WO2020225890A1 - Point group analysis device, method, and program - Google Patents

Abstract

The present invention makes it possible to accurately determine the connection between a cable and an electric pole.　A point group analysis device that infers a connection point for a cable that is connected to an outdoor structure. The point group analysis device has an inference unit that associates a first area of the cable and a second area of the cable that is included in the first area and thereby infers whether the boundary between the first area and the second area is the connection point. The inference unit infers whether the boundary is the connection point by associating a point group for the second area and the shape of the second area as inferred from a first area model that is a model of the shape of the first area.

Description

Point cloud analyzers, methods, and programs

The present invention relates to a point cloud analyzer, method, and program, and more particularly to a point cloud analyzer, method, and program that models a linear structure from a point cloud consisting of three-dimensional points.

There are a huge number of electric power and communication infrastructure equipment, and in order to ensure the safety of the equipment, companies and local governments need to perform maintenance work on the equipment on a regular basis. For example, in utility pole inspection items, it is important not only to check the deterioration of cracks and deflections of the utility pole alone, but also to understand the number, direction, and tension of cables connected to the utility pole in order to understand the cause of the deterioration. It becomes. In other words, when the bent utility pole is rebuilt (renewal work), it is necessary to understand the connected structures in order to deal with the factors that cause it. For example, understanding the utility pole and surrounding structures, such as whether the tension of the cable connected to the bent utility pole should be relaxed or whether the branch line should be wired in the direction opposite to the cable, will lead to the examination of countermeasures.

However, since the number of infrastructure facilities is enormous and the range in which the facilities exist is wide, there is a problem that the amount of operation of the maintenance work is large.

On the other hand, in recent years, a car equipped with a camera or a laser scanner called a mobile mapping system (MMS (Mobile Mapping System)) travels in the city and travels on the surface of objects such as buildings and roads that are structures around the road. The use of a system that can measure the shape of a building is becoming widespread. This system can record the surface of an object as three-dimensional coordinate information using GPS (Global Positioning System) or IMS (Inertial Measurement Unit). It is expected that this technology will be used to automatically estimate the condition of equipment around roads. Here, the three-dimensional coordinate information is the three-dimensional coordinate information corresponding to the position in the real space, and the relative position in the coordinate information corresponds to the positional relationship in the real space.

By traveling the MMS outdoors and measuring with a laser, the surface shape of the infrastructure structure (hereinafter referred to as the subject) can be recorded with an accuracy of millimeters, and as in Non-Patent Document 1 and Patent Document 1, the cable Development of technology to model a quadratic curve (catenary curve or parabolic curve), etc., and development of technology to automatically extract the amount of deflection of a utility pole by applying a curved cylindrical model to a group of points measured with a utility pole ing.

By applying these technologies, it is possible to determine whether the models are connected by calculating the distance in the three-dimensional space for each model and checking whether the distance is close. For example, if the cable model exists within a certain radius from the utility pole, it is determined that the cable model is connected to the utility pole.

Patent No. 5981886, Nippon Telegraph and Telephone Corporation, Hitoe Arakaki, Jun Shimamura, Hiroyuki Arai, Yukinobu Taniguchi

However, regarding the connection between the cable and the utility pole, if there is an accessory (arm) of the utility pole, the cable exists at a position several meters away from the utility pole, so it is difficult to make a general judgment by threshold processing based on the distance. is there. Also, in urban areas, there are many utility poles that are not connected but are located near the cables, for example, a few centimeters from the cables, so if the threshold distance range is widened, even extra cables will be connected. It will be judged that it was. In other words, it can be seen that it is difficult to determine the connection only from the spatial information of the utility pole model and the cable model.

Further, a point cloud measured by a laser such as MMS has a property that it is difficult to measure a fine object as the distance from the measurement position increases, and it is difficult to measure a connecting fixing bracket between a cable and a utility pole. Difficult means that the laser point is difficult to hit because the object for connection is small. If the cable and the utility pole are separated, for example, there may be an arm (rod-shaped object) to attach, but occlusion occurs because there are many cables at the connection point between the utility pole and the cable. This is because there are many cases where it is difficult to measure the arm itself.

For example, even if the point group around the connection position is missing in the area of about 1 to 2 m, if the cable is modeled, whether or not it intersects the utility pole by extending the model, or spatially It is possible to know whether or not it exists nearby, but as mentioned above, it is difficult to determine the connection using only spatial information.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a point cloud analysis device, a method, and a program for accurately determining the connection between a cable and a utility pole.

In order to achieve the above object, the point cloud analysis device according to the first aspect is a point cloud analysis device that estimates the connection point of the cable connected to the outdoor structure, and is the first point cloud analysis device included in the cable. By associating the region with the second region which is a region included in the cable, it is estimated whether the boundary between the first region and the second region is the connection point. The estimation unit has an estimation unit, and the estimation unit has a shape of the second region estimated from a first region model that models the shape of the first region, and a point cloud of the second region. It is estimated whether it is the connection point by associating it.

The point cloud analysis method according to the second aspect is a point cloud analysis method for estimating a connection point of a cable connected to an outdoor structure, and is a first region included in the cable and the first region. By associating with the second region, which is a region included in the cable, it is estimated whether the boundary between the first region and the second region is the connection point, and in the estimation, the first region is estimated. By associating the shape of the second region estimated from the first region model that models the shape of the region with the point cloud of the second region, it is estimated whether the connection point is the connection point. ..

