JP2022099068A - Buried pipe determination apparatus, exploration apparatus, and buried pipe determination method - Google Patents

Abstract

To provide technique capable of reducing an oversight rate of buried pipes when automatically extracting candidates for the buried pipes while suppressing the increase in teacher data.SOLUTION: A buried pipe determination apparatus comprises: a candidate area vector derivation unit S2 for deriving a feature quantity, which is obtained by quantifying a plurality of predetermined features, for determining that a quadratic curve candidate in a candidate area is a buried pipe and deriving a candidate area vector having each of the plurality of feature quantities as a component for each of the plurality of candidate areas derived by a candidate area derivation unit S3; and a buried pipe candidate extraction unit S1 for classifying the plurality of derived candidate area vectors based on a classification rule group for classifying a buried pipe candidate vector that is a candidate for the buried pipe and a non-buried pipe candidate vector that is not the candidate for the buried pipe to extract the candidate area that is the candidate for the buried pipe.SELECTED DRAWING: Figure 2

Description

The present invention processes the reflected wave of the exploration electromagnetic wave radiated toward the ground when scanned along the scanning line set in the search range including the buried pipe buried in the ground, and the scanning line. A buried pipe determination device, an exploration device, and a buried pipe that can acquire a cross-sectional data image showing the burial status of a buried pipe in a vertical cross-sectional view including and can determine the presence or absence of a buried pipe in the ground based on the cross-sectional data image. Regarding the judgment method.

Conventionally, as a buried pipe determination device, when scanned along a scanning line set in a search range including a buried pipe buried in the ground, the reflected wave of the exploration electromagnetic wave radiated toward the ground is emitted. Processing is performed, a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view including a scanning line is acquired, a quadratic curve image included in the cross-sectional data image is extracted, and the spread of the quadratic curve image is set to 1. It is known to execute a migration process that converges to a point and extracts a buried pipe (see Patent Document 1).
On the other hand, as another buried pipe determination device, the reflected wave of the exploration electromagnetic wave radiated toward the ground when scanned along the scanning line set in the search range including the buried pipe buried in the ground. Is processed, a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view including a scanning line is acquired, the cross-sectional data image is machine-learned, the characteristics of the buried pipe are extracted by AI, and the cross-sectional data image is used. Research is also being conducted to extract buried pipes.

Japanese Unexamined Patent Publication No. 10-141907

In the buried pipe determination device that executes the above migration process and extracts the buried pipe from the cross-sectional data image, when the quadratic curve image contains a quadratic curve with a lot of noise, that is, a strong signal such as horizontal stripes is included. If so, there is a problem that the horizontal stripe is mistakenly extracted as a buried pipe.
In addition, since the degree of convergence of the quadratic curve is determined by the signal strength, there is also a problem that it is greatly affected by a strong signal from the boundary of the stratum.
On the other hand, the technique of machine learning the cross-sectional data image to extract the buried pipe requires a large number of cross-sectional data images as teacher data for learning, and it is difficult to detect the unlearned buried pipe. was there.
In addition, in the cross-sectional data image, it is necessary to consider the soil quality, water content ratio, stone mixing condition, pavement type, presence / absence of reinforcing bars in the pavement, influence of proximity pipe, influence of excavation trace, and diameter / material of buried pipe.・ It is necessary to consider the influence of the burial depth, and in the technique of machine learning the cross-sectional data image to extract the buried pipe, the hit rate tends to be low when it is judged, that is, the overlook rate of the buried pipe tends to be high. There was room for improvement.

The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to reduce the oversight rate of buried pipes when automatically extracting candidates for buried pipes while suppressing an increase in teacher data. The point is to provide the technology to obtain.

The buried pipe determination device for achieving the above purpose is
When scanned along a scanning line set in a search range including a buried pipe buried in the ground, the reflected wave of the exploration electromagnetic wave radiated toward the ground is processed and the scanning line is included. A buried pipe determination device capable of acquiring a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view and determining the presence or absence of the buried pipe in the ground based on the cross-sectional data image. The composition is
A candidate region derivation unit for deriving a candidate region including one quadratic curve candidate as a possible signal of the buried pipe from the cross-sectional data image, and a candidate region derivation unit.
With respect to each of the plurality of candidate regions derived by the candidate region derivation unit, a plurality of predetermined features for determining that the quadratic curve candidate in the candidate region is the buried pipe are quantified. A candidate region vector derivation unit for deriving the created feature quantity and deriving a candidate region vector having each of the plurality of the feature quantities as a component.
The plurality of derived candidate region vectors are classified based on a classification rule group for classifying into a buried pipe candidate vector that is a candidate for the buried pipe and a non-buried pipe candidate vector that is not a candidate for the buried pipe, and the buried pipe is used. It is provided with a buried pipe candidate extraction unit for extracting the candidate region as a pipe candidate.

