JP7320260B2 - Pedal localization device and sawing system - Google Patents

Description

The present invention relates to a pulp position estimating device for estimating the position of a pulp, and a sawing system provided with the pulp position estimating device and a sawing machine.

When lumbering raw wood cut from forests, wood from which the bark has been stripped from the raw wood (hereinafter collectively referred to as logs), and flitches which have been roughly sawn into logs, the growth It is often necessary to locate the pith, which is the center of the annulus. In the following description, annual rings, which are typical growth rings, are used as representative examples, and wood with a remaining pith, including logs and flitches, is referred to as core wood. When sawing the core wood, for example, it is necessary to specify the position of the pith to decide whether to saw the pith-containing so-called core lumber or the pith-free lumber as the so-called heart-free lumber. In addition, lumber products tend to warp or bend depending on the positional relationship with the pith, or the material differs, which affects the yield and material price, so it is necessary to specify the pith position.

Furthermore, when cutting a piece of core wood several times, when changing the direction of the core wood in the circumferential direction between cuttings, the position of the rotation center and the pith of the core wood is often misaligned. In that case, it is often rational to specify the position of the pith and set it as the center of rotation. Patent Literature 1 proposes estimating the position of the pith by automatically detecting the normal direction of annual rings on a photographed image instead of specifying the position with the naked eye of a worker.

JP 2011-88436 A

However, since tree rings contain nodes, it is sometimes difficult to accurately extract tree rings as circular arcs. It may not be possible to estimate accurately.

SUMMARY OF THE INVENTION In view of the above circumstances, it is an object of the present invention to provide a pulp position estimation apparatus capable of accurately estimating the position of the pulp, and a sawing system capable of performing sawing according to the accurately estimated position of the pulp. do.

The apparatus for estimating the position of the spinal cord of the present invention, which solves the above object,
a learning model storage unit for storing a learning model constructed by supervised learning based on area image data representing a target area including the pith on the butt end surface and position data specifying the position of the pith in the area image data; ,
an acquisition unit that acquires target image data representing all or part of the butt surface;
an estimating unit that uses the learning model to estimate the position of the pith on the butt surface of the target image data from the target image data acquired by the acquiring unit ;
The position data is data representing the size of an area that includes the pith and is narrower than the target region and the position of the area with respect to the target region,
A predetermined positional relationship is established between the area and the pith contained in the area,
The estimating unit performs object detection based on the area from the target image data acquired by the acquiring unit, and estimates the position of the pith based on the predetermined positional relationship,
The predetermined positional relationship is a positional relationship in which the position of the pith coincides with the center position of the area .

The target area may be an area of the entire butt surface, or a partial area including the pith on the butt surface.

Further, the area image data may be a plurality of area image data obtained by rotating the butt surface in the circumferential direction and photographing the butt surface at different rotation angle positions.

The supervised learning is learning based on teacher data consisting of a set of the area image data and the position data. That is, the learning is based on teacher data obtained by adding the information of the position data to the area image data.

According to the spinal cord position estimating apparatus of the present invention, the estimating unit can accurately estimate the position of the ankle.

The positional relationship in which the position of the pith coincides with the central position of the area simplifies the estimation of the position of the pith.

Note that the positional relationship may be such that the position of the pith coincides with a position shifted by a predetermined distance in a predetermined direction from the center position of the area.

Further, a second estimation unit may be provided for estimating the density of annual rings or the average interval of annual rings (average width of annual rings) up to the n-th tree around the position of the pith estimated by the estimation unit. For example, the learning model storage unit stores, in addition to position data obtained by estimating the position of the pith, a learning model constructed by supervised learning based on tree-ring data relating to tree-ring density or average tree-ring width, The second estimation unit may estimate the density of tree rings or the average width of tree rings from the target image data acquired by the acquisition unit. Regarding the density of tree rings, it may be estimated to which group it belongs to among a plurality of groups regarding density of tree rings (eg, high density, normal, low density). It may be estimated to which of a plurality of groups (for example, the average width of tree rings is wide, normal, or narrow). If the density of tree rings or the average width of tree rings can be estimated, then the position of the n-th tree ring can also be estimated from the estimated position of the pith. The n-th annual ring may be an annual ring (for example, the 15th annual ring) that serves as a boundary between mature and immature wood.