The program according to the third aspect includes the first region included in the cable and the first region by the estimation unit of the point cloud analyzer that estimates the connection point of the cable connected to the outdoor structure to the computer. By associating the region with the second region, which is a region included in the cable, it is estimated whether the boundary between the first region and the second region is the connection point, and in the estimation, the first region is estimated. By associating the shape of the second region estimated from the first region model that models the shape of one region with the point cloud of the second region, it is estimated whether the connection point is the connection point. It is a program to execute the process as if it were.

According to the point group analyzer, method, and program of the present invention, it is possible to accurately determine whether or not a utility pole is connected to a linear structure such as a cable or a drop wire, and the utility pole is affected by the influence of an arm or the like. It is possible to accurately determine the connection of a cable that is separated from the cable or a cable that passes near the utility pole but is not connected to the utility pole. Further, when connecting a cable away from a utility pole, it is possible to prevent an erroneous determination that a linear structure that is not a communication line or a linear structure that is not a power line is connected to a utility pole.

It is an image diagram of the two utility poles and the cables existing around them as seen from directly above. It is an image diagram of the two utility poles and the cables existing around them as viewed from directly above, and is an image diagram when the cables are divided. It is an image diagram for showing the deformation of the cable connected to a utility pole. It is an image diagram of the estimation result of the connection boundary point position when searching the connection boundary point. It is an image diagram when the model parameter is estimated in each area. It is a block diagram which shows an example of the functional structure of the point group analysis apparatus which concerns on 1st Embodiment. It is a figure which shows an example of the setting range of the area of interest for each wire model. It is a figure which shows an example of the structure of the boundary point detection part of 1st Embodiment. It is a schematic block diagram which shows an example of the computer which functions as a point cloud analysis apparatus. It is a flowchart which shows an example of the processing flow by the program which concerns on 1st Embodiment. It is a flowchart which shows an example of the process flow of the calculation of the connection boundary degree which concerns on 1st Embodiment. It is a block diagram which shows an example of the functional structure of the point cloud analysis apparatus which concerns on 2nd Embodiment. It is a figure which shows an example of the structure of the boundary point detection part of the 2nd Embodiment. It is a figure for demonstrating the normal angle when there is an external force. It is a figure which shows an example of the shape of the cable when there is an external force and when there is no external force. It is a flowchart which shows an example of the process flow of the calculation of the connection boundary degree which concerns on 2nd Embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

The method according to the embodiment of the present invention is a technique for estimating the connection relationship between the cable and the surrounding structure from the point cloud measured by the point cloud analysis. In particular, it is a technique for determining the connection between a cable and a utility pole.

In the embodiment of the present invention, the "cable" is represented as a linear structure for infrastructure equipment including all optical fiber cables, metal cables for telephones, high-voltage lines for electric power, and low-voltage lines. On the other hand, for example, an elongated structure such as a rope for a hanging curtain in a shopping district or a clothesline for washing is represented as a linear structure that is not a cable.

These linear structures can be expressed as a curve model from the measured point cloud by applying a curve model as in Non-Patent Document 1. In the following description, the cable is modeled as a curve model (called a wire model) by the point cloud analysis technique, and the modeled one is used to determine whether or not the utility pole and the curve model are connected. The wire modeling of the cable can be automatically generated by, for example, an existing method.

One of the merits of modeling is that you can check the physical shape information of the cable of interest for a certain cable of interest. For example, when you want to check the length, if you model it, you can calculate it from the distance between the end points (start point and end point), and by calculating the distance between the models, you can calculate the separation distance between the cable of interest and the surrounding cable. It is also possible to ask.

In other words, in the state of the measured point cloud, it is not known which point is the point that measured the surface of which subject, but by point cloud analysis, it is possible to grasp which point represents what, and by modeling the subject's surface. Know the physical information.

The entire processing outline of the embodiment of the present invention will be described. 1A and 1B are image views of the two utility poles and the cables existing around them as viewed from directly above. FIG. 1B shows an image diagram when the connection is determined and the cable is divided. The end point of the cable model (wire model) is indicated by the broken line, and the connection position (connection point) of the cable and the utility pole is indicated by the alternate long and short dash line.

As shown in FIG. 1A, it is assumed that there is a connection relationship between the utility poles 1 to the utility pole 2 to which the wire model belongs. In this case, first check the existence of the utility pole model. As shown in FIG. 1B, when it is determined that the utility pole is connected, the wire model is divided and the connection relationship between the utility poles is re-registered.

As shown in FIG. 2, in the cable 1 connected to the utility pole, the shape of the cable is deformed from the shape of the quadratic curve because the cable 1 is deformed at the position connected to the utility pole (connection position Q). .. Since the cable 2 is not deformed, it can be sufficiently expressed as one quadratic curve. Actually, there is measurement noise in the measured point cloud, and it may be difficult to visually understand the inflection point as clearly as the exaggerated image diagram.

When not connected to a utility pole, when focusing on a certain area, the approximation accuracy is high even with one quadratic curve model (the difference between the two models is small). On the other hand, when the shape is deformed due to the connection with the utility pole, the shapes deviate from each other, so that the approximation accuracy is higher when the two curves are approximated with the position as the boundary. Therefore, it is possible to estimate the position of the connection boundary point by examining whether or not the approximation accuracy is higher when the region is modeled separately.

In the embodiment of the present invention, paying attention to the above-mentioned properties, the degree of connection boundary is obtained as an index for determining whether or not the utility pole is connected at the connection position. In the embodiment of the present invention, the connection position between the cable and the utility pole is hereinafter referred to as a connection boundary point. FIG. 3 is an image diagram of the estimation result of the connection boundary point position when the connection boundary point is searched. In the schematic diagram of FIG. 3, a region _Pi (attention region) of a certain size and a connection boundary degree at the position Q _i are expressed centering on the attention position Q _i set by the user. As shown in FIG. 3, the connection boundary degree determines that the connection point exists at the position of the peak (maximum value) of the connection boundary degree.