The method for determining the buried pipe to achieve the above purpose is
When scanned along a scanning line set in a search range including a buried pipe buried in the ground, the reflected wave of the exploration electromagnetic wave radiated toward the ground is processed and the scanning line is included. It is a buried pipe determination method capable of acquiring a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view and determining the presence or absence of the buried pipe in the ground based on the cross-sectional data image. The composition is
A candidate region derivation step for deriving a candidate region including one quadratic curve candidate as a possible signal of the buried pipe from the cross-sectional data image, and
With respect to each of the plurality of candidate regions derived in the candidate region derivation step, a plurality of feature quantities for determining that the quadratic curve candidate in the candidate region is the buried pipe are derived, and a plurality of features are derived. A candidate region vector derivation step for deriving a candidate region vector having each of the feature quantities as a component, and a process for deriving the candidate region vector.
The plurality of derived candidate region vectors are classified based on a classification rule group for classifying into a buried pipe candidate vector that is a candidate for the buried pipe and a non-buried pipe candidate vector that is not a candidate for the buried pipe, and the buried pipe is used. The point is to execute the buried pipe candidate extraction step of extracting the candidate area as a pipe candidate.

According to the above feature configuration, first, the candidate region derivation unit derives a candidate region including one quadratic curve candidate as a signal that may be a buried pipe from the cross-sectional data image. By deriving a candidate area including the next curve candidate and executing the subsequent buried pipe determination process, the determination logic in the buried pipe determination process can be determined as compared with the case where the buried tube determination process is executed using the entire cross-sectional data image. It is possible to sufficiently reduce the amount of arithmetic processing and the arithmetic processing time until a result is obtained by simplification.

By the way, in the conventional machine learning technique for extracting buried pipes, the cross-sectional data image is directly machine-learned, but according to the above feature configuration, it is determined that the quadratic curve candidate in the candidate region is the buried pipe. In order to determine the features to be used in advance and calculate the value of the feature amount related to the feature in the candidate area, the required amount of teacher data is determined as compared with the case of directly machine learning the cross-sectional data image. Can be reduced.
Here, for example, in the candidate region of the cross-sectional data image as shown above in FIG. 7, the brightness is along the depth direction (arrow Z direction) along the straight line P passing substantially the center in the exploration direction (arrow X direction). Consider a graph (bottom of FIG. 7) plotting values (intensities).
In graphs obtained from the candidate regions, generally, as shown at the bottom of FIG. 7, the maximum amplitude (L1 at the bottom of FIG. 7) and the frequency of the wave at which the maximum amplitude is obtained (FIG. 7). The value calculated from L2 below) and can be obtained.
The maximum amplitude for each of the plurality of candidate regions corresponding to the buried pipe candidates, the plurality of candidate regions corresponding to the possible buried pipes, and the plurality of candidate regions corresponding to the non-buried pipe candidates. A plot of the distribution is shown in FIG. 8, and a plot of the frequency distribution is shown in FIG.
As shown in FIGS. 8 and 9, in both the buried pipe candidate and the non-buried pipe candidate, the maximum amplitude value and the frequency distribution are often overlapped, and these feature quantities alone are buried. It can be seen that it is difficult to separate the pipe candidate and the non-buried pipe candidate, that is, to extract the buried pipe candidate.

Therefore, according to the above feature configuration, the candidate region vector derivation unit determines that the quadratic curve candidate in the candidate region is an embedded pipe for each of the plurality of candidate regions derived by the candidate region derivation unit. Derived feature quantities related to a plurality of predetermined features for the purpose, derived a candidate region vector having each of the plurality of feature quantities as a component, and the buried pipe candidate extraction unit derived the plurality of candidate region vectors. A candidate area that is a candidate for a buried pipe is extracted by classifying based on a classification rule group that classifies a candidate buried pipe candidate vector that is a candidate for a buried pipe and a non-buried pipe candidate vector that is not a candidate for a buried pipe.
That is, the candidate region vector includes feature quantities related to a plurality of features as components, and the candidate region vector corresponds to a buried pipe candidate or a non-buried pipe candidate depending on the combination of these values. Since it is classified, it is possible to effectively classify both, which were difficult to classify by a single feature.
Further, only the classification rule group for classifying the candidate for buried pipe and the candidate for non-buried pipe, that is, a plurality of derivations of the threshold vector having a threshold for each of the components of the candidate region vector are derived by using machine learning. Compared to the case of machine learning the cross-sectional data image, the number of teacher data can be significantly reduced.
From the above, it is possible to realize a buried pipe determination device that can reduce the oversight rate of buried pipes when automatically extracting candidates for buried pipes while suppressing an increase in teacher data.

Further features of the buried pipe determination device
The candidate area derivation unit is
A cutting step of calculating a change in brightness along a predetermined straight line on the cross-sectional data image and cutting out a rectangular area circumscribing the quadratic curve candidate in a form of extracting only a region having a large change in brightness, and the cutting. A value based on the shape of the quadratic curve candidate cut out in the output step, a value based on the brightness value of the quadratic curve candidate, or a value based on the brightness value around the rectangular region is included in the quadratic curve candidate. The first candidate region derivation process for executing the exclusion step of excluding those that do not exist and deriving as a candidate region including the rectangular region corresponding to the remaining quadratic curve candidate, and
Matching that derives the rectangular region as a candidate region when the similarity between the cutting step and the quadratic curve candidate included in the rectangular region and the reflected wave waveform corresponding to the known buried pipe is equal to or higher than a certain level. The point is that at least one of the second candidate region derivation process for executing the process is executed.