The sawing system of the present invention, which solves the above objects,
a holding device that holds the core wood in which the pith remains in a non-rotatable manner in the circumferential direction;
a sawing machine for cutting the core wood held by the holding device;
A learning model storage unit for storing a learning model constructed by supervised learning based on area image data representing a target area including the pith on the butt end surface and position data specifying the position of the pith in the area image data, estimating the position of the pith on the butt surface of the target image data from the target image data obtained by the obtaining unit using the acquisition unit that acquires target image data representing all or part of the butt surface and the learning model; a spinal cord position estimation device having an estimation unit ;
a photographing device for photographing a butt surface of the core wood before being held by the holding device and transmitting data of the photographed image to the acquisition unit;
a rotating device that rotates the core wood in the circumferential direction around the position of the pith estimated by the estimating unit and adjusts the sawing posture of the core wood on the sawing machine;
The acquisition unit acquires image data transmitted from the photographing device as the target image data, and the holding device holds the core lumber whose lumber attitude has been adjusted by the rotation device. It is characterized by

The core wood referred to here includes not only raw wood cut from the mountains, but also wood from which the bark has been stripped from the raw wood, as well as flitches that have been coarsely sawn into logs (the same shall apply hereinafter). .

According to the lumbering system of the present invention, since the pulp position estimation device of the present invention is provided, the position of the pulp can be accurately estimated. Can do lumber. In addition, by rotating the core wood in the circumferential direction around the position of the pith and arranging the sawing posture, it is possible to cut wood with good yield, rational wood cutting, and wood cutting that considers warping, bending, and material. It becomes possible. You can also decide in advance whether to mill with the pith in or without the pith.

In addition, the photographing device photographs an area of a figure circumscribing a circle whose center is aligned with the center position of the butt surface and has a predetermined ratio of 1/3 or more and 1 or less of the diameter of the butt surface, or the area of the circle. It may be something to do.

The inventor of the present application found out in what range the probability of the pith being located is high, based on the center position of the butt end surface. In other words, it has been found that if an area with a predetermined ratio of 1/3 is used as the narrowest area, the pith will not be out of the imaging area. In addition, since the predetermined ratio of 1/3 is a value with some margin, the diameter here may be the diameter in the longest direction or the average value (for example, the diameter in the vertical direction and the diameter in the horizontal direction). direction diameter).

Also, the graphic area may be a polygonal area (for example, a rectangular area or more).

Further, the target area in the apparatus for estimating the position of the pulp is also a figure area circumscribing a circle whose center is aligned with the center position of the buttress surface and has a diameter equal to the predetermined ratio of the diameter of the buttress surface, or the area of the circle. A region is preferred. In other words, it is preferable that a predetermined ratio in the imaging region imaged by the imaging device is the same as a predetermined ratio in the target region of the spinal cord position estimating device.

According to the apparatus for estimating the position of the pulp of the present invention, the position of the pulp can be accurately estimated, and according to the sawing system of the present invention, sawing can be performed according to the accurately estimated position of the pulp.

FIG. 10 is a schematic functional block diagram of a device for estimating the position of the spinal cord during machine learning; It is the figure which showed the butt end surface of two types of logs typically. FIG. 4 is a diagram showing a plurality of area image data; FIG. 3 is a schematic functional block diagram of a lumber system having a pulp position estimating device when estimating the pulp position; FIG. 5 is a side view showing a portion on one end side of the rotating device shown in FIG. 4; FIG. 5 is a front view of the band saw machine shown in FIG. 4 as viewed from the right side of FIG. 4;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.

The spinal cord position estimation device of this embodiment performs machine learning. Here, first, the machine learning performed by the spinal cord position estimation device will be described.

FIG. 1 is a schematic functional block diagram of a device for estimating the position of the spinal cord during machine learning.

The spinal cord position estimation device 10 shown in FIG. 1 includes a learning unit 11 . The learning unit 11 is provided with teacher data. The teacher data is obtained by adding position data information to the area image data. In other words, the teacher data is data obtained by annotating (tagging) information about the position of the pith found by a person to the area image data.

FIG. 2 is a diagram schematically showing the butt end surfaces of two types of logs. Both of these two types of logs have a diameter of about 300 mm to 400 mm, which is a so-called large-diameter log. Any butt surface shown in FIG. 2 is a butt surface on the base side. In FIG. 2, only the heartwood portion shown in gray shows annual rings, and the sapwood portion shown in white does not show annual rings. In the following explanation, annual rings are taken as an example, but the same applies to growth rings.