In the embodiment of the present invention, the determination is made based on the degree of connection boundary, which is an index for determining whether or not the utility pole is connected. Specifically, as shown in FIG. 4, in the two regions P _ia and P _ib set for the target area P _i at the target position Q _i on the cable, to estimate the model parameters at each area, each area It is considered that the lower the correlation of the parameters in, the higher the possibility of division. The symbol i is a number for distinguishing the region of interest, and takes an integer value of i = 1,2,3, ..., Np (Np is the total number of regions of interest). In the example of FIG. 4, has been drawn only i = 1 area of the region P _i, the left area when the area P _ia is that half the area P _i at the location of the Q, region P _ib right area It has become.

Here, instead of analyzing the entire cable by dividing it into two regions, it is important to pay attention to the region within a certain range from the attention position as shown in FIG. 4 with two connecting boundary points on one cable. This is because there is a possibility that there is more than that. That is, when there is only one connection boundary point, there is no problem even if the entire cable is set to two regions and the processing of the embodiment of the present invention is performed, but when there are two or more, the connection boundary degree is correct. Is not calculated, so it is necessary to analyze each part of the area.

Here, the model parameters to be estimated are all the model parameters in the divided cable. For example, when there are three connecting boundary points, four models can be obtained. Since the model parameters are obtained from the data (point cloud) having the X, Y, and Z coordinate values, the model represented by the model parameters is a concept including the position and the direction. From the wire model based on the model parameters, the angle between the vertical direction and the model (the plane on which the quadratic curve exists), that is, the inclination direction of the model with respect to the horizontal plane can be known. In addition, if there is a ground point cloud, the minimum ground clearance from the ground and the three-dimensional position that is the minimum ground clearance can be obtained from the distance from the model.

Also, in order to be connected to a utility pole, it is necessary that there is a utility pole to be connected around the cable of interest. Therefore, in the embodiment, the connection boundary point position is estimated accurately by considering not only the shape of the cable of interest but also the relative positional relationship with the surrounding utility poles.

[First Embodiment]
<Configuration of point cloud analysis device according to the first embodiment>

FIG. 5 is a block diagram showing an example of the functional configuration of the point cloud analysis device 10 according to the first embodiment.

As shown in FIG. 5, the point cloud analysis device 10 includes a calculation unit 20, a three-dimensional data storage unit 12, and an input unit 14.

The three-dimensional data storage unit 12 is a device that stores point cloud information. Here, the three-dimensional point cloud is assumed to be data mounted on a moving body such as a fixed laser sensor or MMS, and the point cloud representing a three-dimensional point on an object is measured while scanning the measurement position. doing. The measurement sensor may be a device capable of measuring the distance between the subject and the sensor, such as a laser range finder, an infrared sensor, or an ultrasonic sensor.

A measurement system mounted on a moving body is, for example, a laser range finder mounted on a GPS-equipped vehicle or an airplane equipped with GPS and measured while moving to create an outdoor environment. It is a system that measures the three-dimensional position of the surface of an object, for example, a cable, a building, a guardrail, a road ground, or the like. In the first embodiment, the data input to the three-dimensional data storage unit 12 measures the position on the surface of the object of the subject by using an MMS equipped with GPS and a laser range finder on the vehicle. Let it be a three-dimensional point cloud which is the measurement result. In the first embodiment, the object of the subject is a utility pole, a cable, and other branch lines related to the utility pole and a structure existing around the utility pole.

Further, the three-dimensional data storage unit 12 holds each of the wire models representing the cables obtained from the acquired point cloud. It also holds each of the pole models that represent utility poles. The wire model and the utility pole have information on the connection relationship of which utility pole is connected from the end point position of the wire model.

The wire model is an Nth-order polynomial or spline curve existing in three-dimensional space, or a piecewise smooth continuous line. Mathematically, it means a continuous line segment that cannot be differentiated in some sections, and for example, y = | x | (absolute value) that cannot be differentiated at the origin corresponds.

The wire model is represented as the central axis of the cable by a continuous line with two end points. Two endpoints can be determined on the central axis, and physical values such as the three-dimensional position and the ground height of the wire model can be uniquely determined using only one parameter according to the distance from one endpoint. Defined as a model. Here, in the wire model, it is not necessary to distinguish between the start point and the end point at the end points, and it is sufficient that two end points can be set on the central axis. A wire model can be detected for a three-dimensional point cloud by an existing method. Hereafter, it is assumed that the central axis of the cable is represented by a wire model, ignoring the thickness, and that the wire model detected in advance from the point cloud is stored in the three-dimensional data storage unit 12, and the wire model. The method of detecting the connection boundary point position will be described in.

In the embodiment of the present invention, the wire model corresponds to one cable stretched in a certain utility pole section. That is, a cable that is deformed by tension or the like or a cable that is not deformed is represented by a wire model. For example, when the cable of interest is deformed due to the connection with the utility pole, the cable is represented by two quadratic curves. The wire model is composed of these two quadratic curves, and there is a restriction that the three-dimensional positions of the two model end points match at the connection boundary point position.

For example, a deformed cable is represented by a wire model composed of a plurality of quadratic curves, and a cable without deformation is represented by a wire model composed of one quadratic curve. The number of quadratic curves constructed is determined by the connection boundary points, and the quadratic curves that make up the same wire model are discontinuous at the connecting positions. That is, the wire model is a line segment having two end points composed of one quadratic curve or a plurality of quadratic curves.

