JP2021032736A - Wear amount measurement system - Google Patents

Abstract

To provide a wear amount measurement system that can more accurately measure the amount of wear of a part of a component of a working machine from an image including the component.SOLUTION: A wear amount measurement system 1 comprises calculation devices 2A, 3A that calculate the amount of wear of a component of a working machine 5 from a captured image including the component. The calculation devices 2A, 3A comprise: a feature point extraction unit 22 in which extraction of feature points is machine-trained by using, as teacher data, a learning image g1 including a specific kind of component of the working machine 5 and a plurality of feature points a1-a10, b1-b10 that specify the shape of a worn part of the component in the learning image g1, and which extracts, from a captured image G1, feature points A1-A10, B1-B10 of a part in the captured image G1; and a wear amount calculation unit 37A that calculates the amount of wear of the part from the feature points A1-A10, B1-B10.SELECTED DRAWING: Figure 3

Description

The present invention relates to a wear amount measuring system that calculates the amount of wear of a part of a part from an image including a part of a work machine.

Conventionally, continuous use of a work machine may cause wear of parts that make up the work machine, and it is important to manage the wear state of these parts in order to use the work machine in a stable manner. ..

From this point of view, for example, Patent Document 1 obtains an image of a worn part of a work machine, measures the worn edge of the worn part from the acquired image, and wears from the result of the measured wear edge. A wear determination system that identifies the wear state of a part has been proposed.

Special Table 2016-504643

However, in the wear determination system shown in Patent Document 1, the wear state of the part to be worn is specified from the acquired image, but the difference in the result of the wear state even in the same wear state depending on the imaging angle of the wear part. May become large and it may not be possible to pinpoint the exact state of wear.

The present invention has been made in view of these points, and an object of the present invention is to measure the amount of wear that can more accurately measure the amount of wear of a part of a part from an image including a part of a work machine. Provide a system.

In view of the above problems, the wear amount measuring system according to the present invention is a wear amount measuring system including a calculation device for calculating the wear amount of the parts from an image captured including the parts of the work machine. , The learning image including a specific type of part of the work machine and a plurality of feature points for specifying the shape of the part where the part is worn in the learning image are used as training data to extract the feature point. It is provided with a feature point extraction unit that is machine-learned and extracts feature points of the portion in the captured image from the captured image, and a wear amount calculation unit that calculates the wear amount of the portion from the feature points.

According to the present invention, the feature point extraction unit includes a learning image including a specific type of component of a work machine, and a plurality of feature points that specify the shape of a portion where the component wears in the learning image. As teacher data, the extraction of feature points is machine-learned. Therefore, for example, when the captured image captures the shape of the part of the component at an imaging angle at which the amount of wear can be calculated, feature points are extracted, but the captured image is captured at such an imaging angle. If not, the feature points are not extracted. As a result, since the feature points are extracted from the captured image in which the amount of wear can be calculated accurately, the amount of wear of the part of the part can be measured more accurately from the image including the part of the work machine.

It is a schematic side view of the work machine which is the object of measurement of the wear amount measuring system which concerns on 1st to 3rd Embodiment of this invention. It is a schematic conceptual diagram of the wear amount measurement system which concerns on 1st Embodiment of this invention. It is a block diagram of the arithmetic unit of the wear amount measurement system shown in FIG. It is a block diagram for demonstrating the machine learning of the part detection part and the feature point extraction part shown in FIG. It is a schematic diagram which showed the image for learning by the part detection part shown in FIG. It is a schematic diagram which showed the learning image by the feature point extraction part shown in FIG. It is a schematic diagram which displayed the result by the part detection part and the feature point extraction part shown in FIG. 3 on the image of a sprocket. It is a figure for demonstrating the method of calculating the wear amount by the wear amount calculation part from the result shown in FIG. 5A. It is a graph for demonstrating the calculation of the component replacement by the replacement time calculation unit shown in FIG. It is a figure which showed the display screen of the communication terminal after setting in the reference gauge setting part shown in FIG. It is a work flow figure by the wear amount measuring system shown in FIG. It is a block diagram which concerns on the modification of the part detection part and the feature point extraction part shown in FIG. It is a block diagram of the arithmetic unit of the wear amount measurement system which concerns on 2nd Embodiment of this invention. It is a schematic diagram which set the shooting guide by the guide setting part shown in FIG. 11 in an image. It is a schematic diagram for demonstrating the correction by the feature point correction part shown in FIG. It is a work flow figure by the wear amount measuring system shown in FIG. It is a block diagram which concerns on the modification shown in FIG. It is a block diagram of the arithmetic unit of the wear amount measurement system which concerns on 3rd Embodiment of this invention. It is a schematic diagram which set the shooting guide in the captured image of the tooth of the bucket of the work machine shown in FIG. It is a schematic diagram which set the region and the feature point in the learning image of the tooth shown in FIG. 17A. It is a schematic diagram which displayed the area and the feature point in the captured image of tooth. It is a schematic diagram in which a shooting guide is set in the captured image of the roller of the work machine shown in FIG. It is a schematic diagram which set the region and the feature point in the learning image of the roller shown in FIG. 18A. It is a schematic diagram which displayed the region and the feature point in the image taken by the roller. It is a schematic diagram in which a shooting guide is set in the captured image of the shoe plate of the work machine shown in FIG. It is a schematic diagram which set the region and the feature point in the learning image of the shoe plate shown in FIG. 19A. It is a schematic diagram which displayed the region and the feature point in the image taken by the shoe plate. It is a schematic diagram in which a shooting guide is set in the captured image of the link of the work machine shown in FIG. It is a schematic diagram which set the region and the feature point in the learning image of the link shown in FIG. 20A. It is a schematic diagram which displayed the area and the feature point in the captured image of a link.

Hereinafter, some embodiments of the wear amount measuring system according to the present invention will be described with reference to the drawings.

1. 1. About the work machine 5 In the present embodiment, the work machine 5 to be measured for wear is not particularly limited, but is, for example, a hydraulic excavator or the like. Hereinafter, an example of the configuration of the work machine 5 to be measured by the wear amount measurement system 1 of the present embodiment will be described first, and then the wear amount measurement system 1 of the present embodiment will be described in detail.

As shown in FIG. 1, the work machine 5 has a crawler type traveling body 51, a swivel body 52 rotatably provided on the traveling body 51, and an articulated type mounted on the swivel body 52 as a hydraulic load. It is equipped with a front work machine 53. A cabin 54 is provided on one of the left and right sides of the front portion of the swivel body 52, and a front working machine 53 that rotates in the vertical direction to perform work such as excavation is provided in the center of the front portion of the swivel body 52. A counterweight 56 for maintaining the balance of the weight of the machine body is provided at the rear part of the swivel body 52.