FIG. 2(a) shows the end face C1 of a log that is relatively thin and has low roundness. On the butt surface C1 shown in FIG. 2(a), knots k can be seen from the sapwood portion C12 to the heartwood portion C11. The pith (centre of annual rings) is, for example, like a dark dot, and it is easy for a person to find the position of the pith. In FIG. 2, the pith p is indicated by a cross. In the butt end surface C1 shown in FIG. 2(a), the position of the pith p is shifted downward from the center of gravity position g.

FIG. 2(b) shows the end face C2 of a log that is relatively thick and has a high roundness. In the end face C2 shown in FIG. 2(b), dirt d is seen in the core portion C21. Also, on the butt end surface C2, the position of the pith p is displaced upward from the center of gravity position g, but compared to the butt end surface C1 shown in FIG.

The area image data given to the learning unit 11 shown in FIG. 1 is image data obtained by photographing the buttress surface on the root side. This region image data is data representing a target region including the pith on the butt surface. may be used as the target area. More specifically, the target area in the area image data is a figure area circumscribing a circle whose center is aligned with the center position of the butt surface and whose diameter is a predetermined ratio of 1/3 or more and 1 or less of the diameter of the butt surface, or Any area of the circle may be used. Since this area is closely related to the imaging area described later, detailed description will be given later. Here, the target area is assumed to be the area of the entire butt surface. Further, the region image data may be image data obtained by photographing the buttress surface on the proximal side, or may be image data obtained by photographing the buttress surface on each of the proximal side and the distal side.

The training data given to the learning unit 11 is data obtained by adding information of position data specifying an area including the pith and narrower than the target area to the area image data. That is, the teacher data is data obtained by adding information on the size of the area and information on the position of the area with respect to the target area to the area image data. In FIG. 2, this area is indicated by a two-dot chain line. Furthermore, a predetermined positional relationship is established between the area and the pith included in the area. The predetermined positional relationship referred to here is a positional relationship in which the position of the pith coincides with the center position of the area. Note that the positional relationship may be such that the position of the pith coincides with a position shifted by a predetermined distance in a predetermined direction from the center position of the area. By establishing a predetermined positional relationship, when the position of the area is determined, the position of the pith is also determined. Although the shape of the area shown in FIG. 2 is square, the shape of the area is not particularly limited, and may be polygonal or circular. In addition, it is preferable that the size of the area is such that three or more annual rings are included outside the pith. In estimating the position of the pith in the estimation unit to be described later, annual rings serve as marks as patterns. On the end face C2 shown in FIG. 2(b), the size of the area used for the end face C1 shown in FIG. 2(a) does not include three or more annual rings. The area used for the butt surface C2 is larger than the area used for the butt surface C1 shown in FIG. In addition, the heartwood portion and the sapwood portion have different colors, and when trying to learn a region including a boundary where the color changes, the learning unit 11 learns a pattern called "pith" including the change in color. Therefore, it is preferable that the size of the area is a size that can be accommodated within the heartwood portion, since the color contrast and the like may affect the detection accuracy. On the other hand, if the area becomes too large, there is a high possibility that the pith will be displaced from the central position of the area. As a more specific example of the position data, when the shape of the area is rectangular, the coordinate values of the center of the area and the horizontal The data of the length (width) in the direction and the length (height) in the vertical direction are the position data.

FIG. 3 is a diagram showing a plurality of area image data.

FIG. 3 shows images of eight butt surfaces. The image of the butt surface shown in FIG. 3(a) is the same as the image of the butt surface C1 shown in FIG. 2(a). The image of the butt surface shown in FIG. 3(b) is an image obtained by rotating the image of the butt surface shown in FIG. 3(a) clockwise by 45°. The image of the butt surface shown in FIG. 4A is rotated clockwise by 90°, and the image of the butt surface shown in FIG. The image is rotated 135° around. The image of the butt surface shown in FIG. 3(e) is an image obtained by horizontally reversing the image of the butt surface shown in FIG. 3(a), and the image of the butt surface shown in FIG. The image of the butt surface shown in FIG. 3( b ) is a horizontally reversed image, and the image of the butt surface shown in FIG. 3( g ) is a horizontally reversed image of the butt surface shown in FIG. The image of the butt surface shown in FIG. 3(h) is an image obtained by horizontally reversing the image of the butt surface shown in FIG. 3(d). That is, the images shown in FIGS. 3(a) to 3(h) are images of the same butt surface C1.