The input unit 14 is a user interface such as a mouse or a keyboard, and receives parameters used by the point cloud analysis device 10 as input. The parameter is, for example, information for collating a model such as a utility pole position and a cable position obtained in advance with a point cloud. Further, the input unit 14 may be an external storage medium such as a USB (Universal Serial Bus) memory that stores measurement information.

The calculation unit 20 includes a region of interest setting unit 30, a boundary point detection unit 32, and a connection division unit 34. The boundary point detection unit 32 is an example of the estimation unit.

The calculation unit 20 performs the following processing for each of the wire models representing the cables stored in the three-dimensional data storage unit 12.

The area of interest setting unit 30 window-searches the area at the end or the middle of the wire model for the wire model including the quadratic curve model representing the cable, which is obtained from the point cloud consisting of three-dimensional points on the object. The attention area obtained by the above method, which is divided into a first area and a second area, is set.

Here, the area of interest means an area of a certain size set by the user for searching for the position where the cable is deformed by connecting with the utility pole. Specifically, the area of interest is set while shifting the area of interest at regular intervals from the end point of the wire model representing a certain cable to the position of the other end point. If there is a utility pole in the middle of the wire model, pay attention to the position from a predetermined distance on one side to the position of the end point on the other side with respect to the position of the wire model corresponding to the utility pole in the middle part. Set the area of interest while shifting the area at regular intervals.

Speaking of two-dimensional image analysis processing, window search corresponds. It corresponds to the process of setting a certain region ROI (Region of Interest) set by the user at various positions on the input data. In the embodiment of the present invention, since the connection boundary point is on the cable, it is sufficient to set this ROI at a position between the end points of the wire model representing the cable.

FIG. 6 is a diagram showing an example of the setting range of the region of interest for each wire model. As shown in FIG. 6, the region of interest P is set in a range based on the start point and the end point of the end points. In addition, any one of the end points may be arbitrarily determined as a start point and an end point. D indicating the size of the region is a value set by the user, and d = 2.0 [m] in the present embodiment.

The boundary point detection unit 32 compares the information regarding the first region with the information regarding the second region based on the point cloud and the wire model included in the region of interest for each of the regions of interest. The connection boundary degree, which represents the degree to which the division positions of the first region and the second region are the connection points between the cable and the utility pole, is calculated. Further, the boundary point detection unit 32 detects the connection boundary point, which is the connection point between the cable and the utility pole represented by the quadratic curve model, based on the connection boundary degree calculated for each of the regions of interest.

FIG. 7 is a diagram showing an example of the configuration of the boundary point detection unit 32 of the first embodiment. As shown in FIG. 7, the boundary point detection unit 32 includes a connection area error estimation unit 40 and a connection boundary degree calculation unit 42.

Here, the principle of obtaining the connection boundary degree as a model approximation error by the connection boundary degree calculation unit 42 will be described. In the first embodiment, it is the method shown by the following realization method 1 or realization method 2. Estimating the connecting boundary degree determined from the shape of the cable in the attention region P _i, determines a connection boundary points by thresholding.

For example, there are the following realization method 1 and realization method 2 using the error obtained in the first region or the second region, and any of the methods may be used. In the first embodiment, the realization method 2 is used.

The realization method 1 is a quadratic curve model obtained from either the first region or the second region, as shown in the following equation (4), and represents the ratio at which the cable shape in the other region can be estimated. The connection boundary degree is obtained as the model approximation error E. It is considered that the correlation is so high that the shape of the cable (point group on the cable) in the other region can be predicted by the quadratic curve model estimated from one region. That is, for each of the two regions, it is output that the larger the approximation error of the point group when using the curve model obtained from the other region, the larger the connection boundary degree.

This means that the more difficult it is to predict the shape of one region from the other region, the higher the degree of connection boundary. That is, it means that the shapes of the cables existing in both regions are correlated so that the other region can be estimated from one region, that is, the deformation does not occur at the boundary point position of the region of interest.

Position Q all of the points in the target region in the _i P _i, the point group p _i1 in the region of interest _1, the point group in the region of interest 2 When p _i2, connecting boundary degree determined by the following equation E.

... (4)

here,

Is the error between the point p _i1 included in the first region and the quadratic curve model M _i2 obtained from the point cloud included in the second region. In other words, a quadratic curve in the first region to be estimated from the secondary curve model M _i2, an error between the point p _i1 of the first region.

Is the error between the point p _i2 included in the second region and the quadratic curve model M _i1 obtained from the point cloud included in the first region. In other words, the secondary curve of the second region to be estimated from the secondary curve model M _i1, which is an error between the point p _i2 of the second region.

N2 is the number of point clouds included in the second region. N1 is the number of point clouds included in the first region.

In the realization method 2, as shown in the following equation (5), the model approximation error is obtained by subtracting the larger of the first error and the second error from the attention area error. The region of interest error is an error between the quadratic curve model obtained from the point cloud included in the region of interest and the point cloud included in the region of interest. The first error is an error between the quadratic curve model obtained from the point cloud included in the first region and the point cloud included in the first region. The second error is an error between the quadratic curve model obtained from the point cloud included in the second region and the point cloud included in the second region. The connection boundary degree E may be obtained by the following equation (5).

... (5)

The larger the amount of deformation of the cable of interest due to the generation of large tension such as the connection point, the smaller the first error and the second error than the area of interest error. For equation (5), the average of the first error and the second error may be subtracted from the region of interest error instead of Max.