The traveling body 51 is provided with an idler 51A1 at the front end and a sprocket 51A2 driven by a traveling motor (not shown) at the rear end. A link (link assembly) 51B is bridged between the idler 51A1 and the sprocket 51A2, and a roller 51C for supporting the link 51B is provided between the idler 51A1 and the sprocket 51A2. A shoe plate 51D to be installed on the ground is attached to the link 51B. By configuring the traveling body 51 in this way, the work machine 5 can move on rough terrain.

The front work machine 53 is, for example, a machine provided on the front side of the swivel body 52 and driven by a hydraulic drive device to perform work such as excavation work. In the front working machine 53, for example, a boom 53A, an arm 53B, and a bucket 53C are connected to each other, and these are operated by hydraulic cylinders 53a, 53b, and 53c. A plurality of teeth (claw portions) 53D are fixed to the tip of the bucket 53C at intervals in the width direction.

Here, when the work machine 5 is driven by the traveling body 51, the teeth (parts) of the sprocket (part) 51A2, the support portion (part) supporting the link 51B of the roller (part) 51C, and the link (part) 51B The connecting portion (part) and the convex portion (part) on which the shoe plate (part) 51D touches the ground are worn. Further, the tips of the plurality of teeth (parts) 53D are worn due to work such as excavation work. Therefore, in the following first and second embodiments, the wear amount measuring system 1 measures the wear amount of the teeth 51a of the sprocket 51A2, and in the third embodiment, the wear amount of the parts of these parts is measured.

[First Embodiment]
2. Configuration of Wear Amount Measurement System 1 As shown in FIG. 2, the wear amount measurement system 1 according to the first embodiment is a part of a component constituting a work machine 5 imaged by a plurality of communication terminals 3, 3, ... It is a device that calculates the amount of wear of the portion by the arithmetic device 3A of each communication terminal 3 and the arithmetic device 2A of the server 2 via the network. In the present embodiment, the wear amount measuring system 1 includes the server 2. For example, a series of applications for calculating the wear amount of the part from the image of the part described above can be installed in the storage unit 3b of the arithmetic unit 3A. For example, it may be composed of only each communication terminal 3.

The server 2 includes at least an arithmetic unit 2A, and the arithmetic unit 2A includes an arithmetic unit 2a composed of a CPU and the like, and a storage unit 2b composed of a ROM, a RAM, and the like. In the present embodiment, the storage unit 2b stores teacher data such as a learning image described later, a program for executing the site detection unit 21 and the feature point extraction unit 22, and the calculation unit 2a stores these. A device that executes a program. In this specification, the site detection unit 21 and the feature point extraction unit 22 shown in FIGS. 3 and 3 and later are shown as a portion in which the program stored in the storage unit 2b is executed by the calculation unit 2a. The server 2 is accessible by, for example, each pre-registered communication terminal 3, and receives calculation results such as an image captured by each communication terminal 3 and the amount of wear of a part, and these results are stored in a storage unit. It may be stored in 2b.

The communication terminal 3 is a general mobile terminal or tablet terminal, and includes an arithmetic unit 3A, a touch panel display 3D, and an image pickup device 3E. The arithmetic unit 3A includes an arithmetic unit 3a composed of a CPU and the like, and a storage unit 3b composed of a ROM, RAM and the like. The storage unit 3b stores a program for executing the image acquisition unit 33, the extraction determination unit 34, the wear amount calculation unit 37A, etc., which will be described later, and the calculation unit 3a is a device that executes this program. In this specification, the image acquisition unit 33, the extraction determination unit 34, the wear amount calculation unit 37A, and the like shown in FIGS. 3 and 3 are shown as parts for executing the program stored in the storage unit 3b in the calculation unit 3a. ..

The touch panel display 3D is composed of a display device 3B and an input device 3C. The display device 3B is a display screen, and displays images and images captured by the image pickup device 3E, calculation results calculated by the calculation device 3A, and the like. The input device 3C is a device in which an operator inputs predetermined data or the like to the arithmetic unit 2A. The input device 3C is a feature of a part of a component described later, which is displayed on an input button (icon) and a display device (display screen) 3B in which an operator inputs a command signal for acquiring an image to the image pickup device 3E. It has a marker (icon) that corrects the position of the point.

The image pickup device 3E is a device for capturing a digital image and a digital image of a digital camera or the like, and in the present embodiment, the image pickup device 3E is a device for displaying an image (moving image) at a predetermined timing according to the activation of an application for executing wear measurement. Display on 3B. However, this moving image is not recorded in the storage unit 3b. Based on the input signal input by the operator from the input device 3C, the image acquisition unit 33 uses the image (moving image) displayed on the display device 3B, and the image pickup device 3E acquires the captured image (still image). This captured image is recorded in the storage unit 3b.

3. 3. About the arithmetic unit 2A of the server 2 In the present embodiment, the arithmetic unit 2A of the server 2 includes an identification device 20 composed of a site detection unit 21 and a feature point extraction unit 22 as software. The identification device 20 identifies the teeth 51a of the sprocket 51A2 from the above-mentioned image of the sprocket, is generated by the learning unit 40 installed in the storage unit 2b of the server 2, and is executed by the calculation unit 2a. However, the identification device 20 may be provided in the arithmetic unit 3A of the communication terminal 3. Since FIG. 3 is a block diagram for explaining the utilization by the part detection unit 21 and the feature point extraction unit 22, which are trained models, the learning unit 40 used in the learning phase is omitted in FIG. ing.

In the present embodiment, the site detection unit 21 detects the position of the tooth, which is the site of the component, from the image of the sprocket 51A2 as a component of the work machine 5. The site detection unit 21 includes a learning image including the sprocket 51A2 as a specific type of component of the work machine 5, and a region of teeth 51a (specifically, a rectangular region) which is a wearable portion of the sprocket 51A2 in the learning image. ) Is used as teacher data, and the detection of the region of the tooth 51a is machine-learned from this teacher data. Therefore, the site detection unit 21 detects the region R of the teeth 51a of the sprocket 51A2 in the captured image from the captured image (specifically, the image acquired by the image acquisition unit 33) captured by the imaging device 3E.

