JP7627619B2 - Item type determination method and item type determination system - Google Patents

Description

The present invention relates to an item type determination method and an item type determination system for determining the type of an approximately rectangular parallelepiped item.

As described in Patent Document 1, there is a technology in which RFID tags containing identification information indicating the type of item are attached to multiple items, and the type of item is determined by reading the RFID (Radio Frequency Identifier) tags.

JP 2019-202841 A

However, with conventional technology, it was necessary to attach an RFID tag to every item in advance, and it was not possible to identify items that did not have an RFID tag attached.

The present invention has been made in consideration of the above problems, and aims to provide an item type determination method and an item type determination system that determine the type of an item using point data obtained by three-dimensionally detecting the unevenness of the surface of an item.

In order to achieve the above object, one of the present inventions is an item type determination method that uses a detector that acquires the distance to the item as multiple three-dimensional point data and determines the type of the item based on the shape characteristics of the detection surface portion of the approximately rectangular parallelepiped item, including the surface facing the detector, and includes a point cloud acquisition process that acquires point data of multiple locations on the detection surface portion of the item as a point cloud, an edge detection process that detects a pair of opposing edges of the detection surface portion of the item based on the point cloud acquired in the point cloud acquisition process, a straight line shape extraction process that extracts a straight line shape having linearity extending in a direction along the edge based on the point cloud existing between the pair of edges detected by the edge detection process, and a type determination process that determines the type of the item based on the straight line shape extracted by the straight line shape extraction process.

In order to achieve the above object, another aspect of the present invention is an item type determination system that includes a detector that acquires the distance to an item as multiple three-dimensional point data, and an item type determination device that determines the type of item based on the shape characteristics of the detection surface portion of the item, including the surface of the item facing the detector. The item type determination device includes a point cloud acquisition unit that acquires point data of multiple locations on the detection surface portion of the item as a point cloud, an edge detection unit that detects a pair of opposing edges of the detection surface portion of the item based on the point cloud acquired by the point cloud acquisition unit, a straight line shape extraction unit that extracts a straight line shape having linearity extending in a direction along the edge based on the point cloud existing between the pair of edges detected by the edge detection unit, and a type determination unit that determines the type of the item based on the straight line shape extracted by the straight line shape extraction unit.

The present invention makes it possible to distinguish between objects that have linear irregularities on their surfaces and other objects.

1 is a perspective view showing an item type determination system to which an item type determination method is applied; 1 is a perspective view showing the vicinity of a transfer device of an article type determination system. FIG. 1 is a block diagram showing a functional configuration of an item type determination system. 13 is a flowchart showing a process flow of a section detection unit. 11A to 11C are diagrams illustrating the processing state at each stage of the process of the section detection unit. 13 is a flowchart showing a process flow of an edge detection unit. 11A to 11C are diagrams showing the processing state of the edge detection unit at each step of the process. 13 is a flowchart showing a process flow of a straight line shape extraction unit. 11A to 11C are diagrams illustrating the processing state at each stage of the process of the straight line shape extraction unit. 10 is a flowchart showing a process flow of an item type determination unit.

Below, an embodiment of the product type determination method and product type determination system according to the present invention will be described with reference to the drawings. Note that the following embodiment is an example to explain the present invention, and is not intended to limit the present invention. For example, the shapes, structures, materials, components, relative positional relationships, connection states, numerical values, mathematical expressions, the contents of each step in the method, and the order of each step shown in the following embodiment are examples, and may include contents not described below. In addition, geometric expressions such as parallel and orthogonal may be used, but these expressions do not indicate mathematical strictness, and include errors, deviations, etc. that are substantially acceptable. In addition, expressions such as simultaneous and identical also include a substantially acceptable range.

The drawings are schematic diagrams in which emphasis, omissions, or proportions have been appropriately adjusted in order to explain the present invention, and may differ from the actual shapes, positional relationships, and proportions.

In addition, multiple inventions may be collectively described below as a single embodiment. Some of the content described below is also described as an optional component of the present invention.

Figure 1 is a perspective view showing an item type determination system to which the item type determination method is applied. Figure 2 is a perspective view showing the vicinity of a transfer device 130 of the item type determination system. The item type determination system 100 is a system that determines the type of item 200 placed on a rack 110, and includes a detector 140 and an item type determination device 160 (not shown in Figures 1 and 2). In this embodiment, the item type determination system 100 is a device that can automatically transport the item 200 that has been brought in and automatically transfer the item 200 to a storage position for the item 200, and also transport the item 200 that has been automatically transferred from the storage position, and is incorporated in a so-called automated warehouse that includes a transport device 120 and a transfer device 130.