As in the above feature configuration, the candidate region derivation unit calculates the change in brightness along a predetermined straight line on the cross-sectional data image, and extracts only the region with a large change in brightness. Using the cutting step of cutting out the region, the value based on the shape of the quadratic curve candidate cut out in the cutting step, the value based on the brightness value of the quadratic curve candidate, or the brightness value around the rectangular region. A first candidate area derivation process, a cutting process, and an exclusion step of excluding those not included in the quadratic curve candidates and deriving them as a candidate area including a rectangular area corresponding to the remaining quadratic curve candidates. Derivation of the second candidate region for executing a matching step of deriving the rectangular region as a candidate region when the similarity between the quadratic curve candidate included in the rectangular region and the reflected wave waveform corresponding to the known buried pipe is equal to or higher than a certain level. By executing at least one of the processing, it is possible to exclude the quadratic curve candidate that is not clearly caused by the buried pipe as the candidate region, and the subsequent derivation process of the buried pipe candidate by the buried pipe candidate derivation unit is performed. Efficiency can be improved.

By the way, the inventors have found the following features as effective features for classifying buried pipe candidates and non-buried pipe candidates by adopting them as one of the components of the candidate region vector.

For example, the candidate region vector derivation unit uses the position information derived based on the position of the candidate region derived by the candidate region derivation unit in the cross-sectional data image as one of the features, and uses the feature as one of the features. It is preferable to derive the quantified feature amount as one of the constituent elements of the candidate region vector.

Further, for example, the candidate region vector derivation unit includes the width and height of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit, values related thereto, or the quadratic included in the candidate region. At least one of the quadratic coefficient and the linear coefficient of the curve candidate and the ratio of the quadratic coefficient and the linear coefficient is one of the features, and the feature amount obtained by quantifying the feature is one of the constituent elements of the candidate region vector. It is preferable to derive the coefficient.

Further, for example, the candidate region vector derivation unit derives a brightness value along the width direction at a predetermined position in the height direction of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit in the width direction. At least one of the amplitude, wavelength, and number of peaks in the waveform when the brightness value is plotted at each of the positions in is one of the features, and the feature amount obtained by quantifying the feature is the configuration of the candidate region vector. It is preferable to derive it as one of the elements.

Further, for example, the candidate region vector derivation unit features the luminance values of a plurality of pixels in a standardized region in which the candidate region derived by the candidate region derivation unit is standardized to a predetermined size. It is preferable to derive the feature quantity obtained by quantifying the feature as one of the constituent elements of the candidate region vector.

Further, for example, when the candidate region derived by the candidate region derivation unit is used as the target candidate region, the candidate region vector derivation unit has a predetermined range around the target candidate region in the cross-sectional data image. When the number of the candidate regions existing in the area or the candidate region derived by the candidate region derivation unit is used as the target candidate region, a predetermined range around the target candidate region in the cross-sectional data image. One of the values based on the positional relationship between the candidate area and the target candidate area existing in the area is set as one of the features, and the feature amount obtained by quantifying the feature is the component of the candidate area vector. It is preferable to derive it as one.

Further features of the buried pipe determination device
With respect to the candidate region vector, a threshold vector generation process for generating a threshold vector having random components, a plurality of buried pipe vectors as teacher data which are the candidate region vectors known to be the buried pipe, and the buried pipe. When a parent group including a plurality of non-buried tube vectors as the teacher data, which is the candidate region vector known not to be a tube, is classified into two by the threshold vector, the candidate region included in one of the classifications. When the ratio of the buried pipe vector in the vector is greater than or equal to the determination ratio higher than the ratio of the buried pipe vector in the parent group, the classification rule derivation process of adding the threshold vector to the classification rule group is executed. Equipped with a classification rule derivation unit
The buried pipe candidate extraction unit classifies the buried pipe candidate vector and the non-buried pipe candidate vector based on the plurality of classification rule groups derived by the classification rule derivation unit, and classifies the buried pipe candidate vector. The point is to extract the candidate area.

According to the above feature configuration, the classification rule derivation unit executes a vector generation process for randomly generating a threshold vector, and also has a plurality of buried pipe candidate vectors and buried pipes, which are candidate region vectors known to be buried pipes. When a parent group containing a plurality of non-buried pipe candidate vectors, which are known candidate region vectors, are classified into two by a threshold vector, the buried pipe vector among the buried pipe candidate vectors included in one of the classifications. When the ratio of is greater than or equal to the judgment ratio of the buried pipe vector in the parent group, the classification rule derivation process of adding the threshold vector to the classification rule group is executed. Therefore, in machine learning, the threshold vector is randomly selected. Since machine learning is used to determine whether or not the generated threshold vector predominantly classifies the buried pipe vector and the non-buried pipe vector as well as being generated, the machine is based on the cross-sectional data image as teacher data. Compared with the case of performing learning and extracting buried pipe candidates, the number of required teacher data (buried pipe vector and non-buried pipe vector) can be significantly reduced.

Further, the exploration device provided with the buried pipe determination device described so far also works effectively as an exploration device that exerts the action and effect described so far.

It is a schematic block diagram of the exploration apparatus which concerns on embodiment of this invention. It is a block diagram of the control device which concerns on embodiment of this invention. It is a figure for demonstrating the processing by a candidate area derivation part. It is a figure for demonstrating the processing by a candidate area derivation part. This is an example of a diagram in which a buried pipe candidate and a non-buried pipe candidate are included in the area derived by the candidate area derivation unit. It is a figure for demonstrating the flow which the buried pipe candidate extraction part extracts a buried pipe candidate based on a classification rule. It is a figure which shows the amplitude and frequency of the reflected wave which is a feature quantity which constitutes the vector of the candidate area of the exploration data. It is a graph which shows the existence probability (density) for every amplitude in the candidate for a buried pipe, the candidate for a buried pipe, and the candidate for a non-buried pipe. It is a graph which shows the existence probability (density) for each frequency of a candidate for a buried pipe, a candidate for a buried pipe, and a candidate for a non-buried pipe.

