JP6958810B2 - Shape recognition program and shape recognition device for leaf-shaped crops - Google Patents

Description

The present invention relates to a shape recognition program for recognizing the shape of a leaf-shaped crop and a shape recognition device.

Conventionally, a system for selecting and packaging leaf-shaped crops has been provided. For example, in Patent Document 1, the size of a leaf-shaped crop is determined from image data obtained by imaging the leaf-shaped crop, and the size of the leaf-shaped crop is determined based on the determination result. A program that operates the system to perform the above work has been introduced.

Japanese Unexamined Patent Publication No. 2007-331782

However, since the orientation of the leaf-shaped crops shown in the inspection image is not always constant, if the shape and size determination results of the leaf-shaped crops vary (error), the work such as sorting and packaging in the system is correct. It may not be done. Therefore, when handling foliage crops, it is important to accurately recognize the shape of the foliage crops.

Therefore, in view of such circumstances, it is an object of the present invention to provide a shape recognition program and a shape recognition device for leafy crops that can accurately recognize the shape of the leafy crops.

The shape recognition program for leafy crops of the present invention
Computer,
Area setting process for setting a scanning area including at least a curved area curved so as to bulge in the out-of-diameter direction around the center of gravity of the leaf-shaped crop in order to extract a leaf stalk region from the image of the leaf-shaped crop reflected in the inspection image. From the scanning area set in the setting process, the first extraction process for extracting the overlapping area overlapping with the leaf-shaped agricultural product, and the short and small areas having a predetermined length or less are further extracted from the overlapping area extracted in the first extraction process. After each process of the second extraction process is repeatedly executed a plurality of times while expanding the scanning area so that the curved area is displaced in the out-of-diameter direction about the center of gravity, the second extraction process is performed. A peduncle region extraction means for extracting the peduncle region by executing a region merging process for combining a plurality of short and small regions extracted in step 1.
By removing the petiole region extracted by the petiole region extracting means from the image of the leaf-shaped crop, the leaf blade region is extracted as a leaf blade region extracting means for extracting the leaf blade region of the leaf-shaped crop.

According to the shape recognition program having the above configuration, the petiole region and the leaf blade region, that is, the petiole shape and the leaf blade shape are separately extracted from the image of the leaf-shaped crop reflected in the inspection image, thereby producing the leaf-shaped crop. You will be able to recognize the shape accurately.

In addition, the shape recognition program for leaf-shaped crops of the present invention
The computer
Using the information about the petiole region extracted by the petiole region extracting means and the information about the leaf blade region extracted by the leaf blade region extracting means, it is made to function as a direction deriving means for deriving the orientation of the leaf-shaped agricultural product. ,
The orientation derivation means derives the center of gravity of the petiole region extracted by the petiole region extraction means and the center of gravity of the leaf blade region extracted by the petiole region extraction means, and derives the center of gravity of the petiole region and the center of gravity of the leaf blade region. The direction of the straight line connecting the center of gravity of the leaf blade region may be determined to be the direction of the leaf-shaped agricultural product.

The orientation of the petiole crop is deeply related to the orientation and position of the petiole. Since it is configured to derive the above, the orientation of the foliage crop can be detected after adding the information about the petiole region.

Further, since the direction of the straight line connecting the center of gravity of the petiole region and the center of gravity of the leaf blade region is determined as the orientation of the leaf-shaped crop, the orientation of the leaf-shaped crop can be easily obtained.