As shown in FIG. 4, the site detection unit 21 includes a local feature amount extraction unit 21a, an object classifier 21b, and a region identification unit 21c. The local feature amount extraction unit 21a extracts the local feature amount of the image while selecting a region to be a rectangular plurality of pixels included in the captured image by shifting by a predetermined number of pixels. In the present embodiment, HOG (Histograms of Oriented Gradients) features (feature vectors) for each region are extracted. The HOG feature amount is a histogram of the gradient of the brightness of the image.

The object detector 21b is machine-learned from the features of the image G1 of the teeth (parts) 51a of the sprocket (parts) 51A2 of the work machine 5. Specifically, the object classifier 21b is a discriminant function machine-learned by an SVM (Support Vector Machine), and is learned by the learning unit 40.

In the present embodiment, the learning unit 40 is provided in the arithmetic unit 2A of the server 2. However, the learning unit 40 may be provided in another system, for example, and the site detection unit 21 and the feature point extraction unit 22 generated by the other system may be installed in the arithmetic unit 2A of the server 2. As shown in FIG. 5A, the learning unit 40 includes a learning image storage unit 41 in which a large number of learning images g1 in which the teeth (parts) 51a of the sprocket (part) 51A2 are captured, and each of the learning image g1. It is provided with a site region storage unit 42 that is associated with and stores regions r1, r2, ... Of the teeth 51a. The site area storage unit 42 stores two rectangular areas r1 and r2 in the learning image g1, and specifically, stores the coordinates of the four corners of the areas r1 and r2.

The learning unit 40 further includes a feature point storage unit 43, a first learning unit 44, and a second learning unit 45. The first learning unit 44 is a learning device having an SVM, and generates an object classifier 21b. Here, the SVM is basically a two-class discriminator. In the learning phase, the HOG features extracted from each of a large number of learning images (normal examples) that identify the area where the teeth (parts) 51a of the sprocket (part) 51A2 are shown, and the teeth of the sprocket 51A2 are not shown. From a large number of learning images (negative examples) in a region (or difficult to recognize as sprocket teeth), the discrimination boundary with the HOG feature extracted from each is machine-learned by SVM. In the present embodiment, the images extracted by the local feature amount extraction unit 21a are sequentially input to the first learning unit 44 with respect to the learning image g1 stored in the learning image storage unit 41. In machine learning, the images of the areas r1 and r2 stored in the site area storage unit 42 are used as a positive example, and the images of the other areas are used as a negative example. Will be generated.

At the time of utilization, the HOG feature amount extracted for each pixel group is input to the captured image G1 by the local feature amount extraction unit 21a into the object classifier 21b which is a trained model. As a result, as shown in FIG. 6A, the teeth 51a of the sprocket 51A2 can be identified in the captured image G1 by the rectangular regions R1 and R2. The region identification unit 21c identifies the coordinates of the four corners of each region R1 and R2 in the captured image G1 based on the identification result of the object classifier 21b (see, for example, FIG. 6A). This result is sent to the extraction determination unit 34, which will be described later.

As shown in FIG. 5B, the feature point extraction unit 22 has a shape of the above-mentioned learning image g1 including the sprocket (part) 51A2 of the work machine 5 and a tooth (part) 51a in which the sprocket 51A2 is worn in the learning image g1. The extraction of the feature points is machine-learned using a plurality of feature points a1 to a10 and b1 to b10 for specifying the above as teacher data. The feature point extraction unit 22 extracts feature points A1 to A10 and B1 to B10 of the teeth (sites) 51a of the sprocket 51A2 in the captured image G1 from the captured image G1 (see, for example, FIG. 6A).

Specifically, the feature point extraction unit 22 includes a feature point classifier 22a and a feature point identification unit 22b. In the present embodiment, the feature point classifier 22a is a cascade type classifier, and is generated by the second learning unit 45 of the learning unit 40. In the present embodiment, the second learning unit 45 has a random forest (Random Forest) algorithm. In machine learning, a large number of learning images g1 stored in the learning image storage unit 41 are associated with an image whose feature amount is extracted by the local feature amount extraction unit 21a and each learning image g1. The coordinates of the feature points a1 to a10 and b1 to b10 stored in the point storage unit 43 and the data of the areas r1 and r2 stored in the site area storage unit 42 are input.

Specifically, as teaching data in the learning phase, in order to specify the shape of the tooth 51a of the sprocket 51A2, the feature points a1 to a7 and the feature points b1 to b7, which are the features of the shape from the tooth ridge to the tooth ridge, Feature points a8 to a10 and feature points b8 to b10 near the tooth base on the circumference 51b of the sprocket 51A2 are stored for each learning image g1.

The circumference 51b is a circumference formed at a constant distance from the center of rotation of the sprocket 51A2. The feature points a8 and a10 (b8 and b10) are the intersections of the perpendicular line drawn from a1 and a7 (b1 and b7) located at the apex of the mountain of the tooth 51a to the circumference 51b and the circumference 51b. On the other hand, the feature point a9 (b9) is an intersection of the perpendicular line drawn from a4 (b4) located at the valley point of the tooth 51a to the circumference 51b and the circumference 51b.

By inputting such teacher data, in the present embodiment, the second learning unit 45 machine-learns the extraction of the feature points A1 to A10 and the feature points B1 to B10 by the random forest, and generates the feature point classifier 22a. To do.

At the time of utilization, the HOG feature amount extracted for each pixel group is input to the captured image G1 by the local feature amount extraction unit 21a into the feature point classifier 22a which is a trained model. As a result, as shown in FIG. 6A, the feature points A1 to A10 and B1 to B10 of the shape of the teeth 51a of the sprocket 51A2 are extracted in the rectangular regions R1 and R2 and in the vicinity thereof in the captured image G1. The feature point specifying unit 22b specifies the coordinates of the feature points extracted in the captured image G1. This result is sent to the area / feature point output unit 35.

In this embodiment, SVM is used for the first learning unit 44 and a random forest is used for the second learning unit 45 as the machine learning algorithm. However, as the machine learning algorithm, commonly known algorithms such as Decision Tree, Booting, or Real AdaBoost, and neural networks may be used. Further, in the present embodiment, the HOG feature amount is illustrated for machine learning, but other feature amounts such as LBP (Local Binary Pattern) feature amount and Har-like feature amount may be used, or a plurality of feature amounts may be used. It may be used in combination.

4. About the arithmetic unit 3A of the communication terminal 3 In the present embodiment, the arithmetic unit 3A of the communication terminal 3 includes an image acquisition unit 33, an extraction determination unit 34, an area / feature point output unit 35, and a wear amount calculation as software. A unit 37A, a replacement time calculation unit 37B, and a reference gauge setting unit 37C are provided (see FIG. 3). These are installed in the storage unit 3b of the arithmetic unit 3A of the communication terminal 3 as one application for measuring the amount of wear, and are executed by the arithmetic unit 3a.