The reason for taking the difference between the first error and the second error, whichever is larger (max value), is to deal with the case where there is a small accessory in either area. In this case, it is conceivable that the approximation error becomes small only in one region and the deviation from the attention region error becomes large. Therefore, in order to suppress the determination that there is a connection boundary point even though there is almost no deformation, the difference is calculated using the average value and the max value. In order to estimate the presence or absence of deformation from the approximation error, it is necessary to suppress the influence of measurement error and accessories.

As described above, the connection region error estimation unit 40 according to the present embodiment has an attention region error, which is an error between the quadratic curve model obtained from the point group included in the attention region and the point group included in the attention region. The first error, which is the error between the quadratic curve model obtained from the point group included in the first region and the point group included in the first region, and the point group obtained from the second region 2 The second error, which is the error between the next curve model and the point cloud included in the second region, is estimated.

The connection boundary degree calculation unit 42 calculates the connection boundary degree based on the model approximation error, which is the difference between the region of interest error and the larger of the first error and the second error, according to the above equation (5). ..

The boundary point detection unit 32 determines the peak position of the connection boundary degree at the connection point of the cable represented by the quadratic curve model based on the connection boundary degree calculated for each of the regions of interest by the connection boundary degree calculation unit 42. Detected as a connection boundary point.

The connection dividing unit 34 divides the wire model based on the connection boundary point detected by the boundary point detection unit 32 and the position of the utility pole. In the division of the wire model, for example, it is determined from the position of the connection boundary point whether or not the utility pole exists in a predetermined range, and if it exists, the wire model is divided and the utility pole connected at the end point of the wire model. Reconnect. The threshold value of the distance to the utility pole may be determined based on the standard length of the arm metal installed on the utility pole. In the embodiment of the present invention, it is set to 2 [m].

Regarding the degree of connection boundary, it may be determined that the connection is made when the threshold value is exceeded. This threshold value depends on the measurement conditions of the laser sensor and the like. Therefore, the threshold value may be determined as, for example, the average value of the connection boundary degree obtained in the region of interest at the position where the actual utility pole and the cable are connected. When it is desired to detect the connection boundary point position with high recall (without omission), a threshold value may be set as the minimum value of the connection boundary degree obtained in the region of interest at the position where the cable and the utility pole are actually connected.

FIG. 8 is a schematic block diagram showing an example of a computer functioning as a point cloud analysis device. The point cloud analysis device 10 is realized by the computer 84 shown in FIG. 8 as an example. The computer 84 includes a CPU 86, a memory 88, a storage unit 92 that stores the program 82, a display unit 94 that includes a monitor, and an input unit 96 that includes a keyboard and a mouse. The CPU 86, the memory 88, the storage unit 92, the display unit 94, and the input unit 96 are connected to each other via the bus 98.

The storage unit 92 is realized by an HDD, SSD, flash memory, or the like. A program 82 for making the computer 84 function as the point cloud analysis device 10 is stored in the storage unit 92. The CPU 86 reads the program 82 from the storage unit 92, expands the program 82 into the memory 88, and executes the program 82. The program 82 may be stored in a computer-readable medium and provided.

<Operation of the point cloud analyzer according to the first embodiment>

Next, with reference to FIG. 9, the operation of the point cloud analysis device 10 according to the first embodiment will be described. Note that FIG. 9 is a flowchart showing an example of the processing flow by the program 82 according to the first embodiment.

When the point cloud analysis device 10 according to the first embodiment is instructed to execute the point cloud analysis process by the operation of the operator, the CPU 86 reads out and executes the program 82 stored in the storage unit 92. The following processing is executed for the wire model. Further, the following processing executed for each quadratic curve model can be calculated in parallel in the program 82.

First, in step S100, the attention area setting unit 30 is obtained in advance from the input unit 14, the wire model group representing the cable between the three-dimensional point cloud and the utility pole, the pole model group representing the utility pole, and the input unit 14. Get the parameters.

In step S102, the area of interest setting unit 30 is the area of interest obtained by window-searching the wire model including the quadratic curve model around the utility pole at the end or intermediate portion of the wire model, and is the first area. and setting the divided region of interest P _i to the second region. The region of interest _Pi is set to move at a predetermined interval S [m] within a predetermined range including the position corresponding to the electric pole of the quadratic curve model.

In step S104, the boundary point detection unit 32 compares the information regarding the first region with the information regarding the second region based on the point cloud and the quadratic curve model included in the region of interest for each of the regions of interest. Then, the connection boundary degree is calculated.

In step S106, the boundary point detection unit 32 determines whether or not the connection boundary degree has been detected up to the end point of the quadratic curve model, and if the boundary point detection unit 32 has detected the end point, the process proceeds to step S108, and if not, the end point is not detected. Returning to S102, the next region of interest _Pi is set, and the process is repeated.

In step S108, the boundary point detecting unit 32 detects the center position of the region of interest P _i the connecting boundary of the quadratic curve model becomes the maximum value as a connecting boundary points. The maximum value is, for example, a point at which the peak is as shown in FIG.

In step S110, the connection dividing unit 34 divides the wire model based on the connection boundary point detected by the boundary point detection unit 32 and the position of the utility pole.

Next, the details of the process in step S104 will be described. FIG. 10 is a flowchart showing an example of a flow of processing for calculating the connection boundary degree according to the first embodiment.