The shape recognition device for leaf-shaped crops of the present invention
It is equipped with a processing device that executes a process for recognizing the shape of the leaf-shaped crop from the inspection image showing the leaf-shaped crop.
The processing device is
An area setting process for setting a scanning area including at least a curved area that bulges outward in the radial direction around the center of gravity of the foliage crop in order to extract a foliage region from the image of the foliage crop reflected in the inspection image. A first extraction process for extracting an overlapping area overlapping the leaf-shaped agricultural product among the scanning areas set in the area setting process, and a short area having a predetermined length or less is further extracted from the overlapping area extracted by the first extraction process. After each process of the second extraction process is repeated a plurality of times while expanding the scanning area so that the curved area is displaced in the out-of-diameter direction about the center of gravity, the second extraction process is performed. A peduncle region extraction means for extracting the peduncle region by executing a region merging process for combining a plurality of short and small regions extracted in the process.
It has a leaf blade region extraction means for extracting a leaf blade region of the leaf crop by removing the petiole region extracted by the petiole region extraction means from the image of the leaf crop.

According to the shape recognition device having the above configuration, the petiole region and the leaf blade region, that is, the petiole shape and the leaf blade shape are separately extracted from the image of the leaf-shaped crop displayed on the inspection image, thereby producing the leaf-shaped crop. You will be able to recognize the shape accurately.

Further, the shape recognition device for leaf-shaped crops of the present invention is
It has a direction deriving means for deriving the orientation of the leaf-shaped agricultural product by using the information about the petiole region extracted by the petiole region extracting means and the information about the leaf blade region extracted by the leaf blade region extracting means.
The orientation derivation means derives the center of gravity of the petiole region extracted by the petiole region extraction means and the center of gravity of the leaf blade region extracted by the petiole region extraction means, and derives the center of gravity of the petiole region and the center of gravity of the leaf blade region. The direction of the straight line connecting the center of gravity of the leaf blade region may be determined as the direction of the leaf-shaped agricultural product.

The orientation of the petiole crop is deeply related to the orientation and position of the petiole. Since it is configured to derive the direction of the foliage crop, it is possible to derive the orientation of the foliage crop after adding the information on the petiole region.

Further, since the direction of the straight line connecting the center of gravity of the petiole region and the center of gravity of the leaf blade region is determined as the orientation of the leaf-shaped crop, the orientation of the leaf-shaped crop can be easily obtained.

As described above, according to the shape recognition program and the shape recognition device of the leaf-shaped crop of the present invention, it is possible to obtain an excellent effect that the shape of the leaf-shaped crop can be accurately recognized.

FIG. 1 is a schematic plan view of a shipping system including a shape recognition device according to an embodiment of the present invention. FIG. 2 is a side schematic view of a shipping system including the shape recognition device according to the embodiment. FIG. 3 is a block diagram of the hardware configuration of the shape recognition device according to the embodiment. FIG. 4 is a plan view of the stage of the shape recognition device according to the embodiment. FIG. 5 is a functional block diagram of the shape recognition device according to the embodiment. In FIG. 6, (a) is an explanatory diagram of an inspection image, and (b) is an explanatory diagram of an inspection image in a state where a background region and a leaf region are extracted. 7A and 7B are explanatory views of a state in which a scanning area is set in the inspection image, and FIG. 7B is an explanatory view of a state in which an overlapping area is extracted from the inspection image by the scanning area. In FIG. 8, (a) is an explanatory diagram of a state in which an expanded scanning region is set in an inspection image, and (b) is an explanatory diagram of a state in which an overlapping region is extracted from the inspection image by the scanning region. 9A and 9B are explanatory views of a state in which a further expanded scanning area is set in the inspection image, and FIG. 9B is an explanatory view of a state in which an overlapping area is extracted from the inspection image by the scanning area. In FIG. 10, (a) is an explanatory diagram of a state in which short and small regions are connected, and (b) is an explanatory diagram of a state in which a petiole region is extracted. FIG. 11 is an explanatory diagram of a process of deriving the orientation of the leaf-shaped crop by the shape recognition device according to the embodiment. FIG. 12 is a main flowchart for explaining the operation of the shape recognition device according to the embodiment. FIG. 13 is a sub-flow chart of the leaf region extraction process by the leaf region extraction means of the shape recognition device according to the embodiment. FIG. 14 is a sub-flow chart of the petiole area extraction process by the petiole area extraction means of the shape recognition device according to the embodiment. FIG. 15 is a sub-flow chart of the direction derivation process of the perilla by the direction derivation means of the shape recognition device according to the same embodiment.