The object 200 to be determined by the object type determination system 100 is not particularly limited as long as it is approximately rectangular. Approximately rectangular includes shapes formed by six flat rectangular faces, and is described as including shapes with rib-like or flange-like protrusions, handle-like depressions or holes, etc. Specifically, there are two types of object 200: a first type of object with flat surface characteristics without linear unevenness on the surface, such as cardboard boxes, paper boxes, and wooden boxes, and a second type of object with surface characteristics with linear unevenness on the surface, such as containers, trays, and foldable containers.

The rack 110 is a facility for storing a plurality of roughly rectangular parallelepiped objects 200 arranged so that the detection surface 201 of the object 200 facing the area in which the conveying device 120 moves is aligned along a predetermined arrangement direction (X-axis direction in the figure). In the present embodiment, the rack 110 includes a shelf 111 for holding the object 200 in a placed state, and a support 112 for supporting the shelf 111. The shelf 111 is flat, and the position in which the object 200 is stored is not particularly limited. The rack 110 stores a mixture of a first type of object 200 and a second type of object 200. Note that, although the rack 110 is shown on one side as viewed in the direction of movement of the conveying device 120 in the figure, the rack 110 may be arranged on both sides.

The conveying device 120 is a device that holds and conveys the item 200, and is not particularly limited as long as the detector 140 can be attached to it. Examples of the conveying device 120 include an unmanned guided vehicle that holds the item 200 and travels autonomously on a floor surface, and a tracked cart that holds the item 200 and travels along a predetermined track such as a rail. In the case of this embodiment, the conveying device 120 is a so-called stacker crane that includes a rail 121, a cart 122 that travels on the rail 121, a mast 123 that is attached upright to the cart 122 and moves together with the cart 122, and a lifting platform 124 that can hold the item 200 and moves up and down along the mast 123.

The transfer device 130 is a device that transfers the item 200 between the rack 110 and the lifting platform 124 of the transport device 120, and transfers the item 200 by moving it in the depth direction (Y-axis direction in the figure) perpendicular to the arrangement direction of the items 200 (X-axis direction in the figure) in a horizontal plane. The type of transfer device 130 is not particularly limited, and examples include a device that transfers the item 200 by clamping both opposing sides of the item 200, a device that hooks a claw on the rear side or front side of the item 200 and transfers the item 200 while sliding it, and a device that scoops up the item 200 and transfers it.

In this embodiment, the transfer device 130 is attached to the lifting platform 124 of the transport device 120, and is capable of transferring the item 200 between the rack 110 and the lifting platform 124. When the racks 110 are disposed on both sides of the transport device 120, the transfer device 130 is configured to be able to transfer the item 200 to either of the racks 110 on either side.

The detector 140 is a sensor that acquires the distance between the detector 140 and multiple points on the detection surface portion 201 of the article 200, including the surface facing the detector 140, as multiple three-dimensional point data. The type of the detector 140 is not particularly limited, but examples include three-dimensional distance measuring sensors such as a LiDAR (Laser Imaging Detection and Ranging) sensor and a TOF (Time of Flight) camera.

The location where the detector 140 is attached is not particularly limited, but it is preferable to attach it to the location where the item 200 is transferred, for example, because the relative positional relationship between the transfer position and the item 200 can be accurately detected. In the present embodiment, the detector 140 is attached to the lifting platform 124 of the conveying device 120. The number of detectors 140 provided in the item type determination system 100 is not particularly limited, but in the present embodiment, two detectors 140 are provided lined up in the arrangement direction of the items 200. The distance between the two detectors 140 is set to include a range of distances that can acquire point data of a predetermined density or more and can detect at once the expected maximum width of the item 200 stored in the rack 110 in the arrangement direction and the area covering the gap between the adjacent items 200 set on both sides. In the present embodiment, the detectors 140 are attached one each on both sides of the width direction (X-axis direction in the figure) of the lifting platform 124. In other words, the detectors 140 are arranged on both sides of the area through which the item 200 transferred by the transfer device 130 passes. This makes it possible to accurately detect the positional relationship between the area where the item 200 is being transferred and the item 200 stored on the nearby rack 110.