The buried pipe determination device S, the exploration device 3 provided with the buried pipe determination device S, and the buried pipe determination method according to the embodiment are particularly used when automatically extracting candidates for the buried pipe K while suppressing an increase in teacher data. The present invention relates to a technique capable of reducing the oversight rate of the buried pipe K.

[Exploration device]
As shown in FIGS. 1 and 2, the exploration device 3 according to the embodiment of the present invention is set on the scanning line SL set in the search range SA including the buried pipe K buried in the ground by being pushed by an operator. It is scanned along. When the exploration device 3 is scanned along the scanning line SL, it emits an exploration electromagnetic wave toward the ground, processes the reflected wave of the exploration electromagnetic wave, and performs a vertical cross-sectional view including the scanning line SL. A buried pipe determination device S for acquiring a cross-sectional data image showing the burial status of the buried pipe K and determining the presence or absence of the buried pipe in the ground based on the cross-sectional data image is provided. Further, the buried pipe determination device S is configured so that the cross-sectional data image is displayed on the display unit S6 installed on the upper part of the exploration device 3.

More specifically, the exploration device 3 is a hand-push type or self-propelled radar exploration device, and travels along a scanning line SL set in a predetermined search range SA to be searched for the buried pipe K. While traveling, the exploration device 3 radiates an electromagnetic wave for exploration into the ground from the antenna A. If a buried object such as a buried pipe K exists in the propagation path of the radiated electromagnetic wave, it is reflected there. The reflected wave that is reflected and returned is processed by the buried pipe determination device S, and a cross-sectional data image for evaluating the existence of the target buried object (buried pipe K) is output.

The cross-section data image is appropriately subjected to image processing by the buried pipe determination device S, and is visualized on the display unit S6 as a cross-section data image as shown in FIG. 1 (b). The operator can grasp the underground position of the buried pipe K to be worked from the cross-sectional data image displayed on the display unit S6.

The search range SA is generally a specific section such as a sidewalk or a road in a jurisdiction area, and a scanning line SL is set in a predetermined pattern within the search range SA. At that time, the reference marker M defining the scanning line SL is assigned to the ground surface of the search range SA as an index.

The reference marker M that defines the scanning line SL as used herein is, for example, a character or symbol indicating the start point, intermediate point, end point, etc. of the scanning line SL, and may be drawn directly on the ground surface with chalk or paint. , A marker such as a traffic cone may be placed on the ground. Alternatively, the reference marker M that defines the scanning line SL can be created by a method of drawing a line indicating the scanning line SL or a method of placing a rope on the scanning line SL. That is, the actual scanning position and the exploration data at the scanning position are related to each other by the position of the reference marker M on the ground surface.

In the example of FIG. 1, the reference marker M (M1, M2, M3) is a chalk on the ground consisting of a black circle at each starting point of the three scanning lines SL (SL1, SL2, SL3) and an arrow indicating the operation direction. It is a drawn index.

In the present embodiment, the search device 3 scans a plurality of scan lines SL parallel to each other and having the same length in the search range SA. More specifically, at least three scanning lines SL (SL1 to SL3) are scanned, and a cross-sectional data image is acquired for each of the scanning lines SL1 to SL3.

The interval L of the scanning lines SL is determined based on the state of the buried pipe K existing under the search range SA. Specifically, for example, they are provided at intervals of 50 cm. If the buried state of the buried pipe K is known in advance (for example, when it is known that the buried pipe K exists in a straight line over a long distance), it may be provided at a wider interval.

As shown in FIG. 2, the exploration device 3 is roughly divided into an antenna A, a buried pipe determination device S, and a display unit S6. The buried pipe determination device S signals the electromagnetic wave received by the antenna A, and the display unit S6 displays an image or the like that visualizes the cross-sectional data image signal-processed by the buried pipe determination device S to the operator.

The antenna A of the exploration device 3 is preferably composed of a plurality of antenna elements. The exploration device 3 has a high-frequency power supply and a transmitter (neither shown) for radiating a pulsed electromagnetic wave in the microwave region toward the ground through the antenna A at a predetermined repeating frequency, and the probe 3 in the ground through the antenna A. It is provided with a receiving unit (not shown) for receiving the reflected wave reflected from the antenna.

By the way, the exploration device 3 is moved from the reference marker M, which is an index drawn on the ground with chalk, as a starting point, or along a scanning line SL set from a predetermined position as a starting point. The exploration device 3 is equipped with a position detection sensor unit (not shown) that detects the moving distance or the moving point. The position data detected by the position detection sensor unit is sent to the buried pipe determination device S, combined with the reflected wave reflected from the ground, and used to create a cross-sectional data image. The position detection sensor unit can be easily constructed by a rotary encoder connected to a traveling wheel, but may be constructed by a GPS or a gyro.

The display unit S6 comprises a flat panel display that displays a cross-sectional data image in a form that can be visually confirmed by an operator. As shown in FIG. 1 (b), the flat panel display is provided on the upper surface of the exploration device 3 in an inclined posture so as to be clearly visible to the operator who pushes the exploration device 3.