Hereinafter, the shape recognition program and the shape recognition device for the leaf-shaped crop according to the embodiment of the present invention will be described with reference to the attached drawings.

The shape recognition program is configured to function as a shape recognition device for leaf-shaped crops for identifying the shape of the leaf-shaped crops based on an inspection image obtained by imaging the leaf-shaped crops.

Further, the shape recognition device is incorporated into, for example, a leaf-shaped crop shipping system for inspecting and preparing for shipping of the leaf-shaped crop.

In the present embodiment, the shape recognition device and the shape recognition program will be described by taking as an example the fact that the shape recognition device is incorporated in the shipping system. The leaf-shaped crop according to the present embodiment is a large leaf, but a leaf-shaped crop of another species may be used as long as the leaf has a thickness smaller than the plane size.

As shown in FIGS. 1 and 2, the shipping system 1 includes an accommodating tool 2 for accommodating a plurality of perilla, a shape recognition device 3 for recognizing the shape of the perilla taken out from the accommodating tool 2, and the same. It is provided with a transport device 4 for transporting the perilla whose shape has been recognized by the shape recognition device 3, and a processing device 5 for performing processing for acquiring and shipping the perilla transported by the transport device 4.

The accommodating tool 2 is configured to accommodate a plurality of large leaves in a stacked state.

As shown in FIGS. 2 and 3, the shape recognition device 3 includes a stage 30 on which the perilla to be imaged is placed, an image pickup device 31 for imaging the perilla mounted on the stage 30, and the image pickup device. A processing device 32 for recognizing the shape of the perilla based on an inspection image which is an image of the perilla captured by 31 and a storage device 33 connected to the processing device 32 by wire or wirelessly are provided. There is.

On the stage 30, the perilla leaves taken out from the container 2 are placed one by one. In the present embodiment, the container is provided by a suction transport unit (hereinafter referred to as a first suction transport unit) P1 that sucks and transports the perilla located at the uppermost part of the plurality of large leaves stacked in the container 2. It is carried from 2 to stage 30.

The suction transport unit holds the perilla by sucking one of the front surface or the back surface of the perilla that faces upward, and when the suction state for the perilla is released, the perilla leaves (falls) from the suction transport unit. It is configured in.

Further, the stage 30 according to the present embodiment has a transparent mounting plate portion 300, and the perilla leaves taken out from the accommodating tool 2 are placed on the mounting plate portion 300 (see FIG. 4).

The image pickup apparatus 31 is arranged on the lower side of the stage 30, specifically, on the lower side of the mounting plate portion 300. Then, the image pickup apparatus 31 is configured to take an image of the perilla placed on the mounting plate portion 300 from one side (the front side in the present embodiment) facing the lower side of the front side or the back side. There is.

The image pickup apparatus 31 is arranged on the upper side of the stage 30 (mounting plate portion 300), and the large leaves mounted on the mounting plate portion 300 are viewed from one side of the front surface or the back surface facing the upper side. It may be configured to image.

In the shape recognition device 3, the processing device 32 is configured to execute various means for recognizing the shape of the perilla by executing the shape recognition program.

Specifically, the shape recognition device 3 executes the shape recognition program on the processing device 32, and as shown in FIG. 5, the image acquisition means 60 for acquiring the inspection image, the entire region where the large leaves are reflected from the inspection image. Leaf region extraction means 61 for extracting (hereinafter referred to as leaf region), petiole region extraction means 62 for extracting petiole region (hereinafter referred to as petiole region) of large leaves based on inspection images, large leaves based on inspection images It is configured to function as a leaf blade region extracting means 63 for extracting a leaf blade region (hereinafter referred to as a leaf blade region) and a direction deriving means 64 for deriving the orientation of a large leaf based on the petiole region and the petiole region. Has been done.

