JP4644816B2 - Ground penetrating radar apparatus and image signal processing method - Google Patents

Description

  The present invention relates to an underground radar apparatus and an image signal processing method for visualizing a buried object in the ground as a three-dimensional stereoscopic image.

JP 09-033194 A JP 2004-198195 A

  Traditionally, landmines represented by anti-tank landmines and antipersonnel landmines have been used in conflict areas around the world, and even today, countless landmines remain buried, It is extremely difficult.

  In the post-conflict mine removal, so-called humanitarian mine removal, life on the land that is the minefield is assumed, so all the landmines buried in the minefield must be removed. The actual situation is that these mines are difficult to detect and remove.

  In other words, the types of landmines include, for example, those that detonate in response to the magnetic field of a landmine detector, those that detonate by applying pressure several times, and those that are composed of metal or nonmetal. There is.

  As a result, in the case of landmines that detonate in response to a magnetic field, there is a possibility of detonation just by bringing a metal detector closer. In addition, in the case of a mine that detonates by applying pressure several times, there is a possibility that it will not detonate when a mine-treated vehicle passes, and it may detonate when a vehicle other than a mine-treated vehicle that passes subsequently passes. is there.

  Moreover, depending on the type of radar, it may be misrecognized in the underground, such as mine fragments that have already detonated, reflectors such as moisture in the soil, or soil that contains a large amount of metal components. There was a possibility that it would end up.

  On the other hand, as an apparatus for detecting such landmines, an apparatus using a metal detector as disclosed in Patent Document 1 or an electromagnetic wave propagated into the ground as disclosed in Patent Document 2, and from the ground There is known an apparatus that detects a reflected wave of the image with a sensor and forms an image based on the detection result.

  By the way, in the underground radar apparatus and the image signal processing method configured as described above, due to various factors such as the type of mine and the constituent material described above, in the metal detector disclosed in Patent Document 1, There are many misrecognitions, and it takes time for the removal work, and there is a problem that skill is required to specify the detailed position of the buried position (including the depth).

  In addition, in the case of the device disclosed in Patent Document 2, the underground state is represented by a horizontal sectional view and a vertical sectional view so that the three-dimensional position of the buried object can be easily understood visually. Since it is not displayed as a three-dimensional stereoscopic image, it is difficult to easily grasp the shape of the buried object, and it is difficult to specify whether it is a landmine.

  Further, when such three-dimensional imaging is possible, when a general image analysis technique is applied, there is a problem that reliability is low in order to specify the presence or absence of an embedded object.

  In addition, as one of the buried objects, in the case of a buried object that can be specified in advance and has a special shape such as a landmine, the pre-patterned three-dimensional stereoscopic image is compared with the detected three-dimensional object image pattern. Pattern matching methods that identify whether or not a land mine has been considered are also considered, but these pattern matching methods, for example, in the soil containing metal components described above, adhere to the surroundings of land mines and land mines. Even the soil containing the metal component is recognized as a buried object and patterned, and there is a possibility that the two are not matched, resulting in a problem of low reliability.

  On the other hand, a ground penetrating radar (GPR) radiates a radio wave from an antenna into the ground, and visualizes the ground by receiving a radio wave reflected from a buried object in the ground or a layer boundary surface. Device.

  The GPR data received by the ground penetrating radar is a three-dimensional data including a two-dimensional matrix in a horizontal plane in the X and Y directions with reference to the ground surface, and a Z direction (depth direction) from the ground surface toward the ground. As shown in FIG. 1, a three-dimensional stereoscopic image can be calculated based on the GPR data.

  In addition, in FIG. 1, although the presence and position of the land mine J as an underground buried object can be confirmed clearly, in many cases, clutter that becomes unnecessary reflection due to inhomogeneity of gravel and the like contained in the ground C is generated, and it is difficult to distinguish landmine J and clutter C from a simple three-dimensional image.

  By the way, since the horizontal image data for each depth as the GPR data described above surely includes the information of the reflector, it is possible to clearly determine the position of the landmine. However, when this becomes a large amount of three-dimensional GPR data, it is very difficult to find the object while eliminating the clutter C.

  Therefore, if the reflected wave from the embedded object contained in such 3D GPR data can be detected efficiently and the position of the embedded object can be detected accurately and simply, the work at the mine removal work site can be done quickly. In addition, downsizing of the apparatus can be realized.

  In order to solve the above problems, the present invention provides a ground penetrating radar apparatus and an image signal processing method capable of improving the detection accuracy of an object embedded in the ground and improving the reliability. Objective.

  In order to achieve the object, the image signal processing method of the present invention calculates numerical values of each element of the three-dimensional matrix data in the horizontal plane direction with respect to the ground surface and the depth direction from the ground surface to the ground. A three-dimensional image is created based on a three-dimensional data acquisition step to be acquired, a moving average calculation step for obtaining a moving average in a horizontal plane with reference to the ground surface, and the three-dimensional data acquisition step and the moving average calculation step. An imaging step, a peak sharpness calculating step for obtaining the peak sharpness from the cross-correlation of adjacent images, and a buried object identifying step for identifying the presence or absence of the buried object from a condition with the maximum peak sharpness. It is characterized by being.

