Y11), (X12, Y12) and (X13, Y13) are shown as reference coordinates.

[0085] The controller 15 sets reference coordinates (X21, Y21), (X22, Y22), and (X23, Y23) corresponding to the reference coordinates of V1 for feature region V2, sets reference coordinates (X32, Y32) and (X33, Y33) corresponding to the reference coordinates of V1 for feature region V3, and sets reference coordinates (X41, Y41), (X42, Y42), and (X43, Y43) corresponding to the reference coordinates of V1 for feature region V4.

[0086] Based on the aforementioned reference coordinates, controller 15 overlaps multiple feature regions V1~V4 to generate a composite image VA of feature regions V1, V2, V3, and V4 (see reference). Figure 6 ).

[0087] Next, in step S4, the controller 15 performs a process that infers the area where the item M exists through the item area inference unit 17. That is, the controller 15 determines whether the synthesized image VA fits the preset reference shape B.

[0088] Figure 7 This is a schematic diagram illustrating an example of a reference shape B established to determine the area where an object exists.

[0089] When the controller 15 determines that the synthetic image VA fits the reference shape B (step S4: Yes), in step S5, the area enclosed by the reference shape B inscribed by the synthetic image VA is determined as the area where the item M exists.

[0090] When the controller 15 determines that the synthesized image VA does not fit the reference shape B (step S4: no), it repeatedly performs the process starting from step S1.

[0091] The controller 15 uses another group of images, consisting of multiple images captured consecutively at a time later than the image group G, to perform image processing after step S2.

[0092] By executing the series of image processing steps S1 to S5 described above, the controller 15 can infer the area where the object M exists based on the image acquired by the 2D camera 11.

[0093] The controller 15 also performs operations on the images acquired by the 3D camera 12 in the image acquisition unit 10. Figure 4 The same image processing is used. The controller 15 infers the area where the object M exists based on the region inferred from the composite image generated using multiple images acquired by the 2D camera 11 and the height information acquired by the 3D camera 12.

[0094] [Effects of the item detection device and method]

[0095] According to the embodiment of the present invention, the article detection apparatus 1 and the article detection method executed by the article detection apparatus 1 acquire multiple images of the article by changing the relative position with respect to the article M in the image acquisition unit 10, and generate a composite image VA using the multiple images acquired by the image acquisition unit 10, and infer the area where the article M exists using the composite image VA.

[0096] Therefore, the item detection device 1 can improve the detection accuracy of items.

[0097] When the item detection device 1 determines by the item area inference unit 17 that the composite image VA of the feature areas V1, V2, V3 and V4 of the multiple images is inscribed in the reference shape B, the area enclosed by the reference shape B inscribed in the composite image VA can be determined as the area where the item M exists.

[0098] Therefore, the item detection device 1 can improve the detection accuracy of items.

[0099] Furthermore, according to the item detection device 1, since the area where the item M exists is inferred from the composite image VA generated by using multiple images P1 to P4 obtained by changing relative positions, the detection accuracy of the item can still be improved even when at least a part of the item M has the same background color as the item M, or when at least a part of the item M is transparent and the background color is detected inside the shape of the item M.

[0100] Furthermore, according to the article detection device 1, since the image acquisition unit 10 acquires an image of the article M conveyed by the conveying device 20 along the conveying direction F, even if the image acquisition unit 10 is fixed relative to the article detection device 1, it can still easily acquire multiple images whose relative position to the article M has changed. As a result, the device configuration of the article detection device 1 can be simplified.

[0101] Furthermore, according to the item detection device 1, the image acquisition unit 10 acquires a planar image of the item M using the 2D camera 11 and acquires the height information of the item M using the 3D camera 12. Since the item region inference unit 17 infers the area where the item M exists based on the region inferred from the composite image obtained by the image obtained by the 2D camera 11 and the height information acquired by the 3D camera 12, the detection accuracy of the item M can be improved.