The image acquisition means 60 is configured to acquire an image in which a large leaf to be recognized for shape is reflected. The image acquisition means 60 according to the present embodiment controls the image pickup device 31 so as to image the perilla mounted on the mounting plate portion 300, and further, the image of the perilla image captured by the image pickup device 31 is used as an inspection image. It is configured to get. That is, the image acquisition means 60 is configured to acquire an inspection image directly from the image pickup apparatus 31.

As shown in FIGS. 6A and 6B, the leaf region extraction means 61 includes the leaf region RE1 from the inspection image T and the region outside the leaf region RE1 (that is, the region in which the background is reflected). It is configured to distinguish it from the BE. In the present embodiment, the region BE outside the leaf region RE1 is referred to as the background region BE, and the following description will be given.

More specifically, the leaf region extraction means 61 executes a background region identification process for identifying the background region BE from the inspection image T and a leaf region identification process for identifying the leaf region RE1 from the inspection image T. It is configured.

In the background region identification process, the leaf region extraction means 61 selects the background region from the entire inspection image T based on the background region identification information that serves as a reference for determining whether or not each region of the inspection image T is a background. It is configured to identify the BE.

For example, information (color information) indicating a color included in the background area is set in the background area identification information, and the leaf area extraction means 61 uses the color information of the background area identification information in the inspection image T as the color information. The area having the corresponding color may be configured to be identified as the background area BE.

When setting the color information in the background area identification information, it is possible to determine the color information based on the color included in the image of the stage 30 in which the perilla is not placed. Further, the color information may be information indicating a single color or information indicating a range of a color gamut.

Further, the leaf region extraction means 61 determines the leaf region based on the leaf region identification information as a reference for determining whether or not the large leaf is reflected in each region of the inspection image T in the leaf region identification process. It is configured to identify RE1.

For example, information indicating a color (color information) is set in the leaf region identification information, and the leaf region extraction means 61 has a color corresponding to the color information possessed by the leaf region identification information in the inspection image T. The area may be configured to identify the area as the background area BE.

The color information to be set in the leaf region identification information can be set based on, for example, an image obtained by capturing a large leaf. In the present embodiment, the leaf region identification information is based on an inspection image T obtained by imaging a large leaf with an imaging device 31. It is set.

Further, the leaf region extraction means 61 according to the present embodiment has different colors for the region identified as the background region BE in the background region identification process and the region identified as the leaf region RE1 in the leaf region identification process. It is configured as shown. That is, the leaf region extraction means 61 is configured to color-code the leaf region RE1 of the inspection image T and the background region BE with different colors.

The petiole region extraction means 62 sets a scanning region including at least a curved region curved so as to bulge in the out-of-diameter direction around the center of gravity of the leaf-shaped agricultural product, a preset scanning region and a leaf region RE1. A first extraction process for extracting an overlapping area that overlaps with and a second extraction process for further extracting a short area having a predetermined length or less from the overlapping area extracted by the first extraction process, which are set in advance. It is combined by an expansion process that executes the first extraction process after expanding the scanning area, an area combination process that combines a plurality of short and small areas extracted in the second extraction process to create a combined area, and an area combination process. It is configured to execute the selection process of selecting the area having the longest dimension among the areas.

As shown in FIG. 7A, the region setting process according to the present embodiment is an annular shape centered on the center of gravity C1 of the leaf region RE1 (specifically, a circle having a constant curvature over the entire circumference in the circumferential direction). It is configured to set the scanning region PE1 (annular).

As described above, the first extraction means is configured to extract the region where the scanning region PE1 and the leaf region RE1 overlap as the overlapping region PE2. Note that FIG. 7B illustrates a state in which a plurality of overlapping regions PE2 are extracted on one inspection image T. Further, the preset scanning area is a scanning area that has already been set when the first extraction means is executed (in the present embodiment, the scanning area set in the area setting process or an extension). It is a scanning area expanded by processing).