  That is, according to the image signal processing method of the present invention, in a two-dimensional (one-dimensional position-depth) or three-dimensional (two-dimensional position-depth) data space acquired along a one-dimensional or two-dimensional scanning line. The spatial correlation coefficient in a range adjacent to a specific depth direction is calculated, and since the ground radar data has a continuous pattern in the area where the spatial correlation coefficient is large, the reflected wave from the target is captured and the target is detected. The detection rate of the target (embedded object) can be increased from the continuity of the reflected wave from the target.

  According to the underground radar apparatus and the image signal processing method of the present invention, it is possible to improve the reliability related to the presence or absence of an embedded object.

  Next, an embodiment according to the underground radar apparatus and the image signal processing method of the present invention will be described with reference to the drawings.

  The underground radar apparatus of the present invention is mounted on a work vehicle such as a GPR radar, a work vehicle such as a demining vehicle equipped with the GPR radar, and a detection signal (GPR data) from the GPR radar. And a personal computer as a calculation unit. The personal computer may receive and process a detection signal (GPR data) from the GPR radar at a remote location via a wireless communication device mounted on the work vehicle.

  Hereinafter, the processing method on the personal computer of the underground radar apparatus in the present invention will be described.

(Mineland image extraction algorithm)
[Two-dimensional imaging]
The GPR measurement data is composed of a three-dimensional matrix in the X direction, the Y direction, and the Z direction. A horizontal plane image in the Z direction (depth direction) is obtained by imaging the plane formed by the X direction and the Y direction of the matrix. Buried images such as landmines appear continuously in the depth direction, and even if they disappear once, the colors appear again. This is because the value of the reflected wave from the landmine changes from positive to negative. On the other hand, in the case of clutter, it changes minutely and irregularly.

  FIG. 2 shows the change in GPR waveform at the depth Z of the value of the XY coordinate point where the land mine exists. In FIG. 2, the portion sandwiched between two two-dot chain lines near a depth of 0.2 m is the value of the reflected wave from the landmine.

[Moving average image]
3A and 3B show examples of GPR horizontal plane images at arbitrary depths when the depth is changed. FIG. 3A shows a case of a landmine, and FIG. 3B shows a case of clutter. Show. An actual image is a color-processed image such as a thermosensor that becomes red as the power increases.

  As shown in FIG. 3 (A), the magnitude of the reflection from landmines varies continuously with depth, whereas the reflection from clutter is not as shown in FIG. 3 (B). Change to rules. By taking a moving average of the power of the waveform, the reflected wave from the landmine becomes a stable change, making it easy to distinguish from the clutter. Now, as shown in FIG. 4, the moving average of the XY plane is calculated by setting the average width of the depth change in the depth direction of the horizontal image as W and the shift width as Dz. As an example, FIG. 5 shows a part of an image obtained when W = 50 (4 cm) and Dz = 25 (2 cm). The image shown in FIG. 5C is a landmine image.

  In FIG. 5C, it can be seen that landmines are clearly distinguished from clutter by taking a moving average. Moreover, it can confirm that it has appeared continuously. Further, it can be seen from FIG. 5C that the coordinate point having the maximum value in the mine image is the mine position. From these features, if the similarity between adjacent images in the moving average image is high, it can be specified that they are landmine images. Furthermore, the coordinate point taking the maximum value is the landmine position.

[Evaluation of similarity using spatial correlation method]
A spatial cross-correlation function is used to evaluate the similarity between images. First, the GPR signal is represented by Equation 1.

この数１のそれぞれを、2次元フーリエ変換することにより、数２に示すように、フーリエスペクトルが得られる。

By performing two-dimensional Fourier transform on each of the equations 1, a Fourier spectrum is obtained as shown in equation 2.

この数２より、数３に示すクロス・パワースペクトルを求めることができる。

From this equation 2, the cross power spectrum shown in equation 3 can be obtained.

さらに、この数３を逆フーリエ変換することにより、数４に示す相互相関関数を得ることができる。

Furthermore, the cross correlation function shown in Formula 4 can be obtained by performing inverse Fourier transform on Formula 3.

  Examples of obtaining cross-correlation between adjacent moving average images are shown in FIGS. The cross-correlation image based on the clutters shown in FIGS. 6A and 6B (FIG. 6C) has a gentle peak at the center, but the mine images shown in FIGS. 7A and 7B. It can be seen that the cross-correlation image based on each other (FIG. 7C) has a sharp peak at the center. That is, the higher the similarity between two images, the sharper the peak of this cross-correlation image.

  In order to quantify such similarity, as shown in FIG. 8 and FIG. 7C, peak half-value widths δx and δy are obtained in the X and Y directions, and the sharpness of the peaks = δx · δy is evaluated. . An example in which the change of the value for each depth is graphed is shown in FIG.