In FIG. 3 as well, in each figure, the pith p is indicated by an X mark, and a square area centered on the position of the pith p is indicated by a chain double-dashed line. As shown in FIG. 3, when each region is viewed in the direction shown in FIG. 3, it can be seen that eight region image data can be prepared from one butt surface. As a result, it is possible to prepare teacher data to which information about the area indicated by the two-dot chain line is added for each of the eight area image data. In this way, eight training data are prepared from each of the large number of images of the butt end surface and given to the learning unit 11 .

The learning unit 11 performs supervised learning based on teacher data to construct a learning model. In the learning unit 11, learning proceeds by a technique such as deep learning.

The spinal cord position estimation device 10 also includes a learning model storage unit 12 , and the learning model constructed by the learning unit 11 is stored in the learning model storage unit 12 .

Next, the pith position estimating device 10 for estimating the position of the pith on the butt end surface using the constructed learning model will be described with an example of incorporating it into a sawing system.

FIG. 4 is a schematic functional block diagram of a lumber system equipped with a pulp position estimating device when estimating the pulp position. This functional block diagram shows the lumber system as seen from directly above.

The sawing system 1 shown in FIG. 4 includes, in addition to the pulp position estimation device 10 when estimating the position of the pulp, an input device 20, a conveying device 30, an imaging device 40, a rotating device 50, a material feeding device 60, a pair of clamping plates 70, and A band saw machine 80 is provided.

A plurality of logs are supplied to the loading device 20 . The loading device 20 loads the logs L one by one into the conveying device 30 as indicated by the arrow a.

The conveying device 30 conveys the input log L to a conveying position beside the photographing device 40 as indicated by an arrow b. The log L transported to the transport position is sent to the photographing device 40 as indicated by an arrow c.

A sensor for measuring the diameter of the butt surface in two orthogonal directions and a sensor for measuring the length of the log are provided between the transport position of the transport device 30 and the photographing device 40. The lumber shown in FIG. In the system 1, coordinate information (x, y, z) representing the three-dimensional center position of the butt end surface can be obtained by these sensor measurements.

The photographing device 40 has a first camera 41 and a second camera 42 for photographing the butt surfaces of both ends (base side and tail side) of the log L. As shown in FIG. The photographing device 40 uses the first camera 41 and the second camera 42 to photograph the entire butt surface, but instead of photographing the entire butt surface, only a partial area including the pith on the butt surface may be photographed. More specifically, the photographing area photographed by the photographing device 40 is a figure circumscribing a circle whose center is aligned with the center position of the butt surface and whose diameter is a predetermined ratio of 1/3 or more and 1 or less of the diameter of the butt surface. Any area or area of the circle may be used. The inventor of the present application found out in what range the probability of the pith being located is high, based on the center position of the butt end surface. In other words, it has been found that if an area with a predetermined ratio of 1/3 is used as the narrowest area, the pith will not be out of the imaging area. In addition, since the predetermined ratio of 1/3 is a value with some margin, the diameter here may be the diameter in the longest direction or the average value (for example, the diameter in the vertical direction and the diameter in the horizontal direction). direction diameter). In the range of 1/3 or more and 1 or less, the closer the ratio is to 1/3, the higher the definition of the image data can be obtained. On the other hand, the closer the ratio is to 1, the more image data with more information (eg, pattern) can be obtained. Furthermore, it is preferable that the area to be photographed here matches the target area of the area image data given to the learning unit 11 . In other words, it is preferable that the predetermined ratio of the imaging area photographed by the imaging device 40 and the predetermined ratio of the target area of the area image data given to the learning unit 11 are the same.

Of the buttress surface on the base side and the buttress surface on the distal side, the photographed image data of only the buttress surface on the base side may be transmitted to the pith position estimation device 10 .

The spinal cord position estimation device 10 shown in FIG. 4 includes an acquisition unit 13 and an estimation unit 14 in addition to the learning model storage unit 12 . The acquisition unit 13 acquires the data of the captured image transmitted from the imaging device 40 as target image data. In a preferred embodiment, the predetermined ratio in the target area of the area image data given to the learning unit 11 in the learning phase is the same as the predetermined ratio in the image area of the target image data acquired by the acquisition unit 13 in the estimation phase. Become. By doing so, it can be expected that the regions of the image data handled in the learning phase and the estimation phase will match, and the accuracy of estimation will improve.