At step S200, connection area error estimator 40, the attention region _{P i,} estimating a first error and the second error of the second region _{P i2} of the first region _{P i1.} Specifically, the first error between the quadratic curve model _Mi1 obtained from the point cloud included in the first region and the point cloud included in the first region _Pi1 , and the points included in the second region. The second error between the quadratic curve model _Mi2 obtained from the group and the point cloud included in the second region _Pi2 is estimated.

At step S202, connection area error estimator 40, the attention region _{P i,} estimates the attention area error of the attention region _{P i.} Specifically, it estimates the attention area error between the point group included in the quadratic curve model M _i region of interest P _i obtained from a group of points included in the target region.

In step S204, the connection boundary degree calculation unit 42 obtains the model approximation error obtained by subtracting the larger of the first error and the second error from the attention area error according to the above equation (5). calculated as the concatenation boundary of the P _i.

As described above, the point cloud analysis device 10 according to the first embodiment provides information about the first region and information about the first region based on the point cloud included in the region of interest and the quadratic curve model for each region of interest. By comparing with the information about the second region, the connection boundary degree indicating the degree to which the division positions of the first region and the second region of the region of interest are the connection points between the cable and the utility pole is calculated, and the region of interest is calculated. Based on the connection boundary degree calculated for each of the above, the connection boundary point which is the connection point of the cable represented by the quadratic curve model is detected. Thereby, the connection point with the utility pole can be detected in consideration of the shape of the cable.

Further, the connection region error estimation unit 40 described above has a first error, which is an error between the quadratic curve model obtained from the point group included in the first region, and the point group included in the second region, and a first error. The second error, which is the error between the quadratic curve model obtained from the point group included in the second region and the point group included in the first region, is estimated, and the connection boundary degree calculation unit 42 performs the above (4). According to the equation, the connection boundary degree may be calculated based on the model approximation error, which is the larger of the first error and the second error.

[Second Embodiment]
In the first embodiment, the connection boundary degree is calculated based on the magnitude of the approximation error, but in the second embodiment, the connection boundary degree is calculated by machine learning using the feature vector. The same parts as those in the first embodiment are designated by the same reference numerals, and the description thereof will be omitted.

FIG. 11 is a block diagram showing an example of the functional configuration of the point cloud analysis device 210 according to the second embodiment.

The boundary point detection unit 232 obtains a predetermined feature amount vector for each of the regions of interest, inputs it to a model for obtaining a connection point with a utility pole learned in advance, and calculates the connection boundary degree.

FIG. 12 is a diagram showing an example of the configuration of the boundary point detection unit 232 of the second embodiment. As shown in FIG. 12, the boundary point detection unit 232 includes a connection area error estimation unit 40, a connection boundary degree calculation unit 242, and a vector calculation generation unit 244. The vector calculation generation unit 244 includes a plane approximation calculation unit 252, a peripheral distance detection unit 254, an arm metal detection unit 255, and a vector generation unit 256.

The vector calculation generation unit 244 represents plane information representing the error between the plane and the point group obtained from the point group included in the attention region, and the periphery representing the information between the quadratic curve model around the attention region and the attention region. The information is calculated to generate a feature vector including the model approximation error, the plane information, and the peripheral information.

Specifically, the feature amount vector is generated by connecting the following feature amounts (1) to (4). (1) and (2) are features related to shape deviation, and (3) and (4) are features related to the relative positional relationship with the surrounding structure. The plane information is the feature amount of (2), and the peripheral information is the feature amount of (3) and (4). As for the connection of the feature amounts, all of them may be combined, or they may be selected and combined.

(1) The model approximation error obtained by the same method as the boundary point detection unit 32 of the first embodiment, and the number of points included in the first region and the second region of the region of interest are used as feature quantities. To do. When there is no connecting point, there is no deformation of the quadratic curve, and it is possible to estimate the shape of one region from either the point group information (model) of the first region and the second region. The model approximation error obtained by the above equation (5) increases as the quadratic curve is deformed. Therefore, the larger the deformation, the higher the probability of existence of the connection point. It should be noted that the more end points, the smaller the number of points included in the region.

(2) The feature amount is the plane approximation error of the region of interest. The plane approximation error is an error between a plane that approximates a point cloud included in the region of interest and a point cloud in the region of interest. When there is a connecting point, tension is generated by the utility pole, so that the plane that approximates the point cloud included in the region of interest is likely to deviate from the quadratic curve shape of the cable (see FIG. 13). That is, it can be determined that the higher the degree of shape deviation from the plane that approximates the point cloud included in the region of interest, the higher the probability of existence of the connecting point. FIG. 14 is a diagram for explaining a normal angle when there is an external force. FIG. 13 is a diagram showing an example of the shape of the cable when there is an external force and when there is no external force.

(3) A feature of the distance between the position of the utility pole and the center position of the area of interest. Due to the characteristics of MMS measurement, the point group near the intersection of utility poles is often missing due to the influence of occlusion. Therefore, even if there is a point cloud defect near the intersection position of the utility pole and the cable of interest, if there is a connection point, the utility pole is often located near the cable. That is, when there is a utility pole in the vicinity, there is a high possibility that a connection point by the utility pole exists.

(4) It is a feature amount of the criterion for judging whether or not the arm is present. This is because it is possible that there is a connecting point in the vicinity when there is an arm.

The plane approximation calculation unit 252 calculates, as plane information, a plane approximation error representing an error between a plane that approximates a point group included in the region of interest and a point group of the region of interest for each of the regions of interest. Further, the plane approximation calculation unit 252 calculates the plane approximation normal angle, which is the angle formed by the plane normal and the horizontal plane that approximate the point group included in the region of interest.