  In FIG. 9, the peak sharpness has a minimum value near a depth of 0.2 m. When this peak sharpness is small, the continuity of continuous reflected waves is strong, that is, there is a high possibility that a reflector exists. The moving average image from which this correlation is obtained is the mine image, and the coordinate at which the maximum value of the image is taken is the mine position.

[Construction of mine location identification algorithm]
Summarizing the algorithm for identifying landmine positions, take the power of each element of the three-dimensional matrix data ・ Calculate the moving average of the XY plane and image each of them ・ Calculate the cross-correlation of adjacent images and peak Find the sharpness ・ Identify the mine image from the condition with the maximum peak sharpness ・ Find the point that takes the maximum value in the mine image ・ Identify the mine coordinates

[Application example]
In the algorithm shown here, the selection of the width W for averaging and the width Dz for shifting it is very important. In order to extract the characteristics of the landmine image, an appropriate average width W and shift width Dz must be set. Next, specific examples to which the invented algorithm is applied will be given.

  The GPR data handled here is measured in dry geology. The target is a TYPE72 type simulated landmine (diameter 7.6 cm, thickness 4 cm). As shown in FIGS. 10 and 11, the landmine position can be specified. At this time, the average width W = 30 (2.4 cm) and the shift width Dz = 15 (1.2 cm).

  In the above description, landmine detection is used as an example of buried object detection. However, the present invention is not limited to landmines. Also, if the GPR signal has continuity in the depth direction even in the case of a non-stationary shape such as the groundwater distribution of soil, such as a landmine, the shape is not steady, but for example, underground petroleum pollutants are targeted. It is also possible to detect abnormal (atypical) objects such as.

1 is a schematic diagram of a three-dimensional stereoscopic image according to an embodiment of the present invention. It is a graph of the change of the value of the XY coordinate point of the landmine position which shows one Embodiment of this invention. 1A and 1B show an embodiment of the present invention, in which FIG. 4A is an explanatory diagram of changes due to the depth of a horizontal plane image in the case of landmines, and FIG. 4B is an explanatory diagram of changes due to the depth of a horizontal plane image in the case of clutter. It is explanatory drawing of the concept of the moving average in the depth direction which shows one Embodiment of this invention. One Embodiment of this invention is shown, (A)-(E) is explanatory drawing of a moving average image. 1A and 1B show an embodiment of the present invention, in which FIG. 1A is an explanatory diagram of an image that is one of cross-correlations of moving average images in the case of clutter, and FIG. Explanatory drawing of an image, (C) is explanatory drawing of the cross correlation image in the case of a clutter. 1A and 1B show an embodiment of the present invention, in which FIG. 1A is an explanatory diagram of an image that is one of cross-correlations of a moving average image in the case of landmines, and FIG. Explanatory drawing of an image, (C) is explanatory drawing of the cross correlation image in the case of a landmine. It is a graph of a peak section showing an embodiment of the present invention. It is a graph of the peak sharpness-depth relationship showing an embodiment of the present invention. 1 is an explanatory diagram of landmine identification according to an embodiment of the present invention. It is a graph of the peak sharpness-depth relationship for showing landmine identification according to an embodiment of the present invention.

Claims (3)

  A GPR radar that radiates a radar wave toward the ground surface and receives a reflected wave reflected in the ground, and obtains and calculates horizontal image data for each arbitrary depth based on the reflected wave received by the GPR radar. A ground penetrating radar apparatus comprising: an arithmetic unit that obtains peak sharpness from the mutual relationship of horizontal image data.   A three-dimensional data acquisition step for acquiring numerical values of each element of the three-dimensional matrix data in a horizontal plane direction with respect to the ground surface and a depth direction from the ground surface into the ground, and a horizontal plane with reference to the ground surface A moving average calculation step for obtaining a moving average of the image, an imaging step for creating a three-dimensional image based on the three-dimensional data acquisition step and the moving average calculation step, and a peak sharpness based on a cross-correlation between adjacent images. An image signal processing method comprising: a peak sharpness calculating step for obtaining a degree; and a buried object specifying step for specifying presence / absence of a buried object from a condition with the maximum peak sharpness. A three-dimensional data acquisition step for acquiring numerical values of each element of the three-dimensional matrix data in a horizontal plane direction with respect to the ground surface and a depth direction from the ground surface into the ground, and a horizontal plane with reference to the ground surface A moving average calculation step for obtaining a moving average of the image, an imaging step for creating a three-dimensional image based on the three-dimensional data acquisition step and the moving average calculation step, and a peak sharpness based on a cross-correlation between adjacent images. The peak sharpness calculation step to determine the degree, the buried object identification step that identifies the presence or absence of the buried object from the condition where the peak sharpness is the maximum, and the buried object coordinates are identified based on the point at which the maximum value of the buried object image is taken An image signal processing method comprising: an embedded object coordinate specifying step.

Images