The exploration device 3 appropriately amplifies the received signal received by the receiving unit, and performs signal processing by the embedded pipe determination device S so that the amplified received signal can be displayed on the display unit S6.
By the way, the buried pipe determination device S has the following buried pipe determination method in order to reduce the oversight rate of the buried pipe K when automatically extracting candidates for the buried pipe K while suppressing the increase in teacher data. Has a configuration to execute.
That is, the buried pipe determination device S is a candidate area deriving unit S3 that executes a candidate area deriving process for deriving a candidate area including one quadratic curve candidate as a possible signal of the buried pipe K from the cross-sectional data image. And, with respect to each of the plurality of candidate regions derived by the candidate region derivation unit S3, a plurality of predetermined features for determining that the quadratic curve candidate in the candidate region is the buried pipe K are quantified. The candidate area vector derivation unit S2 for executing the candidate area vector derivation process for deriving the candidate area vector having each of the plurality of feature amounts as a component, and the plurality of derived candidate area vectors are embedded. Buried to extract candidate areas that are candidates for buried pipe K by classifying them based on a classification rule group that classifies them into a buried pipe candidate vector that is a candidate for pipe K and a non-buried pipe candidate vector that is not a candidate for buried pipe K. It is configured to include a buried pipe candidate extraction unit S1 that executes a pipe candidate extraction step.

The candidate region derivation unit S3 calculates the change in brightness along a predetermined straight line on the cross-sectional data image as the first candidate region derivation process, and extracts only the region having a large change in brightness. A cutting process for cutting out a rectangular area circumscribing the candidate, and a value based on the shape of the quadratic curve candidate cut out in the cutting process, or a value based on the brightness value of the quadratic curve candidate or the brightness value around the rectangular area. Is used to exclude those not included in the quadratic curve candidates, and execute the exclusion step of deriving as a candidate region including the rectangular region corresponding to the remaining quadratic curve candidates.
Here, the rectangular area and the candidate area may include not only one quadratic curve candidate but also another quadratic curve candidate.
In the cutting step, a high-luminance (large signal amplitude) region shown in FIG. 3B is detected from the reflected wave image shown in FIG. 3A, and the detected region is shown in FIG. 3C. The process of cutting out a shape close to the quadratic curve shown is executed. To add an explanation, in FIG. 3B, the difference between the differential filter and the smoothing filter is used. At that time, a convolution matrix having a predetermined size range (5 to 25 pixels for both rows and columns) is used. In FIG. 3 (c), a probabilistically plausible point set is extracted as a quadratic curve candidate from the set of points on the line structure obtained in FIG. 3 (b). Specifically, when fitting with a quadratic curve, a point set having an average position error of 6 pixels or less (more preferably 3 pixels or less) is extracted.
Further, in the above exclusion step, as values based on the shape of the quadratic curve candidate, the width and height of the circumscribing rectangle of the quadratic curve candidate (width and height of the rectangle shown by the broken line in FIG. 4) and the quadratic curve candidate. At least one of the coefficients is used to compare these values with the threshold, and the non-target ones are excluded from the quadratic curve candidates (candidate regions).
As an example, the width of the quadratic curve (width of the circumscribing rectangle of the quadratic curve candidate) excludes those exceeding 200 pixels, and the range of the quadratic coefficient is less than 0.002 and more than 0.02. Exclude.

Explaining with reference to FIG. 4, (a) is excluded because the quadratic coefficient is larger than 0.02, and (b) is excluded because the quadratic coefficient is less than 0.002 and the width exceeds 200 pixels. Only (d) is left as a quadratic curve candidate (candidate area).

Next, the candidate region vector derivation unit S2 determines that the quadratic curve candidate in the candidate region is an embedded pipe with respect to each of the plurality of quadratic curve candidates (candidate regions) derived by the candidate region derivation unit S3. A feature quantity that quantifies a plurality of predetermined features for determination is derived, and a candidate region vector having each of the plurality of feature quantities as a component is derived.
The feature quantity as a component of the candidate region vector derived by the candidate region vector derivation unit S2 preferably includes at least two or more of the following.

For example, the candidate area vector derivation unit S2 uses the position information derived based on the position in the cross-sectional data image of the candidate area derived by the candidate area derivation unit S3 as one of the features, and quantifies the feature. The feature quantity is derived as one of the components of the candidate region vector.
Specifically, with respect to the X coordinate Z coordinate of the center position of the candidate area in the section data image and the center position of the candidate area, the arrangement direction of a plurality of section data images (the arrangement direction of the reference marker M in FIG. 1A). ), Etc. are mentioned.
Since the feature cannot be extracted only from the candidate region, when the feature is used as a component of the candidate region vector, the cross-sectional data image from which the candidate region is extracted is also stored in the storage unit S5 together with the candidate region. I will keep it.
Further, the position information includes not only the area in the circumscribed rectangle of the quadratic curve candidate but also the position in the candidate area including the surrounding area.
Further, the position information includes the position where the quadratic curve candidate is migrated.

For example, the candidate region vector derivation unit S2 is the width and height of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit S3 and the values related to them, or the quadratic curve candidate included in the candidate region. At least one of the coefficient, the linear coefficient, and the ratio of the quadratic coefficient to the linear coefficient is set as one of the features, and the feature quantity obtained by quantifying the feature is derived as one of the components of the candidate region vector.