Figure 3 is a block diagram showing the functional configuration of the product type determination system. The product type determination device 160 is a device that determines the type of product based on the shape characteristics of the detection surface portion 201 of the product 200. The product type determination device 160 includes a point cloud acquisition unit 151, an edge detection unit 155, a straight line shape extraction unit 156, and a type determination unit 157 as processing units that are realized by causing a processor to execute a program. In this embodiment, the product type determination device 160 includes a section detection unit 152.

The point cloud acquisition unit 151 acquires point data of multiple locations on the detection surface 201 of the item 200 from the detector 140 as a point cloud. The data structure of the point data is not particularly limited, but includes, for example, three-dimensional data indicating a relative positional relationship with the transfer device 130. In the case of this embodiment, the data includes data of an orthogonal coordinate system including data on the arrangement direction (X-axis direction in the figure) of the items 200 on the shelf 111 of the rack 110, an orthogonal direction (Z-axis direction in the figure) perpendicular to the arrangement direction on the detection surface 201, and a depth direction (Y-axis direction in the figure) perpendicular to both the arrangement direction and the orthogonal direction.

In this embodiment, the product type determination system 100 includes multiple detectors 140 arranged in the array direction, and the point cloud acquisition unit 151 acquires a point cloud from each of the detectors 140. Since adjacent detectors 140 have a portion of the imaging angle of view overlapping, the point cloud acquisition unit 151 statistically processes the point data included in the overlapping area and combines the point clouds of the multiple detectors 140 into a single image. This makes it possible to process point data related to the detection surface portions 201 of multiple products 200 in the array direction as a single image. Note that instead of combining the point clouds of the multiple detectors 140 into a single image, it is also possible to obtain processing results using images corresponding to the point clouds of each detector 140 and combine the results.

The section detection unit 152 detects an object-free section in which no object 200 exists in the arrangement direction from the point cloud acquired by the point cloud acquisition unit 151. FIG. 4 is a flowchart showing the flow of processing by the section detection unit. FIG. 5 is a diagram showing the processing state at each stage of the process of the section detection unit.

An example of the process flow of the section detection unit 152 will be described. First, the section detection unit 152 executes a trimming process to extract point data contained within a predetermined area from the point cloud acquired from the point cloud acquisition unit 151 (S101, trimming process). The predetermined area is not particularly limited, but can be, for example, a strip-shaped area having a predetermined height that includes the center position of the minimum height (length in the orthogonal direction) of the item 200 stored in the rack 110. Specifically, point data having a value in the orthogonal direction (Z-axis direction in the figure) less than the first threshold value is excluded, and point data having a value in the orthogonal direction (Z-axis direction in the figure) greater than the second threshold value (>first threshold value) is excluded. In other words, point data in the range of the first threshold value or more and the second threshold value or less is extracted. Since the trimming process detects the no-item section using data in a predetermined height range, it is possible to suppress the influence of reflections from the shelf boards 111 of the rack 110 and the influence of ribs, flanges, holes, etc. provided on the surface of the item 200, and it is possible to accurately detect the no-item section. In addition, it is possible to suppress the amount of data of the point cloud and promote the processing of the next process.

Next, the section detection unit 152 projects the point cloud trimmed by the trimming process in the depth direction (transfer direction) to generate a two-dimensional first projection image shown in FIG. 5(a) (S102, first projection process). Specifically, the two-dimensional first projection image is generated by excluding data in the depth direction (Y-axis direction in the figure) from each point data. Here, the section detection unit 152 may project the point cloud trimmed by the trimming process in the up-down direction to generate a two-dimensional first projection image.

Next, the section detection unit 152 performs morphology processing on the first projection image, complementing the points that existed as rough points, and modifies the data so that the part corresponding to the item 200 becomes a single mass (the white part in the image in FIG. 5) as shown in FIG. 5(b) (S103, first morphology step). In this embodiment, the section detection unit 152 performs a closing process that repeats expansion and contraction as the morphology process.

Next, the section detection unit 152 searches for an existence section that indicates the section in which the item 200 exists (S104, existence section rough search process). In the case of this embodiment, the section detection unit 152 updates the part regarded as a white mass by the first morphology process (the white part in FIG. 5(b)) by complementing the disconnected and missing parts in the vertical direction in the image corresponding to the orthogonal direction. When the existence section end search process is executed, the state becomes, for example, as shown in the image in FIG. 5(c).