The second extraction process is executed following the first extraction process when the first extraction process extracts the overlapping region PE2.

The plurality of overlapping regions PE2 include a long overlapping region PE2 and a short overlapping region PE2, but the petiole region extraction means 62 sets the short overlapping region PE2 as a short and small region in the second extraction process. It is configured to be extracted as PE3. The length as a reference for determining whether or not the short region PE3 is present can be appropriately changed, but is preferably set within the range of 2 mm to 5 mm.

The extended process is configured to be executed following the second extraction process. Further, in the expansion process, as shown in FIGS. 7 (a), 8 (a), and 9 (a), the size of the scanning region PE1 set in advance is set, and the curve region is the center of gravity of the leaf region RE1. It is configured to be expanded and reset so as to be displaced in the out-of-diameter direction centered on C1. In the expansion process according to the present embodiment, only the diameter of the scanning region PE1 is expanded without changing the width of the scanning region PE1.

The region binding process is configured to be executed following the second extraction process when the overlapping region PE2 is not extracted by the second extraction process. The petiole region extraction means 62 repeatedly executes the first extraction process, the second extraction process, and the expansion process until, for example, the entire scanning region PE1 expanded by the expansion process is located outside the leaf region RE1. It suffices if it is configured in.

As shown in FIG. 10A, the joining process is configured to superimpose and display the short and small region PE3 extracted a plurality of times by the second extraction process on one image. In the present embodiment, the short and small region PE3 displayed overlaid on one image by the binding process is referred to as a binding region PE4.

The inspection image T shown in FIG. 10A shows a plurality of binding regions PE4, and the plurality of binding regions PE4 include a binding region PE4 composed of a single short and small region PE3. A binding region PE4 composed of a plurality of short and small regions PE3 continuous is included.

The selection process is configured to be executed following the area join process. As shown in FIG. 10B, the petiole region extracting means 62 is configured to select the region having the longest dimension, that is, the petiole region RE2, among the binding regions PE4 in the selection process.

The leaf blade region extraction means 63 is configured to extract the leaf blade region RE3 based on the leaf region RE1 and the petiole region RE2 extracted by the petiole region extraction means 62. In the present embodiment, the leaf blade region RE3 is extracted from the leaf blade region RE1. The region corresponding to the petiole region RE2 is cut out, and the remaining region is used as the leaf blade region RE3.

As a result, the petiole region RE2 and the leaf blade region RE3 constituting one leaf region RE1 are separately extracted (see FIG. 11). That is, the shape of the petiole and the shape of the leaf blade are recognized separately.

As shown in FIG. 11, the orientation deriving means 64 derives the center of gravity C2 of the petiole region RE2, derives the center of gravity C3 of the leaf blade region RE3, and sets the direction of the straight line L connecting the respective centroids C2 and C3 to the direction of the perilla. It is configured to be.

The transport device 4 according to the present embodiment is configured to transport the perilla whose shape has been recognized by the shape recognition device 3 to a place where it can be delivered to the processing device 5. The delivery of the perilla leaves from the shape recognition device 3 to the transfer device 4 is also performed by the suction transfer unit (hereinafter referred to as the second suction transfer unit) P2 (see FIG. 1).

The processing device 5 is configured to perform processing for acquiring and shipping the perilla transported by the transport device 4. The perilla leaves are also delivered from the transport device 4 to the processing device 5 by the suction transport unit (hereinafter referred to as the third suction transport unit) P3 (see FIG. 1).

In the present embodiment, the processing for shipping the perilla is the processing of stacking a plurality of perilla leaves into a bundle, and more specifically, the processing of binding the petiole portion in a state where the plurality of perilla leaves are stacked. However, for example, packaging of perilla leaves and processing such as tagging can be mentioned as an example.