Further, for example, the candidate region vector derivation unit S2 derives the luminance value along the width direction at a predetermined position in the height direction of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit S3 in the width direction. One of the features is at least one of the amplitude, wavelength, and number of peaks in the waveform when the luminance value is plotted at each of the positions of, and the feature quantity that quantifies the feature is one of the components of the candidate region vector. Derived as.
Specifically, in the candidate region of the cross-sectional data image as shown above in FIG. 7, the width direction (arrow X direction) along a straight line passing through a predetermined position (for example, the center position) in the depth direction (arrow Z direction). ), At least one of the amplitude, frequency, and peak value of the signal as the brightness value may be one of the features in the graph in which the brightness value (intensity) is plotted.

Further, for example, the candidate area vector derivation unit S2 is characterized by having the luminance values of a plurality of pixels in the standardized area in which the candidate area derived by the candidate area derivation unit S3 is standardized to a predetermined size. Then, the feature quantity obtained by quantifying the feature is derived as one of the constituent elements of the candidate region vector.
Here, the standardized area is a 2 × 2 window frame, and while scanning on a rectangle, the image is replaced with 2 × 2 by 1 pixel of the maximum brightness in the window frame to make the image 1/2 ×. This is an area obtained by repeating the process of converting to a size of 1/2 until the size is finally 4 × 4.

Further, for example, the candidate area vector derivation unit S2 includes the number of quadratic curve candidates (candidate areas) existing in a predetermined range around the candidate area and the cross-sectional data image when the candidate area is set as the target candidate area. The positional relationship between the quadratic curve candidate existing within a predetermined range around the target candidate area and the target candidate area may be one of the features.

Now, a classification rule derivation unit S4 is provided in order to classify whether the candidate region vector thus obtained corresponds to the buried pipe K or the non-buried pipe.
The classification rule derivation unit S4 has a threshold vector generation process for generating threshold vectors _Z1 to _Zn having random components with respect to the candidate region vector, and a teacher which is a candidate region vector known to be a buried pipe K. Multiple buried pipe vectors X ₁ to X _n (integer of n = 2 or more) as data and non-buried pipe vectors Y ₁ to Y _n (n) as teacher data which are candidate region vectors known not to be buried pipe K. When a parent group containing a plurality of (= integers of 2 or more) is classified into two by the threshold vectors Z ₁ to _Zn , the candidate region vectors X ₁ to X _n and Y ₁ to Y _n included in one of the classifications. When the ratio of the buried pipe vectors X ₁ to X _n is greater than or equal to the judgment ratio higher than the ratio of the buried pipe vectors X ₁ to X _n in the parent group, the threshold vectors Z ₁ to Z _n are set as the classification rule group. Executes the classification rule derivation process to be added.
The determination ratio is a ratio arbitrarily determined when deriving a classification rule according to soil conditions and the like.

Here, first, the buried pipe vector and the non-buried pipe vector will be described with reference to FIG. As shown in FIG. 5, for each of the candidate region known to be the buried pipe K and the candidate region known not to be the buried pipe K, the candidate region vectors X ₁ to X _n and Y ₁ to Y _n . Is derived.
The candidate region vectors X ₁ to X _n and Y ₁ to Y _n have numerical values of the feature quantities corresponding to the above-mentioned plurality of features as constituent elements, and the candidate region vectors X ₁ will be described as an example. Then, when the quadratic coefficient of the quadratic curve candidate (X ₁₁ ), the linear coefficient of the quadratic curve candidate (X ₁₂ ), and one of the brightness values of a plurality of pixels in the standardized region (X ₁₃ ) are featured, The candidate region vector is as shown in [Equation 1] below, with the numerical values of them as structural elements.
X ₁ = (X ₁₁ , X ₁₂ , X ₁₃ ) ... [Equation 1]

The candidate region vectors X ₁ to X _n and Y ₁ to Y _n derived from the candidate regions known to be the buried pipe K are defined as the buried pipe vectors X ₁ to X _n , and are not the buried pipe K. Let the non-buried pipe vectors Y ₁ to Y _n be derived from the known candidate regions.

The classification rule derivation unit S4 further generates threshold vectors Z ₁ to Zn (integer of _n = 1 or more) having random components with respect to the above-mentioned candidate region vector. The threshold vectors Z ₁ to _Zn have the number of components corresponding to the above-mentioned candidate region vectors X ₁ to X _n and Y ₁ to Y _n , and the numerical values corresponding to the components are randomly and automatically calculated. It was generated.

As shown in FIG. 6, the classification rule derivation unit S4 randomly generates a parent group including the buried pipe vectors X ₁ to X _n and the non-buried pipe vectors Y ₁ to Y _n in order to automatically generate the classification rule group. It is classified into two by one of the generated threshold _vectors Z ₁ to Zn (rule 1 in FIG. 6). Here, the classification method is arbitrarily set by the classification rule derivation unit S4. For example, when the buried pipe vector X ₁ is classified by the threshold vector Z ₁ , (X ₁ <Z ₁ , X ₂ <Z ₂ ). , X ₃ > Z ₃ ) are classified into those that satisfy all the conditions and those that do not satisfy at least one or more.
After that, when the parent group is classified into two by the threshold vectors Z ₁ to _Zn , the classification rule derivation unit S4 classifies the candidate region vectors X ₁ to X _n and Y ₁ to Y _n included in one of the classifications. When the ratio of the buried pipe vectors X ₁ to X _n (α in FIG. 6) is greater than the ratio of the buried pipe vectors X ₁ to X _n in the parent group (δ in FIG. 6), the determination ratio θ or more. The classification rule derivation process of adding one of the threshold _vectors Z ₁ to Zn (rule 1 in FIG. 6) to the classification rule group is executed.