Next, the section detection unit 152 compares the distance between the black parts in the horizontal direction (left and right when looking at the drawing) in the image corresponding to the arrangement direction of the section between adjacent white block parts with a predetermined gap threshold (S105). If the distance between the black parts is less than or equal to the gap threshold (S105: Yes), the section detection unit 152 updates the part considered to be a white block by complementing the gap as shown in FIG. 5(d) (S106, gap filling process). Note that "<=" in FIG. 4 indicates less than or equal to.

Next, the section detection unit 152 identifies the white block portion updated in the gap filling process or the white block portion updated in the rough search process for the presence section from one end to the other end in the arrangement direction (horizontal direction in FIG. 5) (inverted triangle marks in (d) and (e) in FIG. 5) as the presence section corresponding to the section where the item 200 exists (S107, presence section identification process). In addition, sections other than the presence section, such as between adjacent presence sections, are identified as no-item sections.

Next, the section detection unit 152 determines the middle position in the arrangement direction (horizontal direction in FIG. 5) of the no-item section as the reference position (the position of the arrow in FIG. 5(e)) (S108, reference position determination step). The criteria for determining the reference position are not particularly limited, but for example, the center position between the ends of adjacent existence sections may be determined as the reference position. In addition, when an object-free section with a length equal to or greater than a predetermined first section threshold exists (left side of FIG. 5(e)), the end of a virtual existence section (triangle mark in FIG. 5(e)) may be set at a position a predetermined distance away from the end of the existence section, and the reference position may be set between the end of the existence section and the end of the virtual existence section. Note that the distance at which the virtual existence section is set may be equal to the first section threshold. In addition, in the case of an object-free section that is not sandwiched between the ends of the existence section and is equal to or less than a predetermined second section threshold (right side of FIG. 5(e)), the end of the image may be set as the reference position.

The edge detection unit 155 detects at least a pair of opposing edges of the detection surface 201 of the article 200 based on the point cloud acquired by the point cloud acquisition unit 151. In this embodiment, the edge detection unit 155 detects two pairs of edges that intersect (orthogonal to) each other as contours. Figure 6 is a flowchart showing the flow of processing by the edge detection unit. Figure 7 is a diagram showing the processing state at each stage of the process of the edge detection unit. Note that the vertical line shown in the center in the left-right direction of Figure 7 indicates the boundary of the areas corresponding to the two detectors 140.

An example of the processing flow of the edge detection unit 155 will be described. First, the edge detection unit 155 projects the point cloud acquired from the point cloud acquisition unit 151 in the depth direction (transfer direction) along the normal direction of the detection surface unit 201, and converts the depth direction data of each point data into the gradation of each dot in a two-dimensional plane to generate a gradated projection image, which is a two-dimensional second projection image shown in Figure 7 (a) (S201, second projection step). In this embodiment, the gradated projection image is represented in grayscale.

Next, the edge detection unit 155 performs morphology processing on the gradated projection image to supplement missing portions of point data and change the data so that the portion corresponding to the item 200 as shown in FIG. 7(b) becomes one block (S202, second morphology step). In this embodiment, the edge detection unit 155 performs a closing process, which repeats expansion and contraction as a morphology process, on the entire image including the existence section detected by the section detection unit 152.

Next, the edge detection unit 155 detects a pair of opposing edges that are the edges of the detection surface 201 of the article 200 (S203, edge detection process). In this embodiment, the edge detection unit 155 converts each of the parts (three parts aligned in the left-right direction in FIG. 7B) that are regarded as a block by the second morphology process for each existence section into a binary image with gradation removed as shown in FIG. 7C, and detects the contour of the detection surface 201 (parts shown by rectangles in FIG. 7C) by image analysis based on this binary image. The edge detection unit 155 detects a pair of lines extending in the vertical direction (corresponding to the orthogonal direction) in the figure of the detected rectangular contour as the edge. In this embodiment, since the contour is detected for each of the point clouds acquired from the two detectors 140, two contours are detected for the existence section including the boundary as shown in FIG. 7C.

The straight line shape extraction unit 156 extracts a straight line shape having linearity extending in a direction along a pair of edges detected by the straight line shape extraction unit 156 through image analysis, based on a group of points existing between the pair of edges detected by the straight line shape extraction unit 156. Figure 8 is a flowchart showing the flow of processing by the straight line shape extraction unit. Figure 9 is a diagram showing the processing state at each stage of the process by the straight line shape extraction unit.