The estimation unit 14 estimates the position of the pith on the butt end surface of the target image data acquired by the acquisition unit 13 using the learning model stored in the learning model storage unit 12 . The estimating unit 14 uses an algorithm such as YOLO (You Only Look Once) to search for an area similar to the area given as teacher data from the end face of the target image data. Although detailed explanation is omitted, this YOLO is an object detection algorithm using a convolutional neural network (CNN: Convolutional Neural Network). A rectangular area circumscribing the object (an area indicating the object) called a bounding box is set in the grid. Note that the algorithm used in the estimation unit 14 is not limited to this YOLO, and may be other object detection algorithms using a convolutional neural network, for example. Upon detecting an area from the end face of the target image data, the estimation unit 14 estimates the position of the pith based on the above-described predetermined positional relationship. That is, an estimation result is derived that the pith exists at the center position of the detected area.

Actually, 500 pairs of area image data and position data (teacher data) were prepared. A predetermined number of sets (for example, 64 sets) randomly selected from the 500 sets are regarded as one set, and 31000 sets of machine learning are performed. Here, although the method of extracting eight pieces of teacher data from one butt face, which was explained with reference to FIG. As the test data, 100 pieces of target image data obtained by respectively photographing the periphery of the pith on different butt surfaces are prepared. rice field. All of the 100 estimation results were within a range of 10 mm from the position judged by humans to be the position of the pith, and a very high percentage of correct answers was obtained. More specifically, the estimated maximum deviation was 8 mm, and the average deviation was 2 mm. Moreover, the standard deviation of the positional deviation was 1.3 mm.

Since both the learning unit 11 and the estimating unit 14 are constructed in the apparatus for estimating the position of the spinal cord 10 described above, the estimation result of the spinal cord position in the estimating unit 14 can be evaluated as to whether the position of the spinal cord is correct or not. , the photographed image data can be used as teacher data. On the other hand, the learning unit 11 and the estimation unit 14 may be provided in different devices. That is, the spinal cord position estimating apparatus 10 is not provided with the learning unit 11, and the spinal cord position can be estimated only if the learning model storage unit 12 storing the learning model is provided.

A rotating device 50, shown schematically in FIG. 4, has a control 59 and also has a rotating center 51, 52 at each of the lengthwise end portions of the log L. As shown in FIG.

5 is a side view showing one end side portion of the rotating device shown in FIG. 4. FIG. In FIG. 5, the horizontal direction of the figure is the length direction of the log, and although not shown, a band saw machine 80 is provided at a distant position on the left side of the figure. One end side of the rotating device 50 shown in FIG. 5 is the side away from the band saw board 80 (opposite to the band saw board side), and the other end side (not shown) is the band saw board side. FIG. 5 shows the controller 59, the end of the log L on the side of the base, and the rotation center 51 whose tip part bites into the butt surface C3 of the log L on the side of the base.

The rotating device 50 shown in FIG. 5 includes a material length cylinder 53 that advances and retreats in the length direction of the log L (see arrow Z), an up-down cylinder 54 that advances and retreats in the vertical direction (see arrow Y) that is the height direction, and in FIG. It has a horizontal cylinder 55 that advances and retreats in the horizontal direction perpendicular to the paper surface, and a motor 56 that rotates the rotation center 51 around the center of the rotation center 51 (see arrow R). Further, the rotating device 50 also has a sensor (not shown) for detecting the position of the butt surface. A motor is omitted from the rotating device 50 on the band saw side (not shown in FIG. 5), and a rotating center 52 is rotatably supported along with the rotation of the rotating center 51 on the one end side. Further, the material length cylinder 53 is not provided on the band saw side, but the vertical cylinder 54 and the horizontal cylinder 55 are provided.

The log photographed by the photographing device 40 shown in FIG. 4 is sent to the rotating device 50 as indicated by an arrow d. At this time, the log is sent to the rotating device 50 so that the center position of the butt surface obtained by the sensor measurement described above coincides with the origin coordinate position managed by the rotating device 50 . The log L sent to the rotating device 50 is supported by a clamping mechanism (not shown) that grips the outer peripheral surface of the log with the center position of the butt surface aligned with the coordinate position of the origin.

The estimation unit 14 of the pith position estimation device 10 transmits the estimation result to the control unit 59 of the rotation device 50 when the estimation of the pith position of the butt end surface is completed. Also, the same photographed image data as the photographed image data transmitted from the photographing device 40 to the acquisition unit 13 is transmitted to the control unit 59 . Further, coordinate information (x, y, z) indicating the central position of the photographed image data on the butt end surface obtained by the sensor measurement described above is also transmitted to the control unit 59 . The control unit 59 applies the estimation result of the position of the spinal cord to the photographed image data, and calculates the three-dimensional coordinate information of the position of the spinal cord based on the coordinate information.