Here, as shown in FIG. 6, the parent group for allocating the threshold vectors Z ₁ to _Zn may include all the teacher data, and the other threshold vectors Z ₁ to _Zn may be used. , It may be a group after being classified. Therefore, in FIG. 6, regarding the threshold vectors Z ₁ to Zn corresponding to rule 2, the buried pipe vectors X ₁ among the candidate region vectors X ₁ to X _n and Y ₁ to Y _n included in _one of the classifications. When the ratio of ~ _Xn (β or γ in FIG. 6) is larger than the judgment ratio θ, the rule 2 is added to the classification rule group, and when the ratio is less than the judgment ratio θ, it is added to the classification rule group. It will not be possible.

The buried pipe candidate extraction unit S1 is buried based on the plurality of classification rule groups derived by the classification rule derivation unit S4, that is, by the threshold _vectors Z ₁ to Zn as a plurality of rules included in the classification rule group. The pipe candidate vector and the non-buried pipe candidate vector are classified, and the candidate area that is a candidate for the buried pipe is extracted.

[Another Embodiment]
(1) In the above embodiment, the buried pipe determination device S is provided integrally with the exploration device 3, but a single unit of the buried pipe determination device S is acquired from the external exploration device 3. A configuration may be adopted in which the cross-sectional data image is acquired via a communication means or the like, and the above-mentioned candidate region derivation step, the candidate region vector degree derivation step, and the buried pipe candidate extraction step are executed.
That is, the simple substance of the buried pipe determination device S is also included in the scope of rights of the present application.

(2) In the above embodiment, the candidate area derivation unit S3 has shown an example of executing the first candidate area derivation process, but other processes may be executed.
For example, in the candidate region derivation unit S3, the similarity between the cutting process executed in the above embodiment and the quadratic curve candidate included in the rectangular region and the reflected wave waveform corresponding to the known buried pipe K is at least a certain level. In some cases, a second candidate area derivation process for executing a matching process for deriving a rectangular area as a candidate area and a configuration for executing the process may be adopted.
Further, the candidate area derivation unit S3 may derive another candidate area in a form of executing both the first candidate area derivation process and the second candidate area derivation process.

(3) In the exclusion step in the first candidate area derivation process of the candidate area derivation unit S3, the area including the inside of the circumscribing rectangle of the cut out quadratic curve or its surroundings is used as a value based on the brightness value of the quadratic curve candidate. The brightness value, more specifically, for example, in the candidate region of the cross-sectional data image as shown above in FIG. 7, the depth direction (arrow Z direction) along the straight line P passing substantially the center in the exploration direction (arrow X direction). ), In a graph (bottom of FIG. 7) in which the brightness value (intensity) is plotted, the amplitude, frequency, average value of the signal, etc. of the signal formed by the brightness value are compared with the threshold value, and a predetermined one is obtained. A configuration that excludes from the quadratic curve candidates (candidate areas) may be adopted.

(4) In the above embodiment, a configuration example in which the classification rule derivation unit S4 is provided is shown, but the classification rule derivation unit S4 is provided in another arithmetic unit different from the buried pipe determination device S, and the classification rule derivation unit is provided. A configuration may be adopted in which the classification rule group is derived in S4, and only the derived classification rule group is stored in the storage unit S5 of the buried pipe determination device S and used as necessary.

It should be noted that the configuration disclosed in the above embodiment (including another embodiment, the same shall apply hereinafter) can be applied in combination with the configuration disclosed in other embodiments as long as there is no contradiction. The embodiments disclosed in the present specification are examples, and the embodiments of the present invention are not limited thereto, and can be appropriately modified without departing from the object of the present invention.

The buried pipe determination device, the exploration device, and the buried pipe determination method of the present invention are techniques that can reduce the oversight rate of buried pipes when automatically extracting candidates for buried pipes while suppressing an increase in teacher data. Can be effectively used as a buried pipe determination device, an exploration device, and a buried pipe determination method.

3: Exploration device K: Buried pipe S: Buried pipe determination device S1: Buried pipe candidate extraction unit S2: Candidate area vector derivation unit S3: Candidate area derivation unit S4: Classification rule derivation unit X: Candidate area vector Y: Non-buried pipe Vector Z: Threshold vector θ: Judgment ratio

Claims (10)