In this embodiment, the contour obtained by image analysis is updated to be several pixels (corresponding to several millimeters) inward.

In this embodiment, the straight line shape extraction unit 156 extracts the straight line shape using the gradient projection image generated by the second projection step (S201). The straight line shape extraction unit 156 also extracts the straight line shape using point data included inside the contour determined by the edge detection unit 155. This makes it possible to search for the straight line shape while excluding the edge parts of the item 200, and prevents erroneous detection of the type of item 200 due to point data corresponding to the edge parts.

The process flow of the straight line shape extraction unit 156 will be described. The straight line shape extraction unit 156 acquires a toned projection image from the edge detection unit 155, and extracts straight line shape candidates contained in the toned projection image as shown in FIG. 9(d) based on a predetermined straight line detection algorithm for the toned projection image within each contour (S301, straight line shape candidate extraction process).

Next, the straight line shape extraction unit 156 removes noise from the extracted straight line shape candidate and updates the data so that it becomes a single line shape corresponding to the rib-like protruding portion on the detection surface portion 201 of the article 200 as shown in FIG. 9(e) (the white part within the outline on the left side of FIG. 9(e)) (S302, noise removal process).

Next, the straight line shape extraction unit 156 extracts straight line shapes from the straight line shape candidates from which noise has been removed (S203, straight line shape extraction process). In this embodiment, the straight line shape extraction unit 156 extracts straight line shape candidates that are parallel or approximately parallel to a pair of edges extending in the up-down direction (corresponding to the orthogonal direction) in the figure as straight line shapes. This results in an image such as that shown in (f) of Figure 9. Note that in this embodiment, straight line shape candidates that are parallel or approximately parallel to a pair of edges extending in the left-right direction (corresponding to the arrangement direction) in the figure are also extracted as straight line shapes.

The type determination unit 157 determines the type of the item 200 based on the linear shape extracted by the linear shape extraction unit 156. In this embodiment, as shown in the flowchart of the type determination process in FIG. 10, if a linear shape is not extracted in the linear shape extraction process (S401: No), the type determination unit 157 determines that the detection surface of the item 200 is a first type of item having a flat surface property (S403), and if a linear shape is extracted in the linear shape extraction process (S401: Yes), the type determination unit 157 determines that the detection surface of the item 200 is a second type of item having irregularities. Finally, the type determination unit 157 notifies the conveying device 120 of the determination result (S104).

In this embodiment, the product type determination device 160 performs the above-mentioned edge detection process, straight line shape extraction process, and product type determination process for each product presence section separated by the section detection unit 152, and notifies the conveying device 120 of each result.

According to the product type determination system 100 of the above embodiment, since the edge of the product 200 has a linear shape, the edge is excluded and the linear shape is extracted to determine the type of the product. This makes it possible to determine the type of the product 200 with high accuracy even in an environment where a first type of product 200 whose detection surface 201 has a flat surface property and a second type of product 200 whose surface has irregularities such as ribs are arranged in a mixed state on the rack 110.

In particular, the accuracy of determining the type of item 200 can be improved by making the determination using the linear shape along a pair of opposing edges of the detection surface portion 201.

In addition, after detecting object-free sections, which are sections between adjacent objects 200 where no object 200 exists, and separating the objects 200 from each other in the point cloud, the type of object 200 is determined for each presence section, so that the types of multiple objects 200 can be quickly notified to the conveying device 120, etc. based on the point cloud acquired at one time.

In addition, by determining the contour based on the binary projection image and extracting straight line shapes based on the gradient projection image within the determined contour, it is possible to effectively exclude the edge of the article 200.

In addition, by performing noise processing, etc. on the extracted straight line shape candidates before extracting the straight line shape, it is possible to extract a straight line shape that is effective for determining the type of item 200 even if the point cloud contains a lot of noise.

The present invention is not limited to the above-described embodiment. For example, the components described in this specification may be combined in any way, or some of the components may be removed to create another embodiment of the present invention. The present invention also includes modifications that are made to the above-described embodiment by those skilled in the art without departing from the spirit of the present invention, i.e., the meaning of the words in the claims.

For example, in the above embodiment, a case where a pair of edges are parallel has been described, but the article type determination method may also be applied when, for example, the detection surface portion 201 is trapezoidal and the pair of edges are not parallel. In this case, there are two types of inclination of the linear shape along the pair of edges. In this case, the type of article 200 may be determined based on the linear shape of at least one of the inclinations.