Peripheral distance detection unit 254 calculates the distance between the area of interest and the nearest utility pole as peripheral information for each of the areas of interest. For the distance to the utility pole, the shortest distance between the straight line (central axis) connecting the lower end and the upper end of the utility pole and the central position of the region of interest may be calculated. Physically, it means the distance from the central position of the region of interest to the foot of the perpendicular to the central axis of the utility pole.

Arm-detection unit 255, for each region of interest, as the peripheral information, as if the criterion arm-exists, points to calculate the shortest distance from the utility pole position to the center position of the region of interest P _i region Calculate the number of groups.

The vector generation unit 256 generates a feature amount vector including a model approximation error, plane information, and peripheral information for each of the regions of interest.

The connection boundary degree calculation unit 242 calculates the connection boundary degree for each of the regions of interest using a predetermined machine learning method based on the feature amount vector generated by the vector generation unit 256. The machine learning method may be any method such as logistic regression analysis and rank learning such as AdaRank and RankNet. The machine learning model used for calculating the connection boundary degree may be learned in advance using the learning data in which the correct answer label is attached to the feature quantity vector. Further, since the model approximation error includes the number of point groups in each of the first region and the second region, the connection boundary degree calculation unit 242 acquires the number of point groups in each region.

<Operation of the point cloud analyzer according to the second embodiment>

Next, the operation of the point cloud analyzer 210 according to the second embodiment will be described. The flow of processing by the program 82 according to the second embodiment is the same as the flowchart showing an example of FIG. 9 of the first embodiment, and in step S108, the connection boundary degree calculation unit 242 generates a vector. Based on the feature amount vector generated by the part 256, the connection boundary degree is calculated by using a predetermined machine learning method.

In the second embodiment, the process of calculating the connection boundary degree in step S104 is different from that in the first embodiment. FIG. 15 is a flowchart showing an example of a flow of processing for calculating the connection boundary degree according to the second embodiment.

In step S1200, the connection area error estimation unit 40 and the connection boundary degree calculation unit 242 calculate the model approximation error of the region of interest. The model approximation error is calculated by performing the same processing as in steps S200 to 204 of FIG.

In step S1202, the connection boundary degree calculation unit 242 acquires the number of point groups in each of the first region and the second region for the region of interest.

In step S1204, the plane approximation calculation unit 252 calculates, as plane information, a plane approximation error representing an error between the plane that approximates the point group included in the region of interest and the point group of the region of interest.

In step S1206, the peripheral distance detection unit 254 detects the distance between the position of the utility pole and the center position of the region of interest as peripheral information.

In step S1208, arm-detection unit 255, a determination of whether the reference arm-exists, calculates the number of the points in calculating the shortest distance from the utility pole position to the center position of the region of interest P _i region .. That is, the amount of the point cloud present is calculated on the line segment from the center position of the region of interest to the foot of the perpendicular line from the center axis of the utility pole. Specifically, the number of points existing in the region from the shortest distance line segment to the distance r is used. The distance r may be set to include, for example, the thickness of the arm, and in this embodiment, r = 0.2.

In step S1210, the vector generation unit 256 generates a feature amount vector including a model approximation error including the number of point groups in each region, plane information, and peripheral information.

As described above, in the point group analysis device 10 according to the second embodiment, the boundary point detection unit 32 includes a model approximation error, plane information, and peripheral information for each of the regions of interest. Is generated, the connection boundary degree is calculated using a predetermined machine learning method, and the connection point, which is the connection point of the cable represented by the quadratic curve model, is calculated based on the connection boundary degree calculated for each of the regions of interest. Detect the boundary point. Thereby, the connection point with the utility pole can be detected in consideration of the shape of the cable.

Also, the feature amount vector includes the feature amount of the relative positional relationship with the surrounding structure (arm, etc.). This makes it possible to estimate the connection relationship by connecting utility poles in consideration of the surrounding structures.

In the learning machine learning, as shown in FIG. 4, as the value of the distance DQi consolidated boundary point positions and Q _i are close, so connecting boundary degree is high, to set the consolidated boundary of each feature amount. In other words, connecting the boundary of extraction with position Q _i may be calculated using a function such as decreases in inverse proportion, for example, in distance.

... (6)

As the machine learning method, for example, logistic regression analysis or rank learning may be used.
For an unknown cable (wire model) that is not used for learning, the connection boundary degree is calculated using the learning result for the feature amount at the point of interest position of the wire model, and the position that becomes the maximum value is the connection candidate position (connection candidate position). It can be obtained as the connection boundary point position).

If the value of the connection boundary degree at this candidate position is equal to or greater than the threshold value, it may be determined that the connection is made. This threshold value may be determined using, for example, the average value of the degree of connection boundary obtained from the feature amount at the actual connection position. If you want to detect the connection boundary point position with high recall (without omission) even if there are some errors, set the threshold value using the minimum value of the connection boundary degree actually obtained from the feature amount at the connection position. You may.

The point cloud analyzer and method have been illustrated and described above as embodiments. The embodiment may be in the form of a program for making the computer function as each part included in the point cloud analysis device. The embodiment may be in the form of a storage medium that can be read by a computer that stores this program.

In addition, the configuration of the point cloud analysis device described in the above embodiment is an example, and may be changed depending on the situation within a range that does not deviate from the gist.

Further, the processing flow of the program described in the above embodiment is also an example, and even if unnecessary steps are deleted, new steps are added, or the processing order is changed within a range that does not deviate from the purpose. Good.