The image acquisition unit 33 acquires the captured image G1 from the image pickup device 3E. In the present embodiment, the image acquisition unit 33 acquires the captured image G1 by the image pickup device 3E at the timing when the operator presses the input button arranged on the input device 3C. Specifically, a real-time image (moving image) captured by the image pickup device 3E is displayed on the display device 3B, and the operator presses an input button (not shown) while checking the image. The image pickup device 3E captures an object, and the image acquisition unit 33 acquires an image that is a still image.

The extraction determination unit 34 determines the execution of extraction of feature points by the feature point extraction unit 22 based on the detection result of the site detection unit 21. Specifically, the extraction determination unit 34 permits the feature point extraction unit 22 to extract the feature points from the captured image G1 when the site detection unit 21 can detect the teeth of the sprocket 51A2.

On the other hand, when the site detection unit 21 cannot identify the tooth of the sprocket 51A2, the shape of the tooth 51a whose wear amount can be evaluated cannot be detected from the captured image G1. In this case, the extraction determination unit 34 does not allow the feature point extraction unit 22 to execute the extraction of the feature points with respect to the captured image G1, and causes the display device 3B to display the image captured by the image pickup device 3E. This prompts the operator to image the teeth 51a of the sprocket 51A2 again using the image pickup device 3E.

In the present embodiment, the feature point extraction unit 22 extracts the feature points A1 to A10 and B1 to B10 of the tooth valley regions R1 and R2 of two adjacent sprockets, but for example, only one region R1 is extracted. If the site detection unit 21 cannot be detected, the position of the region R2 may be estimated from the inner circumference 51b or the like of the sprocket 51A2. In this case, the feature points B1 to B10 are extracted by the feature point extraction unit 22 using the image of the estimated region R2.

The area / feature point output unit 35 displays the results of the site detection unit 21 and the feature point extraction unit 22 on the display device 3B. In the present embodiment, as shown in FIG. 6A, the region / feature point output unit 35 includes the captured image G1 and the regions R1 and R2 of the teeth 51a detected by the site detection unit 21 on the captured image G1 and the feature points. The feature points A1 to A10 and B1 to B10 extracted by the extraction unit 22 are displayed.

The wear amount calculation unit 37A calculates the wear amount of the teeth 51a of the sprocket 51A2 from the feature points A1, A2, A8, and B6. Specifically, in the present embodiment, as shown in FIG. 6B, the distance between the feature point A1 and the feature point A8 is the length X1 close to the tooth height, and the distance between the feature point A2 and the feature point B6 is the tooth width. Corresponds to X2. Since the length X1 is almost unchanged even if the teeth 51a are worn, the wear state of the teeth 51a can be grasped by the amount of decrease in the tooth width X2. In the present embodiment, the value of X2 / X1 can be calculated as the wear rate, and the wear rate of the tooth 51a can be calculated as the wear amount of the tooth 51a.

The replacement time calculation unit 37B calculates the replacement time of the sprocket (part) 51A2 from the amount of wear (wear rate) of the teeth 51a of the sprocket (part) 51A. In the present embodiment, as shown in FIG. 7, the wear amount (wear rate), the usage time, and the data of the sprocket 51A2 are measured in advance, and the point at which the usage time of the sprocket 51A2 and the wear rate are plotted is obtained. To find the simple regression lines (these linear functions). By substituting the calculated wear rate into the obtained linear function, it is possible to calculate the usage time of the sprocket 51A2 to be evaluated so far. Then, the remaining usage time (recommended replacement time), that is, the replacement time of the sprocket 51A2 is calculated by subtracting the current usage time of the sprocket 51A2 from the required replacement time of the sprocket 51A2 (life time of the sprocket 51A2). Can be done. The calculation result of the replacement time calculation unit 37B is displayed on the display device 3B as a result as shown in FIG. 8, for example.

The reference gauge setting unit 37C sets an image of the reference gauge (virtual gauge) 51p according to the surface shape of the teeth 51a before wear on the teeth 51a of the sprocket 51A2 of the captured image G1 displayed on the display device 3B. .. Here, the reference gauge 51p aligns the position of the reference gauge 51p from the line segment between the feature point A1 and the feature point A8 and the line segment between the feature point A2 and the feature point B6 specified by the wear amount calculation unit 37A. In this way, how much the sprocket 51A2 is worn can be easily confirmed via the display device 3B.

5. About the wear amount measuring method The wear amount measuring method using the above-mentioned wear amount measuring system 1 will be described below with reference to FIG. First, in step S91, the above-mentioned application for measuring the amount of wear installed in the communication terminal 3 is started. Next, the process proceeds to step S92, and after the application is started, the teeth 51a and the imaging range of the imaging device 3E are aligned so that the teeth 51a of the sprocket 51A2 are imaged. Next, in step S93, the image pickup device 3E images the sprocket 51A2, which is a part of the work machine 5, by the operation from the input device 3C by the operator.

When the sprocket 51A2 is imaged, the process proceeds to step S94, and the site detection unit 21 detects the region including the teeth 51a of the sprocket 51A2 from the captured image G1. In step S95, if the extraction determination unit 34 determines that the tooth 51a of the sprocket 51A2 has not been detected by the site detection unit 21, the process proceeds to step S92 in order to perform imaging again.

On the other hand, if the extraction determination unit 34 determines in step S95 that the tooth 51a of the sprocket 51A2 has been detected by the site detection unit 21, the process proceeds to step S96, and the process proceeds to step S96 to proceed to the feature points A1 to A10 and B1 to B10 of the teeth 51a. Is extracted (see FIG. 6A). In step S97, the wear amount calculation unit 37A calculates the wear amount (wear rate) of the tooth 51a from the feature points A1, A2, A8, and B6 (see FIG. 6B).

Next, in step S98, the replacement time calculation unit 37B calculates the replacement time of the sprocket (part) 51A2 from the amount of wear (wear rate) of the teeth 51a of the sprocket (part) 51A2 (see FIG. 7). Further, in step S99, the reference gauge setting unit 37C is a reference gauge (virtual gauge) corresponding to the surface shape of the teeth 51a before wear with respect to the teeth 51a of the sprocket 51A2 of the captured image G1 displayed by the display device 3B. An image of 51p is set, and an image showing the difference between the reference gauge 51p and the actual wear and a graph as a guideline for the recommended replacement time are displayed (see FIG. 8).