The configuration of the shipping system 1 according to the present embodiment is as described above. Subsequently, the shape recognition process of the perilla leaves (that is, the operation of the shape recognition device 3 when the shape recognition program is executed) by the shape recognition device 3 in the shipping system 1 will be described.

In the shipping system 1, after the shape recognition device 3 recognizes the shape of the perilla taken out from the container 2, the perilla is transported to the processing device 5 by the transport device 4.

As shown in FIG. 12, the shape recognition device 3 acquires the inspection image T by the image acquisition means 60 (S1). In the present embodiment, the image of the perilla placed on the mounting plate portion 300 from the accommodating tool 2 captured by the imaging device 31 is acquired as the inspection image T. Further, in the present embodiment, after the inspection image T is acquired, the leaf region identification information is set based on the inspection image T.

Then, the leaf region extraction means 61 extracts the background region BE and the leaf region RE1 (S2). In the leaf region extracting means 61, as shown in FIG. 13, the background region identification information is selected (S20), and the region of the color corresponding to the color information of the selected background region identification information is identified as the background region BE (S). S21). Further, the leaf region extraction means 61 identifies the region corresponding to the color information of the leaf region identification information as the leaf region RE1 (S22).

As a result, the inside of the inspection image T is divided (color-coded) into a leaf region RE1 and a background region BE.

Subsequently, as shown in FIG. 12, the petiole region extraction means 62 extracts the petiole region RE2 (S3). In the process of extracting the petiole region RE2 by the petiole region extracting means 62, the scanning region PE1 is set on the inspection image T in the region setting process as shown in FIG. 14 (S30). In the present embodiment, the annular scanning region PE1 centered on the center of gravity C1 of the leaf region RE1 is set in the region setting process (see FIG. 7A).

Next, when the overlapping region PE2 is extracted as a result of performing the first extraction processing (Yes in S31), the second extraction processing further extracts the short region PE3 from the overlapping region PE2 (S32).

Then, after the second extraction process is executed, the expansion process expands the scanning region PE1 (S33). More specifically, the expansion process expands only the diameter of the scanning region PE1 without changing the width of the scanning region PE1.

Further, after the expansion process is executed, the first extraction process and the second extraction process are executed again using the new scanning region PE1.

The first extraction process, the second extraction process, and the extension process are repeatedly executed as long as the overlapping region PE2 is extracted by the first extraction process. When the overlapping region PE2 is not extracted in the first extraction process (No in S31), the plurality of short and small region PE3s extracted in the second extraction process are superimposed on one image in the region combination process. By displaying it, it is designated as a binding region PE4 (S34), and subsequently, in the selection process, the region having the longest dimension is extracted as a petiole region RE2 from the plurality of binding regions PE4 (S35).

As a result, the petiole region RE2 is extracted from the leaf region RE1. That is, the petiole region RE2 (the shape of the petiole) and the leaf blade region RE3 (the shape of the leaf blade) are recognized separately.

After the petiole region extraction means 62 extracts the petiole region RE2, the leaf blade region extraction means 63 extracts the leaf blade region RE3 as shown in FIG. 12 (S4).

The leaf blade region extracting means 63 cuts out a region corresponding to the petiole region RE2 from the leaf region RE1 and sets the remaining region as the leaf blade region RE3. As a result, the leaf blade region RE3 is also extracted from the leaf region RE1, and the petiole of the large leaf and the shape of the leaf blade are recognized (specified).

Subsequently, the orientation deriving means 64 derives the orientation of the perilla (S5). As shown in FIG. 15, the orientation deriving means 64 derives the center of gravity C2 of the petiole region RE2 (S50), derives the center of gravity C3 of the leaf blade region RE3 (S51), and connects the respective centroids C2 and C3. The orientation is the orientation of the perilla (S52).