[0102] Furthermore, according to the item detection device 1, when multiple images (P1~P4) acquired from the image acquisition unit 10 are grouped into an image group G (refer to...) Figure 4 When the synthesized image VA obtained by image processing based on the image group G cannot determine the area where the item M exists, the synthesized image obtained by image processing based on the image group consisting of the next few images can be used to determine the area where the item M exists. Therefore, the item detection device 1 can improve the detection accuracy of the item M.

[0103] [program]

[0104] The program involved in this embodiment can be executed by the computer of the article detection device 1. The program causes the computer to perform the following processes: change the relative position with respect to the article M to acquire multiple images of the article M; use the acquired multiple images to generate a composite image; and infer the area where the article M exists in the generated composite image.

[0105] The program is stored in a storage unit accessible via the controller 15 of the item detection device 1. Alternatively, the program is provided in a storage medium that can be loaded and unloaded relative to the item detection device 1.

[0106] [Label pasting system]

[0107] As an example, the article detection device involved in this embodiment can be combined with a label pasting system that affixes labels to articles to form a label pasting system.

[0108] Figure 8 This is a schematic diagram illustrating the label pasting system 100 according to this embodiment.

[0109] The label pasting system 100 involved in this embodiment is related to... Figure 1 and Figure 2 The item detection device 1 shown is used in combination with a system that affixes a label LB to the item M detected by the item detection device 1.

[0110] The label pasting system 100 includes an item detection device 1, a conveying device 20, a pasting arm 30, a printer 40, a label conveying unit 50, and a downstream detection unit 60. All of the above components are controlled by a controller 70.

[0111] Figure 8 The item detection device 1 and conveying device 20 shown are Figure 1 The item detection device 1 and the conveying device 20 shown are the same.

[0112] The adhesive arm 30 has a label holding part 31. At the front end of the label holding part 31, there is a structure for attracting the label LB to hold it.

[0113] The adhesive arm 30 has an actuator that allows for longitudinal, lateral, and height movement and angle adjustment around an axis.

[0114] Therefore, the adhesive arm 30 can pick up and hold the label LB conveyed by the label conveying unit 50 with the adhesive surface facing the item M. In addition, the adhesive arm 30 can move the label LB to the inbound item M and affix the label LB to the designated position on the item M.

[0115] The printer 40 includes a printer body 41 for printing the information required for the label LB, a label roll 42 formed by the label LB being temporarily adhered to a continuous sheet of release paper before printing, and a backing paper roll 43 in a state where the continuous sheet of release paper is recycled after the label LB is peeled off.

[0116] In this embodiment, printer 40 is a printer that can independently set and print the printing content of item M without being connected to an information processing terminal such as a personal computer.

[0117] The label affixing system 100, equipped with a printer 40, is capable of printing information about an item M onto a label LB before affixing. This information about the item M may include, for example, the manufacturing date, shelf life, and raw material names if it is food.

[0118] The label conveying unit 50 is equipped with a conveyor belt 51 for conveying labels LB that are peeled off from the continuous sheet of release paper, and conveys multiple labels LB individually and continuously.

[0119] The downstream detection unit 60 is disposed in the conveying device 20 downstream of the adhesive arm 30. In this embodiment, the downstream detection unit 60 is a camera that takes pictures of the item M after the label LB has been affixed. The image data acquired by the downstream detection unit 60 is used in the controller 70 to process and determine the affixing status of the label LB.

[0120] The controller 70 is a computer consisting of a microprocessor, storage units such as ROM and RAM, input / output interfaces, and buses that connect them.

[0121] In this embodiment, the controller 70 can determine the affixing position of the label LB relative to the item M based on the detection result of the item M obtained from the item detection device 1.

[0122] Figure 9 This is a diagram illustrating the label LB affixed to item M via label affixing system 100.

[0123] exist Figure 9 The diagram shows states (a), (b), and (c). State (a) is the state where label LB is being conveyed via conveyor belt 51. State (b) is the state where the action of label LB being picked up by adhesive arm 30 and transported to item M while being turned toward the specified direction is being performed. State (c) is the state where the action of label LB being affixed to item M is being performed.

[0124] like Figure 9 As shown, according to the item detection device 1, even if the item M is packaged in a transparent package or a package with the same color as the conveyor surface 20a (background image), the item M can still be detected with high accuracy.