In the first embodiment shown above, the site detection unit 21 is provided. For example, the site detection unit 21 is omitted, and the feature point extraction unit 22 with respect to the captured image G1 features the teeth 51a of the sprocket 51A2. A point may be extracted and the amount of wear of the tooth 51a may be measured from this feature point. In this case, as shown in FIG. 10, the second learning unit 45 uses the learning image g1 of the learning image storage unit 41 and the learning feature points a1 to a10 and b1 to b10 of the feature point storage unit 43. The feature point classifier 22a is generated as the teacher data. In the local feature amount extraction unit 21a described above, the captured image G1 is input from the image acquisition unit 33, and the HOG feature amount of the captured image G1 is calculated. The feature point storage unit 43 extracts the feature points A1 to A10 and B1 to B10 of the teeth 51a from the HOG feature amount of the captured image G1, and the feature point identification unit 22b specifies these coordinates. Therefore, in such a case, in addition to the site detection unit 21 shown in FIG. 3, the extraction determination unit 34 shown in FIG. 3 can be omitted, and the region / feature point output unit 35 can use the regions R1 and R2. Do not output. Further, the site area storage unit 42 and the first learning unit 44 shown in FIG. 3 can also be omitted.

As described above, in the present embodiment, the feature point extraction unit 22 includes the learning image g1 including the sprocket 51A2 as a specific type of part of the work machine 5, and the tooth 51a which is a portion where the sprocket 51A2 is worn in the learning image g1. The extraction of the feature points is machine-learned using a plurality of feature points a1 to a10 and b1 to b10 that specify the shape of the feature points as teacher data. Therefore, for example, when the captured image G1 captures the shape of the part of the component at an imaging angle at which the amount of wear can be calculated, the feature points are extracted due to the learning by the teacher data. However, if the image is not taken at such an imaging angle, the feature points are not extracted. As a result, since the feature points are extracted only for the captured image in which the wear amount can be calculated accurately, the wear amount of the part part should be measured more accurately from the image G1 including the sprocket 51A2 of the work machine 5. Can be done. Furthermore, the state of wear can be grasped more accurately by the feature points as compared with the case where the edge detection of simple image processing is used.

Further, as shown in FIG. 8, the display device 3B displays an image diagram in which a virtual gauge is fitted to the image captured by the image pickup device 3E and an image diagram of a graph showing the correlation between the amount of wear and the recommended replacement time. Using this display as a guide, a safety inspector at the site can show the customer a suggestion such as "Why don't you replace it soon?"

[Second Embodiment]
FIG. 11 is a wear amount measuring system 1 according to the second embodiment. As shown in FIG. 11, the wear amount measuring system 1 according to the present embodiment differs from that of the first embodiment in that it includes a guide setting unit 32 and a feature point correction unit 36, and a second. The point is that the learning unit 45 executes re-learning. Therefore, the same configuration as that of the first embodiment omits a detailed description thereof. In the second embodiment, the guide setting unit 32, the feature point correction unit 36, and the re-learning function are provided, but in the wear amount measurement system 1 of the second embodiment, at least one of these is provided. May be provided.

In the present embodiment, the arithmetic unit 2A selects a photographing guide (reference line) T1 according to the shape of the component to be photographed for the image (moving image) ga1 of the imaging device 3E displayed on the display device 3B. A guide setting unit 32 for setting is further provided. Specifically, as shown in FIG. 12, the photographing guide T1 is displayed to specify the position where the feature points of the sprocket 51A2 are extracted, and in the present embodiment, the teeth 51a of the sprocket 51A2 (specifically). The tooth tip) and the position along the circumference 51b are set.

By providing the photographing guide T1 in this way, the operator only has to image the teeth 51a of the sprocket 51A2 along the photographing guide T1, so that the feature point extraction unit 22 more accurately extracts the feature points. It can be carried out. In the present embodiment, the inclination of the communication terminal 3 with respect to the ground (horizontal plane) is directly reflected in the image displayed by the display device 3B. However, the communication terminal 3 may be provided with a gyro sensor, and the display device 3B may display the photographing guide T1 in a fixed posture on the image ga1 regardless of the inclination of the communication terminal 3 with respect to the ground. As a result, the photographing guide T1 does not change regardless of the posture of the communication terminal 3 with respect to the ground, so that the teeth 51a of the sprocket 51A2 can be imaged more stably. Further, in the present embodiment, the image is taken by the operator, but for example, the image pickup device 3E may automatically take an image after a predetermined time has elapsed from the start of the application.

In the present embodiment, the arithmetic unit 3A further includes a feature point correction unit 36 that corrects the position of the feature point according to the change in the position of the feature point extracted by the feature point extraction unit 22. In the present embodiment, as shown in FIG. 13, the display device 3B displays the feature points extracted by the feature point extraction unit 22 on the captured image G1 together with the captured image G1. FIG. 13 shows feature points A1, A2, A8, and B6 that contribute to the calculation of the amount of wear. In the input device 3C, the operator inputs a change in the positions of the feature points A1, A2, A8, and B6 on the captured image G1 with respect to the feature points A1, A2, A8, and B6 displayed on the display device 3B. Can be done. Specifically, the feature point correction unit 36 enables the operator to tap (select) a feature point whose position should be changed among the positions of the feature points displayed by the display device 3B. There is. The feature point correction unit 36 enlarges and displays the portion including the feature point tapped by the operator via the input device 3C by tapping. The feature point correction unit 36 further moves the feature point from the current position to the position to be corrected while the operator touches the tapped feature point (enlarged display feature point), and the operator moves the feature point from the moved feature point. When you release your finger, the feature points are changed there.

Here, since the positions of the feature points A1 and B6 are not appropriate positions, the operator changes the positions of the feature points A1 and B6 via the input device 3C as shown in FIG. 13, and as a result, the features The point correction unit 36 corrects the coordinates of the feature points A1 and B6. In this way, with the correction of the feature points A1 and B6, the wear amount calculation unit 37A more accurately reduces the wear amount of the teeth 51a of the sprocket 51A2 based on the feature points A1 and B6 corrected by the feature point correction unit 36. Can be calculated.

In the present embodiment, the feature point correction unit 36 makes corrections by the feature points A1, A2, A8, and B6 that contribute to the calculation of the amount of wear. For example, the feature point correction unit 36 makes corrections in the feature point extraction unit 22. The positions of all the extracted feature points may be corrected.