The control unit 59 controls the material length cylinder 53, the vertical cylinder 54, and the horizontal cylinder 55 so that the rotation center 51 can rotate the log L in the circumferential direction around the position of the pith estimated by the estimation unit 14. Controls the amount of forward/backward movement. That is, the control unit 59 controls the advance/retreat amount of the material length cylinder 53 based on the calculated coordinate information of the position of the pith so that the tip portion of the rotation center 51 bites into the butt surface C3 on the butt end side of the log L. to Further, the control unit 59 controls the advance/retreat amount of the vertical cylinder 54 and the horizontal cylinder 55 based on the calculated coordinate information of the position of the pith so that the center of the rotation center 51 matches the position of the pith estimated by the estimation unit 14. make it

As shown in FIG. 4, the center of the butt surface C3 on the base side of the log L is aligned with the position of the pith estimated by the estimating unit 14 by controlling the forward/backward movement of each cylinder by the control unit 59 described above. The tip of the rotation center 51 bites into the end portion of the log L, and the tip of the rotation center 52 bites into the butt surface C4 on the end side of the log L with the center aligned with the position of the pith estimated by the estimation unit 14. I'm in. The rotation centers 51 and 52 described above hold the log L whose tip portion is bitten in from the extending direction of the log L so as not to rotate in the circumferential direction. The log L sandwiched by the rotation centers 51 and 52 rotates in the circumferential direction together with the rotation centers 51 and 52 as the motor 56 rotates.

In addition, since the position of the pith is estimated by the estimating unit 14, it is possible to make a lumbering plan for lumbering as a core lumber or as a heart-free lumber. In addition, it is possible to consider how to insert a saw for high yield when lumbering, that is, lumber plan for lumber with good yield, lumber plan for lumber considering warpage, bending, and material. Furthermore, it is also possible to grasp in advance how the mature lumber portion and the immature lumber portion are included in the lumber product, and to make a lumber plan. The lumbering plan may be determined by a computer program, or may be determined by an operator.

According to the sawing plan determined in this way, the rotating device 50 rotates the log in the circumferential direction by rotating the motor 56, prepares the sawing posture of the log when passing through the band saw machine 80 for the first time, and enters the standby state. .

When the lumbering that has started earlier is completed, the log L, which had been in a ready lumbering posture and was in a standby state, is clamped by a pair of clamping plates 70 shown in FIG. is moved to the material feeding device 60. In FIG. 4 , the pair of clamping plates 70 that clamp the log L in the standby state by the rotating device 50 is indicated by dotted lines, and the pair of clamping plates 70 moved to the material feeding device 60 is also indicated by dotted lines.

As shown in FIG. 4, the material feeding device 60 extends in the length direction of the log L, and the band saw machine 80 is arranged in the middle.

6 is a front view of the band saw machine shown in FIG. 4 as seen from the right side of FIG. 4. FIG. Hereinafter, the horizontal direction in FIG. 6 is referred to as the width direction.

FIG. 6 also shows a part of the material feeding device 60 extending in the direction perpendicular to the plane of the paper. The material feeding device 60 is
It has a material feed rail 61 extending in a direction perpendicular to the plane of the drawing and a vehicle body 62 running on the material feed rail 61 . This car body 62 extends in the length direction of the log. Wheels 63 are provided on both sides of the vehicle body 62 in the width direction, and a plurality of wheels are also provided at intervals in the longitudinal direction of the vehicle body 62 . These wheels 63 are engaged with the material feed rail 61, and the vehicle body 62 travels on the material feed rail 61 by driving means (not shown).

Clamps protruding upward are provided at both ends of the vehicle body 62 in the longitudinal direction. FIG. 6 shows only the clamp 64 on one end, while FIG. 4 shows clamps 64 and 65 on both ends. These clamps 64 and 65 can move back and forth in the longitudinal direction of the vehicle body 62, that is, in the direction in which the log L extends. Similar to the rotation center 51 of the rotation device 50 described with reference to FIG. , C6, and grips the log L from the extending direction of the log L in the circumferential direction so as not to be rotatable. The clamp 64 on one end is rotatable about the center of the clamp 64 by a motor (not shown), and the clamp 65 on the other end is rotatable in accordance with the rotation of the clamp 64 on the one end. pivoted.