As described above, according to the shape recognition device 3 according to the present embodiment, the petiole region RE2 and the leaf blade region RE3, that is, the petiole shape and the leaf blade shape, are obtained from the image of the large leaf of the leaf-shaped agricultural product reflected in the inspection image T. By extracting and separately, the shape of the perilla can be accurately recognized.

Further, in the present embodiment, since the scanning region PE1 is annular, the binding region PE4 to which the short and small regions PE3 included in the overlapping region PE2 are bound always includes a region to be the petiole region RE2. ing.

Further, since the outer peripheral edge of the leaf blade of the large leaf has a shape in which a plurality of sharply extending sharp portions are lined up, the short and small region PE3 includes the one corresponding to a part of the petiole region RE2 in addition to the one corresponding to a part of the petiole region RE2. A leaf blade that corresponds to a sharp portion is also included, but in the present embodiment, a process of excluding a long region from the overlapping region PE2 and a region having the longest dimension among the binding regions PE4 are selected. By performing the treatment, only the petiole region RE2 can be extracted with high accuracy.

Further, in the shape recognition device 3 according to the present embodiment, the orientation deriving means 64 is configured to derive the orientation of the perilla, so if the information on the orientation of the perilla derived by the orientation deriving means 64 is utilized, the shape is formed. It is possible to improve the accuracy of work such as transportation and processing performed on the perilla whose shape is recognized by the recognition device 3.

Since the shipping system 1 of the present embodiment includes a processing device 5 for binding the petiole portion in a state where a plurality of large leaves are stacked, the accuracy of processing for binding the large leaves whose shape is recognized by the shape recognition device 3 can be improved. Can be enhanced.

The shape recognition program and shape recognition device of the present invention are not limited to the above-described embodiment, and it goes without saying that various changes are made without departing from the gist of the present invention.

In the shipping system 1 of the above embodiment, the perilla that has recognized the shape by the shape recognition device 3 is configured to be conveyed to the transfer device 4, but the present invention is not limited to this configuration. For example, the shipping system 1 may be configured to directly convey the perilla whose shape has been recognized by the shape recognition device 3 to the processing device 5 or other devices.

In the above embodiment, the shape recognition program is configured to image the perilla with the image pickup device 31, but it may not be configured to control the image pickup device 31 as long as the inspection image can be acquired. ..

In the above embodiment, the image acquisition means 60 is configured to acquire an inspection image directly from the image pickup apparatus 31, but is not limited to this configuration. For example, the image acquisition means 60 may be configured to store the captured inspection image in the storage device 33 and read the inspection image from the storage device 33 as needed, or may be configured with respect to the shape recognition device 3. It may be configured to read out the inspection image stored in the external storage device connected by wire and the inspection image stored in the external storage device capable of wirelessly communicating with the shape recognition device 3.

Although not particularly mentioned in the above embodiment, the shape recognition program may be stored in the storage device 33 of the shape recognition device 3, and if the processing device 32 is accessible, the shape recognition program is wired to the shape recognition device 3. It may be stored in a connected external storage device, or may be stored in an external storage device capable of wirelessly communicating with the shape recognition device 3.

Further, in the above embodiment, the background area identification information and the leaf area identification information are also stored in the storage device 33 of the shape recognition device 3, but if the processing device 32 is accessible, they are connected to the shape recognition device 3 by wire. It may be stored in an external storage device or an external storage device capable of wirelessly communicating with the shape recognition device 3.

In the above embodiment, the region setting process is configured to set the annular scanning region PE1, but the region setting process is not limited to this configuration. For example, the area setting process may be configured to set the elliptical annular scanning area PE1 or may be configured to set the curved scanning area PE1 composed of only the curved area. ..