Further, in the present embodiment, the arithmetic unit 2A uses the captured image G1 and the feature points A1 to A10 and B1 to B10 including the modified feature points A1 and B6 as further teacher data in the second learning unit 45. The point extraction unit 22 may be relearned. As a result, the accuracy of extracting the feature points by the feature point extraction unit 22 can be further improved.

As shown in FIG. 14, in the wear amount measuring system 1 according to the second embodiment, step S92'is different from that of the first embodiment, and further, steps S141 to S143 are performed between steps S96 and S97. The added point is different from that of the first embodiment. Specifically, in step S92', the operator images the teeth 51a of the sprocket 51A2 in step S93 with the photographing guide T2 aligned with the portions of the sprocket 51A2 (teeth 51a and circumference 51b). As a result, the teeth 51a of the sprocket 51A2 to be measured for the amount of wear are arranged at a specific position of the captured image G1. Therefore, the site detection unit 21 detects the region of the teeth 51a and the feature point extraction unit 22 detects the teeth. The feature points including 51a can be extracted more accurately.

Further, in step S141, it is determined whether the position of the feature point is correct, and if it is correct, the process proceeds to step S97, and the wear amount calculation unit 37A continues from the feature points A1, A2, A8, and B6 to the teeth. The amount of wear (wear rate) of 51a is calculated (see FIG. 6B). On the other hand, if the position of the feature point is incorrect, the operator corrects the position of the feature point and proceeds to step S97 (see FIG. 13). Specifically, the operator taps the feature point whose position is to be changed among (candidates for) the feature point, and the vicinity of the tapped feature point is enlarged and displayed on the display device 3B. When the operator moves the feature point from the enlarged current position to the position to be corrected and the operator releases the finger from the feature point, the feature point is changed there. The wear amount calculation unit 37A calculates the wear amount (wear rate) of the tooth 51a based on the feature points corrected by the feature point correction unit 36. In line with this, in step S143, the second learning unit 45 relearns the feature point extraction unit 22. Through such a series of operations, the amount of wear can be calculated more accurately as the position of the feature point is corrected by the operator, and the replacement time of the sprocket 51A2 can be accurately determined.

Here, in the second embodiment shown above, the captured image G1 is acquired by the input operation (pressing the input button) of the operator. However, the captured image G1 may be acquired by the wear amount measuring system 1 according to the modified example shown in FIG. This modification is mainly different from the second embodiment shown above in that the image pickup permission unit 33A is further provided.

Specifically, as shown in FIG. 15, in the wear amount measuring system 1, the display device 3B is configured to display an image (moving image) from the image pickup device 3E before acquiring the captured image G1. The video (moving image) is also input to the site detection unit 21. The site detection unit 21 is configured to acquire an image from the image and detect the regions R1 and R2 of the teeth 51a of the sprocket 51A2.

The arithmetic unit 3A includes an image pickup permission unit 33A that allows the image pickup device 3E to take an image at the timing when the site detection unit 21 detects the regions R1 and R2 of the teeth 51a of the sprocket 51A2 in the image from the video.

The image acquisition unit 33 causes the image pickup device 3E to capture the image G1 at the timing when the site detection unit 21 detects the regions R1 and R2 of the teeth 51a of the sprocket 51A2, that is, at the timing when the image pickup permission unit 33A permits the image pickup. Outputs a command signal for acquisition. As a result, the image pickup device 3E automatically acquires the captured image G1 regardless of the operation of the operator. Since the acquired image G1 is an image in which the site detection unit 21 detects the regions R1 and R2 of the teeth 51a of the sprocket 51A2, the feature point extraction unit 22 can extract the feature points at accurate positions. ..

[Third Embodiment]
FIG. 16 is a wear amount measuring system 1 according to the third embodiment. As shown in FIG. 16, the difference between the wear amount measuring system 1 according to the present embodiment and that of the second embodiment is that the wear amount is measured for a plurality of types of parts, and the parts are set. A unit 31A, an identification device selection unit 31B, and five identification devices 20A to 20E corresponding to a plurality of types of parts are provided. Therefore, the same configuration as that of the second embodiment omits a detailed description thereof.

In the present embodiment, the arithmetic unit 2A of the server 2 includes identification devices 20A to 20E according to a plurality of types (five types) of parts of the work machine 5. When the site detection unit 21 is omitted as in the modified example of the first embodiment, the arithmetic unit 2A includes a plurality (five) feature point extraction units 22A to 22E.

In the present embodiment, the site detection unit 21A of the identification device 20A detects the region of the tooth (site) 51a of the sprocket 51A2 as a component, and the feature point extraction unit 22A extracts the feature points of the sprocket 51A2 (FIG. See 6A). This is the same as that described in the first and second embodiments.

The site detection unit 21B of the identification device 20B detects the region of the tip (site) 53c of the tooth 53D as a component, and the feature point extraction unit 22B extracts the feature points of the tooth 53D (see FIG. 17C). The site detection unit 21C of the identification device 20C detects the region of the support portion (site) 51 g of the roller 51C as a component, and the feature point extraction unit 22C extracts the feature points of the roller 51C (see FIG. 18C).

The site detection unit 21D of the identification device 20D detects the region of the convex portion (site) 51h of the shoe plate (part) 51D as a component, and the feature point extraction unit 22D extracts the feature points of the shoe plate 51D ( See FIG. 19C). The site detection unit 21D of the identification device 20E detects the region of the convex portion (site) 51h of the link (component) 51B as a component, and the feature point extraction unit 22E extracts the feature points of the link 51B (FIG. 20C). reference).

In the present embodiment, the arithmetic unit 3A of the communication terminal 3 includes a component setting unit 31A and an identification device selection unit 31B. The component setting unit 31A sets one component from a plurality of types (five types) of components, and one component (specifically, a component) is set by input information input by an operator via an input device 3C. Code) is set.

The identification device selection unit (extraction unit selection unit) 31B selects one identification device from a plurality (five) identification devices 20A to 20E according to the parts set by the component setting unit 31A. In the present embodiment, when the operator selects the sprocket 51A2 as a component, the component setting unit 31A sets the sprocket 51A2 as a component, and the identification device selection unit 31B selects the identification device 20A. Using the identification device 20A selected in this way, the amount of wear of the sprocket 51A2 is measured through a series of steps according to the configuration described in the second embodiment. When the modified example of the first embodiment is applied, the identification devices 20A to 20E are composed of the feature point extraction units 22A to 22E, so that the identification device selection unit 31B is the feature point extraction units 22A to 22E. select.