In FIG. 4, the car body 62 is not visible due to the log L, but the car body is in its initial position. The log L clamped by the pair of clamping plates 70 and moved above the vehicle body 62 in the initial position is clamped with the center of the clamps 64 and 65 aligned with the position of the pith estimated by the estimation unit 14. clamps 64, 65 in the circumferential direction so as not to rotate. When the clamping by the clamps 64 and 65 is completed, the pair of clamping plates 70 retreat. In FIG. 4, the pair of retracted clamping plates 70 are indicated by solid lines.

A band saw machine 80 shown in FIG. It has a saw wheel unit 84 . A motor 833 is provided on the base 83 . The saw wheel unit 84 has an upper saw wheel 84a and a lower saw wheel 84b which are vertically spaced apart, and a band saw 841 wound around the upper saw wheel 84a and the lower saw wheel 84b. In FIG. 6, the saw wheel unit 84 is indicated by a dashed line. Note that the band saw machine 80 of this embodiment is of a so-called twin type in which two saw wheel units 84 are arranged side by side in the width direction.

A drive shaft 85b extending in a direction perpendicular to the plane of the drawing is non-rotatably fixed to the central portion of the lower saw wheel 84b. The driving force of a motor 833 is transmitted to the drive shaft 85b, and the drive shaft 85b is rotated by the motor 833 to rotate the lower saw wheel 84b. A fixed shaft 85a extending in a direction orthogonal to the plane of the drawing is passed through the central portion of the upper saw wheel 84a. The fixed shaft 85 a is supported by a spall 831 provided on the base 83 . The spall 831 is vertically movable, and by vertically moving the spall 831, the position of the upper saw wheel 84a is also moved vertically, and the interval between the upper saw wheel 84a and the lower saw wheel 84b can be adjusted. A weight 86 is attached to the fixed shaft 85a. This weight 86 applies an upward force to the fixed shaft 85a by the principle of leverage. By adjusting the spacing between the upper saw wheel 84a and the lower saw wheel 84b by the spall 831, and by the force applied from the weight 86, the band saw 841 wound around the upper saw wheel 84a and the lower saw wheel 84b has a predetermined It gives tension. In the saw wheel unit 84 on the left side shown in FIG. 6, the lower saw wheel 84b to which the driving force of the motor 833 is transmitted rotates clockwise as indicated by the arrow, so that the band saw 841 also travels clockwise. As the band saw 841 travels, the upper saw wheel 84a rotates clockwise as indicated by the arrow. On the other hand, in the saw wheel unit 84 on the right side shown in FIG. 6, the lower saw wheel 84b to which the driving force of the motor 833 is transmitted rotates counterclockwise as indicated by the arrow, thereby rotating the band saw 841 counterclockwise. As the band saw 841 travels, the top saw wheel 84a also rotates counterclockwise as indicated by the arrow.

The base 83 is slidable on the rail member 82 in the width direction. A cylinder 832 is arranged on the outer side in the width direction of each of the pair of bases 83, 83, and by driving the cylinder 832, each of the pair of bases 83, 83 slides in the width direction. As the base 83 slides, the saw wheel unit 84 also slides in the width direction.

An example of sawing a log L using the band saw machine 80 shown in FIG. 6 described above will be described below. The example of sawing here is an example according to a sawing plan made based on the result of estimation of the position of the pith by the estimating unit 14 .

First, the motor 833 is driven so that the saw wheel unit 84 on the left side shown in FIG. 6 runs the band saw 841 clockwise, and the saw wheel unit 84 on the right side shown in FIG. . The interval between the band saws 841, 841 arranged on the left and right is adjusted by driving the cylinder 832 and sliding the base body 83 in the width direction according to the size of the target lumber product such as square lumber according to the lumber plan. back. Next, in FIG. 6, the vehicle body 62 at the initial position, which is the front side of the paper, is run toward the band saws 841, 841 (toward the back side of the paper in FIG. 6), and the log L is passed through the band saws 841, 841 ( (see arrow g in FIG. 4). As a result, the log L is cut into a so-called drum material. After that, the space between the band saws 841, 841 arranged on the left and right is slightly widened by driving the cylinder 832, and the vehicle body 62 is returned to the initial position (see arrow h shown in FIG. 4). After the vehicle body 62 returns to the initial position, the sawing posture of the log L is arranged according to the sawing plan. In this example, the drum-shaped log L clamped between the clamps 64 and 65 is rotated 90 degrees around a rotation axis extending in a direction perpendicular to the plane of the paper, and the interval between the band saws 841 and 841 arranged on the left and right sides is adjusted. is the set interval. Subsequently, the vehicle body 62 is again driven toward the band saws 841, 841 to pass the log L through the band saws 841, 841, thereby cutting the log L in the state of drum material into rectangular timbers. Next, the cylinder 832 is driven to slightly widen the interval between the band saws 841, 841 arranged on the left and right sides, and the vehicle body 62 is returned to the initial position. Finally, the clamps 64 and 65 release the log L and remove the log L from above the vehicle body 62 . When the removal of this log L is completed, the next log L is clamped by a pair of clamping plates 70 from the rotating device 50 shown in FIG.