Further, in the above embodiment, the case where the processing according to the embodiment is realized by the software configuration by using the computer by executing the program has been described, but the present invention is not limited to this. The embodiment may be realized by, for example, a hardware configuration or a combination of a hardware configuration and a software configuration.

10 Point cloud analysis device 12 3D data storage unit 14 Input unit 20 Calculation unit 30 Attention area setting unit 32 Boundary point detection unit 34 Connection division unit 40 Connection area error estimation unit 42 Connection boundary degree calculation unit 210 Point cloud analysis device 232 Boundary Point detection unit 242 Connection boundary degree calculation unit 244 Vector calculation generation unit 252 Plane approximation calculation unit 254 Peripheral distance detection unit 255 Arm metal detection unit 256 Vector generation unit

Claims (9)

1. A point cloud analyzer that estimates the connection points of cables connected to outdoor structures.
   By associating the first region included in the cable with the second region which is the region included in the cable, the boundary between the first region and the second region is the said. It has an estimation unit that estimates whether it is a connection point,
   The estimation unit
   It is estimated whether it is the connection point by associating the shape of the second region estimated from the first region model that models the shape of the first region with the point cloud of the second region. Point cloud analyzer.
2. The estimation by the estimation unit is performed by each process of the attention area setting unit and the boundary point detection unit.
   The attention area setting unit is an attention area obtained by window-searching the wire model for a wire model including a quadratic curve model representing a cable, which is obtained from a point cloud consisting of three-dimensional points on an object. Then, set a plurality of areas of interest divided into the first area and the second area.
   For each of the attention regions, the boundary point detection unit obtains information on the first region and information on the second region based on the point cloud included in the attention region and the quadratic curve model. By comparison, the connection boundary degree indicating the degree to which the division positions of the first region and the second region of the attention region are the connection points with the utility pole was calculated, and calculated for each of the attention regions. The point cloud analysis device according to claim 1, wherein a connection boundary point, which is a connection point between the cable represented by the wire model and the utility pole, is detected based on the connection boundary degree.
3. The point cloud analysis device according to claim 2, further comprising a connection dividing portion that divides the wire model based on the detected connection boundary point and the position of the utility pole.
4. The boundary point detection unit is configured to include a connection area error estimation unit and a connection boundary degree calculation unit.
   The connection area error estimation unit is
   From the first error, which is the error between the quadratic curve model obtained from the point cloud included in the first region and the point cloud included in the second region, and the point cloud included in the second region. The second error, which is the error between the obtained quadratic curve model and the point cloud included in the first region, is estimated.
   The point cloud according to claim 2 or 3, wherein the connection boundary degree calculation unit calculates the connection boundary degree based on a model approximation error which is the larger of the first error and the second error. Analytical device.
5. The boundary point detection unit is configured to include a connection area error estimation unit and a connection boundary degree calculation unit.
   The connection region error estimation unit includes a quadratic curve model obtained from a point cloud included in the attention region, an attention region error which is an error between the point cloud included in the attention region, and the first region. A first error, which is an error between the quadratic curve model obtained from the point cloud and the point cloud included in the first region, and a quadratic curve model obtained from the point cloud included in the second region. Estimate the second error, which is the error from the point cloud included in the second region,
   Claim 2 or claim that the connection boundary degree calculation unit calculates the connection boundary degree based on a model approximation error which is a difference between the attention area error and an error based on the first error and the second error. Item 3. The point cloud analyzer according to item 3.
6. The boundary point detection unit is configured to further include a vector calculation generation unit.
   The vector calculation generation unit represents plane information representing an error between the plane obtained from the point group included in the attention region and the point group, and information between the electric column around the attention region and the attention region. The peripheral information is calculated, and a feature quantity vector including the model approximation error, the plane information, and the peripheral information is generated.
   The point cloud analysis device according to claim 4 or 5, wherein the connection boundary degree calculation unit calculates the connection boundary degree using a predetermined machine learning method based on the feature quantity vector.
7. The vector calculation generation unit includes a plane approximation calculation unit, a peripheral distance detection unit, an arm metal detection unit, and a vector generation unit.
   The plane approximation calculation unit uses, as the plane information, a plane approximation error representing an error between a plane that approximates the point group included in the attention region and the point group of the attention region, and a point group included in the attention region. Calculate the plane-approximate normal angle, which is the angle between the approximated plane normal and the horizontal plane.
   The peripheral distance detecting unit detects peripheral distance information including the distance between the position of the electric column in the periphery and the center position of the region of interest as the peripheral information.
   The arm metal detection unit calculates the number of point clouds in a predetermined area up to the center position of the attention area, which is a criterion for determining whether or not the arm metal exists, as the peripheral information.
   The point cloud analysis device according to claim 6, wherein the vector generation unit generates a feature quantity vector including the model approximation error, the plane information, and the peripheral information.
8. A point cloud analysis method that estimates the connection points of cables connected to outdoor structures.
   By associating the first region included in the cable with the second region which is the region included in the cable, the boundary between the first region and the second region is the said. Estimate whether it is a connection point and
   In the above estimation
   It is estimated whether it is the connection point by associating the shape of the second region estimated from the first region model that models the shape of the first region with the point cloud of the second region. Point cloud analysis method.
9. On the computer
   The first region included in the cable, the first region, and the second region included in the cable are determined by the estimation unit of the point cloud analyzer that estimates the connection point of the cable connected to the outdoor structure. By associating with the region of, it is estimated whether the boundary between the first region and the second region is the connection point.
   In the above estimation
   It is estimated whether it is the connection point by associating the shape of the second region estimated from the first region model that models the shape of the first region with the point cloud of the second region. A program to execute the process as if it were.

Images