In addition, although the case where the contour of the detection surface portion 201 is detected has been described, the type of the item 200 may be determined using point data sandwiched between a pair of edges without detecting the contour.

In addition, although the example shows a case where multiple detectors 140 are provided, it is also possible to provide a single detector 140 that can simultaneously detect an area that is the width of the central item 200 among three items 200 lined up consecutively on the shelf 111 of the rack 110, plus the width of the item-free sections on both sides.

In addition, the type of item 200 was determined for each of the point clouds obtained from the multiple detectors 140, but the type of item 200 may be determined after combining the multiple point clouds into one point cloud.

Although the rack 110 is exemplified as being capable of storing items 200 two-dimensionally in the horizontal and vertical directions, the rack 110 may be capable of storing items 200 one-dimensionally along the direction in which the transport cart travels.

In addition, it is possible to omit the morphology process and the noise removal process.

In addition, in the above embodiment, an example was shown in which morphological processing was performed on the entire image, but the same effect can be achieved by performing morphological processing on each of the existence sections.

The present invention can be used in automated warehouses, logistics centers, factory facilities, and other places where it is necessary to distinguish the type of item.

Reference Signs List 100 Product type determination system 110 Rack 111 Shelf 112 Support 120 Transport device 121 Rail 122 Cart 123 Mast 124 Lifting platform 130 Transfer device 140 Detector 151 Point cloud acquisition unit 152 Section detection unit 155 Edge detection unit 156 Linear shape extraction unit 157 Type determination unit 160 Product type determination device 200 Product 201 Detection surface unit

Claims (6)

1. An article type determination method for determining the type of an article based on shape characteristics of a detection surface portion of a substantially rectangular parallelepiped article, the detection surface portion including a surface facing the detector, using a detector that acquires a distance to the article as three-dimensional multiple point data, the method comprising:
a point cloud acquisition step of acquiring point data of a plurality of points on a detection surface of the article as a point cloud;
an edge detection step of detecting a pair of opposing edges on a detection surface of the article based on the point cloud acquired in the point cloud acquisition step;
a line shape extraction step of extracting a line shape having linearity extending in a direction along the pair of edges detected in the edge detection step, based on a group of points existing between the pair of edges detected in the edge detection step;
a type determination step of determining a type of the article based on the straight line shape extracted by the straight line shape extraction step;
A method for determining the type of an item, comprising: In the edge detection step,
Detecting a contour of a detection surface portion of an article based on the point cloud acquired in the point cloud acquisition step;
In the straight line shape extraction step,
extracting a straight line shape based on a group of points present inside the contour detected by the edge detection step;
The method for determining the type of an article according to claim 1 . In the edge detection step,
projecting the point cloud acquired in the point cloud acquisition step in a depth direction along a normal direction of the detection surface, converting the data in the depth direction into a gradation to generate a gradated projection image; detecting an edge of an article based on the generated gradated projection image;
In the straight line shape extraction step,
Extracting a straight line shape based on the gradation-added projection image;
The method for determining the type of an article according to claim 1 or 2. In the straight line shape extraction step,
extracting straight line shape candidates that are candidates for a portion corresponding to a shape having linearity based on the gradation-added projection image, removing noise from the straight line shape candidates, and then extracting the straight line shape;
The method for determining the type of an article according to claim 3. An article type determination method described in any one of claims 1 to 4, wherein if a straight line shape is not extracted in the straight line shape extraction process, in the type determination process it is determined that the detection surface of the article is a first type of article having a flat surface texture, and if a straight line shape is extracted in the straight line shape extraction process, it is determined in the type determination process that the detection surface of the article is a second type of article having unevenness. A detector that acquires a distance to an article as three-dimensional multiple point data;
an article type determination device that determines the type of the article based on shape features of a detection surface portion of the article, the detection surface portion including a surface facing the detector;
The product type determination device includes:
a point cloud acquisition unit that acquires point data of a plurality of points on a detection surface of an article as a point cloud;
an edge detection unit that detects a pair of opposing edges of a detection surface of the article based on the point cloud acquired by the point cloud acquisition unit;
a straight line shape extraction unit that extracts a straight line shape having linearity extending in a direction along the pair of edges detected by the edge detection unit based on a group of points existing between the pair of edges detected by the edge detection unit;
a type determination unit that determines a type of the article based on the straight line shape extracted by the straight line shape extraction unit;
An item type determination system comprising:

Images