[0125] Therefore, by acquiring data about the item M from the item detection device 1, the label affixing system 100 is able to affix the label LB to the item M at an appropriate position and in an appropriate orientation relative to the shape of the item M.

[0126] For example, such as Figure 9 As shown, the label LB can be moved toward the item M while simultaneously aligning one side of the label LB with the shape of the item M, thereby attaching the label LB to the item M.

[0127] Therefore, it can improve the aesthetics between item M and the label LB pasted on item M.

[0128] [Other Implementation Methods]

[0129] The embodiments of the present invention have been described above. However, the above embodiments only show a part of the application examples of the present invention and are not intended to limit the technical scope of the present invention to the specific configuration of the above embodiments.

[0130] In this embodiment, the article M is conveyed, and the image acquisition unit 10 is fixed. In this invention, the image acquisition unit 10 can be movable as long as its relative position can be changed to capture multiple images of the article M.

[0131] In this embodiment, a sensor that illuminates the object M with detection light such as infrared light or laser and obtains height information based on its reflection time can also be used instead of the 3D camera 12.

[0132] The synthesis processing that generates a synthetic image based on multiple feature regions can also apply image processing other than background subtraction. As an example of image processing applicable in this embodiment, algorithms for detecting objects of known shapes from acquired images can be cited. For example, there are methods that use R-CNN or YOLO to obtain bounding boxes (shapes) from multiple images.

[0133] In addition, there are methods that use image processing techniques such as optical flow and known transport speed information to overlap multiple images that are inferred to be item M, thereby determining the area where item M exists in the image.

[0134] As a method for using conveying speed information, the controller 15 can also perform the following processing: acquire the conveying speed of the conveying device 20 that conveys the item M, correct the offset of the display position of feature region V2 in image P2 relative to feature region V1 in image P1 based on the conveying speed, and synthesize feature regions V1 and V2. The same processing is performed on feature regions V3 and V4.

[0135] This application claims priority based on Japanese Patent Application No. 2024-072244, filed with the Japan Patent Office on April 26, 2024, the entire contents of which are incorporated herein by reference.

Claims

1. An item detection device, characterized in that, have: An image acquisition unit that changes its relative position to the item to acquire multiple images of the item; An image processing unit uses the plurality of images acquired by the image acquisition unit to generate a composite image; as well as The item region inference unit uses the synthetic image generated by the image processing unit to infer the region where the item exists.

2. The article detection device according to claim 1, characterized in that, When the composite image, which contains regions of pixel information that are characteristic of the plurality of images, is inscribed within a pre-defined reference shape, the item region inference unit infers the composite image as the region where the item exists.

3. The article detection device according to claim 1, characterized in that, At least a portion of the item has the same color as the background of the image of the item acquired by the image acquisition unit.

4. The article detection device according to claim 1, characterized in that, At least a portion of the item is transparent.

5. The article detection device according to claim 1, characterized in that, The image acquisition unit acquires an image of the article being transported along a predetermined transport direction.

6. The article detection device according to claim 1, characterized in that, The image acquisition unit includes a first image acquisition unit for acquiring a planar image of the item and a second image acquisition unit for acquiring height information of the item. The item region inference unit infers the region where the item exists based on the region inferred from the composite image generated using the plurality of images acquired by the first image acquisition unit and the height information acquired by the second image acquisition unit.

7. The article detection device according to claim 1, characterized in that, The image processing unit uses the plurality of images acquired from the image acquisition unit to generate a first composite image. When the item region inference unit is unable to infer the area where the item exists from the first composite image, the image processing unit uses the next plurality of images acquired after the plurality of images to generate a second composite image.

8. A method for detecting an article, characterized in that, By changing the relative position of the object, multiple images of the object can be obtained. The acquired multiple images are used to generate a composite image. In the generated synthetic image, the area where the item exists is inferred.

9. A program executable by a computer of an article detection device, characterized in that, The program causes the computer to perform the following processes: By changing the relative position of the object, multiple images of the object can be obtained; Use the acquired multiple images to generate a synthetic image; and In the generated synthetic image, the area where the item exists is inferred.

Images