In the embodiments described above, the position of the spinal cord can be accurately estimated by the spinal cord position estimating device 10 . Moreover, according to the lumbering system 1, by providing the rotating device 50, lumbering can be performed according to the position of the pith that is accurately estimated.

The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope of the claims. For example, the learning model storage unit 12 also stores a learning model constructed by supervised learning based on tree ring data relating to the density or average width of tree rings up to the n-th tree ring centering on the position of the pith, and the estimation unit 14 , the density of annual rings or the average width of annual rings may be estimated from the target image data acquired by the acquisition unit 13 . The n-th annual ring may be an annual ring (for example, the 15th annual ring) that serves as a boundary between mature and immature wood. Regarding the density of tree rings, it may be possible to estimate which group it belongs to among multiple groups (high density, normal, low density) for tree ring density. A plurality of groups (average ring width is wide (eg, 6 mm or more), normal (eg, 4 mm or more and less than 6 mm), or narrow (eg, less than 4 mm)) may be estimated. An algorithm such as YOLO may also be used for estimation in .These estimations for tree rings can determine the position of the tree ring (for example, the 15th tree ring) that is the boundary between mature and immature wood.

Also, the photographed image photographed by the photographing device 40 is binarized and edge-enhanced, the annual rings are detected to some extent by image analysis, and deep learning can be used to improve detection accuracy and correct errors.

Furthermore, the pith position estimation device 10 of the present invention can be applied not only to lumber systems but also to plywood machines (rotary races) and the like.

It should be noted that even if the configuration requirements are included only in the descriptions of the embodiments and modifications described above, the configuration requirements may be applied to other embodiments and other modification examples.

1 lumber system 10 pulp position estimation device 11 learning unit 12 learning model storage unit 13 acquisition unit 14 estimation unit 20 input device 30 transport device 40 imaging device 50 rotation device 51, 52 rotation center 60 material feeding device 70 clamping plate 80 band saw C1 , C2, C3, C4, C5, C6 End face L Log p Marrow

Claims (2)

a learning model storage unit for storing a learning model constructed by supervised learning based on area image data representing a target area including the pith on the butt end surface and position data specifying the position of the pith in the area image data; ,
an acquisition unit that acquires target image data representing all or part of the butt surface;
an estimating unit that uses the learning model to estimate the position of the pith on the butt surface of the target image data from the target image data acquired by the acquiring unit ;
The position data is data representing the size of an area that includes the pith and is narrower than the target region and the position of the area with respect to the target region,
A predetermined positional relationship is established between the area and the pith contained in the area,
The estimating unit performs object detection based on the area from the target image data acquired by the acquiring unit, and estimates the position of the pith based on the predetermined positional relationship,
A pith position estimation device , wherein the predetermined positional relationship is a positional relationship in which the position of the pith coincides with the center position of the area . a holding device that holds the core wood in which the pith remains in a non-rotatable manner in the circumferential direction;
a sawing machine for cutting the core wood held by the holding device;
A learning model storage unit for storing a learning model constructed by supervised learning based on area image data representing a target area including the pith on the butt end surface and position data specifying the position of the pith in the area image data, estimating the position of the pith on the butt surface of the target image data from the target image data obtained by the obtaining unit using the acquisition unit that acquires target image data representing all or part of the butt surface and the learning model; a spinal cord position estimation device having an estimation unit ;
a photographing device for photographing a butt surface of the core wood before being held by the holding device and transmitting data of the photographed image to the acquisition unit;
a rotating device that rotates the core wood in the circumferential direction around the position of the pith estimated by the estimating unit and adjusts the sawing posture of the core wood on the sawing machine;
The acquisition unit acquires image data transmitted from the photographing device as the target image data, and the holding device holds the core lumber whose lumber attitude has been adjusted by the rotation device. A lumber system characterized by:

Images