Here, the guide setting unit 32 selects and sets one of the shooting guides T1 to T5 according to the selected component. Specifically, when the selected component is the sprocket 51A2, the photographing guide T1 described in the second embodiment is set.

When the selected component is the tooth 53D, the guide setting unit 32 displays the image ga2 of the photographing guide T2 including the contour lines on both sides of the central tooth 53D and the lines perpendicular to the contour lines as shown in FIG. 17A. Set to. When the selected component is the roller 51C, the guide setting unit 32 displays an image of the rotation axis of the roller 51C and the photographing guide T3 corresponding to the contour line up and down along the support portion as shown in FIG. 18A. Set to ga3.

Further, when the selected component is the shoe plate 51D, the guide setting unit 32 is orthogonal to a line along the surface of the shoe plate 51D adjacent to the link 51B as shown in FIG. 19A. The center line of the convex portion 51h of the shoe plate 51D is set as the photographing guide T4 in the image ga4. Further, when the selected component is the link 51B, as shown in FIG. 20A, the guide setting unit 32 includes a circle along the connecting hole in which the links 51B are connected to each other, and a member of the link 51B. The contour line is set in the image ga5 as a shooting guide T5.

Here, the generation of the trained model will be described below. In the case of the tooth 53D, as shown in FIG. 17B, the learning image g2 and the regions r31 to r35 of each tooth 53D are used as teacher data in the same manner as described above, as shown in FIG. 17c. The site detection unit 21B in which the detection of the regions R31 to R35 in the captured image G2 has been learned is generated by the first learning unit 44.

Further, the feature in which the extraction of the feature points C1 to C5 in the captured image G2 is learned by using the learning image g2 and the feature points c1 to c5 of each tooth 53D as teacher data by the same method as the above-described method. The point extraction unit 22B is generated by the second learning unit 45.

The feature points to be learned are the intersections of the tooth 53D and the bucket 53C as the feature points c1 and c5, the mounting position of the tooth 53D (the boundary point with the adapter) as the feature points c2 and c4, and the tip of the tooth 53D as the feature point c3. To do.

As shown in FIG. 17C, the region detection unit 21B and the feature point extraction unit 22B, which are trained models, detect the regions R31 to R35 of each tooth 53D for the captured image G2 and extract the feature points C1 to C5. can do. Here, the wear amount calculation unit 37A sets the length of the line segment of the feature point C2 and the feature point C4 to X3, and sets the length from the feature point C3 to the intersection of the line orthogonal to this line segment to X4, and sets X4. By calculating / X3, the amount of wear (wear rate) of the tooth 53D is calculated. Since X4 decreases due to wear and X3 hardly changes, the wear state of the tooth 53D can be grasped by calculating X4 / X3.

In the case of the roller 51C, as shown in FIG. 18B, the image g3 for learning and the region r4 of the roller 51C are used as teacher data in the same manner as described above, and the captured image is as shown in FIG. 18c. The first learning unit 44 generates the site detection unit 21C in which the detection of the region R4 in G3 has been learned.

Further, the feature in which the extraction of the feature points D1 to D8 in the captured image G3 is learned by using the learning image g3 and the feature points d1 to d8 of each roller 51C as teacher data in the same manner as the above-described method. The point extraction unit 22C is generated by the second learning unit 45.

The feature points to be learned are the intersections of the rotation axis of the roller 51C and the contour lines at both ends of the support portion 51g as feature points d1 and d2, and the points along the contour line of the support portion 51g of the roller 51C as feature points d3 to d8. And. When the roller 51C is new, the lines connecting the feature points d3 to d8 are substantially straight lines, and as the wear progresses, these lines become concave curves as shown in FIG. 18B.

As shown in FIG. 18C, the region detection unit 21C and the feature point extraction unit 22C, which are trained models, detect the region R4 of the support portion 51g of each roller 51C with respect to the captured image G3, and feature points D1 to D8. Can be extracted. Here, the wear amount calculation unit 37A sets the distance between the feature point D4 and the feature point D8 in the rotation axis direction as X5, and sets the distance between the feature point D4 and the feature point D8 in the direction orthogonal to the rotation axis as X6. By calculating X5, the amount of wear (wear rate) of the roller 51C is calculated. Since X6 decreases due to wear and X5 hardly changes, the state of wear of the roller 51C can be grasped by calculating X6 / X5.

In the case of the shoe plate 51D, as shown in FIG. 19B, the learning image g4 and the region r5 of each shoe plate 51D are used as teacher data in the same manner as described above, as shown in FIG. 19c. The site detection unit 21D in which the detection of the region R5 in the captured image G4 has been learned is generated by the first learning unit 44.

Further, the extraction of the feature points E1 to E19 in the captured image G4 was learned by using the learning image g4 and the feature points e1 to e19 of each shoe plate 51D as teacher data in the same manner as the above-described method. The feature point extraction unit 22D is generated by the second learning unit 45.

The feature points to be learned are the points on the straight line along the back surface of the shoe plate 51D adjacent to the link 51B as feature points e1 to e4, and the points along the convex portion 51h of the shoe plate 51D and the valley portion between them. The feature points are e5 to e19.

As shown in FIG. 19C, the region detection unit 21D and the feature point extraction unit 22D, which are trained models, detect the region R5 of the convex portion 51h of the shoe plate 51D with respect to the captured image G4, and feature points E1 to E19. Can be extracted. Here, the wear amount calculation unit 37A sets the distance between the feature points E6 and E18 located at the vertices of the convex portions 51h on both sides to X7, and sets the distance from the feature points E12 to the line segment connecting the feature points E6 and E18 and the feature point E2. The wear amount (wear rate) of the shoe plate 51D is calculated by calculating X8 / X7 with the distance from and X8 as X8. Since X8 decreases due to wear and X7 hardly changes, the state of wear of the shoe plate 51D can be grasped by calculating X8 / X7.

In the case of the link 51B, as shown in FIG. 20B, the learning image g5 and the region r6 of each link 51B are used as teacher data and imaged as shown in FIG. 20c by the same method as described above. The first learning unit 44 generates a site detection unit 21E in which the detection of the region R6 in the image G5 is learned.

Further, in the same manner as the above-described method, the learning image g5 and the feature points f1 to f18 of the link 51B are used as teacher data, and the extraction of the feature points F1 to F18 in the captured image G5 is learned. The extraction unit 22E is generated by the second learning unit 45.

As the feature points to be learned, the points on the contour line along the longitudinal direction of each link 51B are designated as feature points f1 to f8, and the points for specifying the positions of the connecting portion and the hole of the link 51B are designated as feature points f9 to f18.