When scanned along a scanning line set in a search range including a buried pipe buried in the ground, the reflected wave of the exploration electromagnetic wave radiated toward the ground is processed and the scanning line is included. A buried pipe determination device capable of acquiring a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view and determining the presence or absence of the buried pipe in the ground based on the cross-sectional data image.
A candidate region derivation unit for deriving a candidate region including one quadratic curve candidate as a possible signal of the buried pipe from the cross-sectional data image, and a candidate region derivation unit.
With respect to each of the plurality of candidate regions derived by the candidate region derivation unit, a plurality of predetermined features for determining that the quadratic curve candidate in the candidate region is the buried pipe are quantified. A candidate region vector derivation unit for deriving the created feature quantity and deriving a candidate region vector having each of the plurality of the feature quantities as a component.
The plurality of derived candidate region vectors are classified based on a classification rule group for classifying into a buried pipe candidate vector that is a candidate for the buried pipe and a non-buried pipe candidate vector that is not a candidate for the buried pipe, and the buried pipe is used. A buried pipe determination device including a buried pipe candidate extraction unit that extracts the candidate area that is a candidate for a pipe. The candidate area derivation unit is
A cutting step of calculating a change in brightness along a predetermined straight line on the cross-sectional data image and cutting out a rectangular area circumscribing the quadratic curve candidate in a form of extracting only a region having a large change in brightness, and the cutting. A value based on the shape of the quadratic curve candidate cut out in the output step, a value based on the brightness value of the quadratic curve candidate, or a value based on the brightness value around the rectangular region is included in the quadratic curve candidate. The first candidate region derivation process for executing the exclusion step of excluding those that do not exist and deriving as a candidate region including the rectangular region corresponding to the remaining quadratic curve candidate, and
Matching that derives the rectangular region as a candidate region when the similarity between the cutting step and the quadratic curve candidate included in the rectangular region and the reflected wave waveform corresponding to the known buried pipe is equal to or higher than a certain level. The buried pipe determination device according to claim 1, wherein at least one of the second candidate region derivation process for executing the process is executed. The candidate region vector derivation unit uses the position information derived based on the position of the candidate region derived by the candidate region derivation unit in the cross-sectional data image as one of the features, and digitizes the feature. The buried pipe determination device according to claim 1 or 2, wherein the feature amount is derived as one of the constituent elements of the candidate region vector. The candidate region vector derivation unit is the width and height of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit and the values related thereto, or the quadratic curve candidate included in the candidate region. At least one of the next-order coefficient, the first-order coefficient, and the ratio of the quadratic coefficient and the first-order coefficient is set as one of the features, and the feature quantity obtained by quantifying the feature is derived as one of the constituent elements of the candidate region vector. The buried pipe determination device according to any one of claims 1 to 3. The candidate region vector derivation unit derives a luminance value along the width direction at a predetermined position in the height direction of the circumscribing rectangle of the candidate region derived by the candidate region derivation unit, and is a position in the width direction. At least one of the amplitude, wavelength, and number of peaks in the waveform when the luminance value is plotted is one of the features, and the feature amount obtained by quantifying the feature is one of the constituent elements of the candidate region vector. The buried pipe determination device according to any one of claims 1 to 4, which is derived as described above. The candidate area vector derivation unit has one of the features of each of the luminance values of a plurality of pixels in the standardized area in which the candidate area derived by the candidate area derivation unit is standardized to a predetermined size. The buried pipe determination device according to any one of claims 1 to 5, wherein a feature amount obtained by quantifying the feature is derived as one of the constituent elements of the candidate region vector. The candidate area vector derivation unit exists within a predetermined range around the target candidate area in the cross-sectional data image when the candidate area derived by the candidate area derivation unit is used as the target candidate area. When the number of the candidate regions or the candidate region derived by the candidate region derivation unit is used as the target candidate region, the candidate region exists within a predetermined range around the target candidate region in the cross-sectional data image. One of the values based on the positional relationship between the candidate area and the target candidate area is set as one of the features, and the feature amount obtained by quantifying the feature is derived as one of the constituent elements of the candidate area vector. The buried pipe determination device according to any one of claims 1 to 6. With respect to the candidate region vector, a threshold vector generation process for generating a threshold vector having random components, a plurality of buried pipe vectors as teacher data which are the candidate region vectors known to be the buried pipe, and the buried pipe. When a parent group including a plurality of non-buried tube vectors as the teacher data, which is the candidate region vector known not to be a tube, is classified into two by the threshold vector, the candidate region included in one of the classifications. When the ratio of the buried pipe vector in the vector is greater than or equal to the determination ratio higher than the ratio of the buried pipe vector in the parent group, the classification rule derivation process of adding the threshold vector to the classification rule group is executed. Equipped with a classification rule derivation unit
The buried pipe candidate extraction unit classifies the buried pipe candidate vector and the non-buried pipe candidate vector based on the plurality of classification rule groups derived by the classification rule derivation unit, and the buried pipe candidate extraction unit. The buried pipe determination device according to any one of claims 1 to 7, wherein the candidate area is extracted. An exploration device provided with the buried pipe determination device according to any one of claims 1 to 8. When scanned along a scanning line set in a search range including a buried pipe buried in the ground, the reflected wave of the exploration electromagnetic wave radiated toward the ground is processed and the scanning line is included. It is a buried pipe determination method capable of acquiring a cross-sectional data image showing the burial status of the buried pipe in a vertical cross-sectional view and determining the presence or absence of the buried pipe in the ground based on the cross-sectional data image.
A candidate region derivation step for deriving a candidate region including one quadratic curve candidate as a possible signal of the buried pipe from the cross-sectional data image, and
With respect to each of the plurality of candidate regions derived in the candidate region derivation step, a plurality of feature quantities for determining that the quadratic curve candidate in the candidate region is the buried pipe are derived, and a plurality of features are derived. A candidate region vector derivation step for deriving a candidate region vector having each of the feature quantities as a component, and a process for deriving the candidate region vector.
The plurality of derived candidate region vectors are classified based on a classification rule group for classifying into a buried pipe candidate vector that is a candidate for the buried pipe and a non-buried pipe candidate vector that is not a candidate for the buried pipe, and the buried pipe is used. A method for determining a buried pipe, which executes a buried pipe candidate extraction step of extracting the candidate area as a candidate for a pipe.

Images