1 ... Shipping system, 2 ... Accommodator, 3 ... Shape recognition device, 4 ... Transfer device, 5 ... Processing device, 30 ... Stage, 31 ... Imaging device, 32 ... Processing device, 33 ... Storage device, 60 ... Image acquisition means , 61 ... Leaf region extraction means, 62 ... Petiole region extraction means, 63 ... Leaf blade region extraction means, 64 ... Direction derivation means, 300 ... Placement plate portion, BE ... Background area, C1 ... Center of gravity, C2 ... Center of gravity, C3 ... Center of gravity, L ... Straight line, PE1 ... Scanning area, PE2 ... Overlapping area, PE3 ... Short area, PE4 ... Bonding area, RE1 ... Leaf area, RE2 ... Petiole area, RE3 ... Leaf blade area, T ... Inspection image

Claims (4)

Computer,
Area setting process for setting a scanning area including at least a curved area curved so as to bulge in the out-of-diameter direction around the center of gravity of the leaf-shaped crop in order to extract a leaf stalk region from the image of the leaf-shaped crop reflected in the inspection image. From the scanning area set in the setting process, the first extraction process for extracting the overlapping area overlapping with the leaf-shaped agricultural product, and the short and small areas having a predetermined length or less are further extracted from the overlapping area extracted in the first extraction process. After each process of the second extraction process is repeatedly executed a plurality of times while expanding the scanning area so that the curved area is displaced in the out-of-diameter direction about the center of gravity, the second extraction process is performed. A peduncle region extraction means for extracting the peduncle region by executing a region merging process for combining a plurality of short and small regions extracted in step 1.
A shape recognition program for a leaf-shaped crop that functions as a leaf blade region extracting means for extracting a leaf blade region of the leaf-shaped crop by removing the petiole region extracted by the petiole region extracting means from the image of the leaf-shaped crop. The computer
Using the information about the petiole region extracted by the petiole region extracting means and the information about the leaf blade region extracted by the leaf blade region extracting means, it is made to function as a direction deriving means for deriving the orientation of the leaf-shaped agricultural product. ,
The orientation derivation means derives the center of gravity of the petiole region extracted by the petiole region extraction means and the center of gravity of the leaf blade region extracted by the petiole region extraction means, and derives the center of gravity of the petiole region and the center of gravity of the leaf blade region. The shape recognition program for a leaf-shaped agricultural product according to claim 1, wherein the direction of a straight line connecting the center of gravity of the leaf blade region is determined to be the direction of the leaf-shaped agricultural product. It is equipped with a processing device that executes a process for recognizing the shape of the leaf-shaped crop from the inspection image showing the leaf-shaped crop.
The processing device is
An area setting process for setting a scanning area including at least a curved area that bulges outward in the radial direction around the center of gravity of the foliage crop in order to extract a foliage region from the image of the foliage crop reflected in the inspection image. A first extraction process for extracting an overlapping area overlapping the leaf-shaped agricultural product among the scanning areas set in the area setting process, and a short area having a predetermined length or less is further extracted from the overlapping area extracted by the first extraction process. After each process of the second extraction process is repeated a plurality of times while expanding the scanning area so that the curved area is displaced in the out-of-diameter direction about the center of gravity, the second extraction process is performed. A peduncle region extraction means for extracting the peduncle region by executing a region merging process for combining a plurality of short and small regions extracted in the process.
A shape recognition device for a leaf-shaped crop having a leaf blade region extracting means for extracting a leaf blade region of the leaf-shaped crop by removing the petiole region extracted by the petiole region extracting means from the image of the leaf-shaped crop. It has a direction deriving means for deriving the orientation of the leaf-shaped agricultural product by using the information about the petiole region extracted by the petiole region extracting means and the information about the leaf blade region extracted by the leaf blade region extracting means.
The orientation derivation means derives the center of gravity of the petiole region extracted by the petiole region extraction means and the center of gravity of the leaf blade region extracted by the petiole region extraction means, and derives the center of gravity of the petiole region and the center of gravity of the leaf blade region. The shape recognition device for a leaf-shaped agricultural product according to claim 3, wherein the direction of a straight line connecting the center of gravity of the leaf blade region is determined to be the direction of the leaf-shaped agricultural product.

Images