As shown in FIG. 20C, the region detection unit 21E and the feature point extraction unit 22E, which are trained models, can detect the region R6 of the link 51B with respect to the captured image G5 and extract the feature points F1 to F18. it can. Here, the wear amount calculation unit 37A sets the distance between the feature points F9 and F18 located on both sides of the link 51B as X9, and from the feature points F12 (or feature points F15) located between them, the feature points The wear amount (wear rate) of the link 51B is calculated by calculating X10 / X9, where the distance to the straight line connecting F9 and F18 is X10. Since X10 decreases due to wear and X9 hardly changes, the state of wear of the link 51B can be grasped by calculating X10 / X9. In the present embodiment, the amount of wear for one link 51B is calculated. For example, if the total length of the plurality of links 51B is set to X9 and the value of X10 / X9 is calculated, a more accurate link can be calculated. The wear of 51B can be grasped.

As described above, in the present embodiment, by selecting these parts for the parts of the plurality of parts of the work machine 5, the amount of wear of each part can be measured, and the parts that are easily worn by the work machine 5 can be measured. It is possible to propose the replacement time to the customer by using one wear amount measuring system 1. When the selected parts are the roller 51C, the shoe plate 51D, and the tooth 53D, the reference gauge setting unit 37C may set the reference gauge according to these shapes as in the case of the sprocket. ..

Although the details have been described above using the embodiment of the present invention, the specific configuration is not limited to this embodiment and the embodiment, and there are design changes within a range not deviating from the gist of the present invention. Also, they are included in the present invention.

In the modified example of the second embodiment, the imaging determination unit determines the result detected by the site detection unit to determine the timing of imaging by the imaging device. Of course, it may be applied to the wear amount measuring system of. Further, if such a determination is made, an image suitable for specifying the amount of wear can be acquired. Therefore, the wear of the part of the component may be specified from the acquired image without using the feature point extraction unit.

Further, in the second and third embodiments, only the second learning unit is provided as the learning unit, but the same learning unit as in the first embodiment may be provided in the arithmetic unit 2A of the second and third embodiments. Good.

In the present embodiment, the replacement time is predicted by calculating the degree of wear as the amount of wear. For example, the size of the gap between the reference gauge and the part where the wear is measured as shown in FIG. 8 is calculated. The device may calculate, output the image of the reference gauge and the part, and the calculation result of the size of the gap to the display device, and the operator or the like may determine the replacement of the target part.

1: Wear measurement system, 2A: Arithmetic logic unit, 3A: Arithmetic logic unit, 3B: Display device, 3C: Input device, 3E: Imaging device, 21: Site detection unit, 22: Feature point extraction unit, 5: Work machine, 34: Extraction judgment unit, 36: Feature point correction unit, 37A: Wear amount calculation unit, 37B: Replacement time calculation unit, 37C: Reference gauge setting unit, 51A: Sprocket (parts), 51a: Teeth (part), 51B: Link (part), 51C: Roller (part), 51h: Convex part, 51D: Shoe plate (part), 51g: Support part (part), 53D: Tooth, 53c: Tip (part)

Claims (10)

It is a wear amount measurement system equipped with an arithmetic unit that calculates the wear amount of the parts from an image captured including parts of a work machine.
The arithmetic unit
Using a learning image including a specific type of part of the work machine and a plurality of feature points for specifying the shape of a portion where the part is worn in the learning image as teacher data, the extraction of the feature point is a machine. A feature point extraction unit that is learned and extracts feature points of the portion in the captured image from the captured image,
A wear amount measuring system including a wear amount calculation unit for calculating the wear amount of the portion from the feature points. The wear amount measuring system includes an imaging device and a display device that displays the captured image from the imaging device.
The arithmetic device uses the learning image and the region of the worn portion of the component in the learning image as teacher data, and the detection of the region is machine-learned from the captured image to the captured image. It further includes a part detection unit that detects the area of the part of the part.
The wear amount measuring system according to claim 1, wherein the display device displays a region of the component detected by the site detection unit on the captured image. The wear amount measuring system according to claim 2, wherein the arithmetic unit further includes an extraction determination unit that determines execution of extraction of the feature points by the feature point extraction unit based on the detection result of the site detection unit. The wear amount measuring system includes an image pickup device and a display device for displaying an image from the image pickup device.
The arithmetic unit further includes a guide setting unit that sets a shooting guide according to the shape of the component on the image displayed by the display device, and an image acquisition unit that acquires the captured image from the imaging device. The wear amount measuring system according to claim 1. In the wear amount measuring system, the display device displays an image from the image pickup device.
The arithmetic unit further includes a guide setting unit that sets a shooting guide according to the shape of the component on the image displayed by the display device, and an image acquisition unit that acquires the captured image from the imaging device. Prepare,
The wear amount measuring system according to claim 2, wherein the image acquisition unit causes the image pickup apparatus to acquire the captured image at a timing when the site detection unit detects a region of the site in an image from the video. The wear amount measuring system includes a display device that displays the feature points extracted by the feature point extraction unit on the captured image together with the captured image.
An input device for inputting a change in the position of the feature point on the captured image with respect to the feature point displayed on the display device is further provided.
The arithmetic unit includes a feature point correction unit that corrects the position of the feature point in accordance with a change in the position of the feature point.
The wear amount measuring system according to claim 1, wherein the wear amount calculation unit calculates the wear amount of the portion based on the feature points corrected by the feature point correction unit. The wear amount measuring system according to claim 6, wherein the arithmetic unit relearns the feature point extraction unit using the captured image and the feature points including the modified feature points as further teacher data. The wear amount measuring system according to claim 1, wherein the arithmetic unit further includes a replacement time calculation unit that calculates a replacement time of the part from the wear amount of the part. The wear amount measuring system includes a display device that displays the captured image and the replacement time.
The arithmetic unit further includes a reference gauge setting unit that sets an image of a reference gauge according to the surface shape of the portion before wear with respect to the portion of the captured image displayed on the display device. 8. The wear amount measuring system according to 8. The arithmetic unit includes a plurality of the feature point extraction units according to a plurality of types of parts of the work machine.
The wear amount measuring system includes a component setting unit that sets one component from the plurality of types of components, and a component setting unit.
Further, an extraction unit selection unit that selects one feature point extraction unit from the plurality of feature point extraction units according to the parts set by the component setting unit is further provided.
The wear amount measuring system according to claim 1, wherein the feature point extraction unit selected by the extraction unit selection unit extracts the set feature points of the component.

Images