JP2016194515A - Imaging method of image for point group data generation, and point group data generation method using image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an imaging method of an image for accurately generating point group data indicating a three-dimensional shape of a target region using a plurality of images taken from a high place using a flight vehicle, crane, or lift, and provide a generation method of the point group data.SOLUTION: An imaging method of an image used for generating point group data indicating a three-dimensional shape of a target region from a plurality of taken images is provided. In this method, the plurality of taken images using a standard lens for generating the point group data are taken so that the plurality of taken images overlap each other by 30% or more and at least some images, of the plurality of images, include at least one of three or more reference marks set at a predetermined interval in a target region in which the plurality of taken images are taken.SELECTED DRAWING: Figure 1

Description

The present invention relates to a method for capturing an image for generating point cloud data representing a three-dimensional shape of a target region using a plurality of captured images, and a method for generating point cloud data using the captured images. In particular, the present invention relates to a method for capturing an image for generating point cloud data that can improve the accuracy of point cloud data, and a method for generating point cloud data using the image.

  Conventionally, photogrammetry that accurately extracts wide-area geographic and topographical information using photographs taken from flying objects such as aircraft has been used for topographic map creation and interpretation analysis of topography and land use. Such photogrammetry is based on the detailed relationship between the aerial photograph and the ground position taken by overlapping the vertical photograph of the ground surface by 60% to 80% along the flight course of the flying object. By measuring automatically, an accurate three-dimensional image can be created and used for measurement and topographic map creation.

  Such aerial photogrammetry is also proposed in Patent Document 1 (Japanese Patent Laid-Open No. 2004-245741). That is, in this document, there is an aerial photogrammetry method for obtaining ground coordinates of a specified point in a plurality of captured images captured from a flying object flying at a predetermined altitude on the ground, the camera position and The camera posture is directly localized by GPS and IMU mounted on the flying object, and an appropriate number of blocks are arranged in the surveying range, consisting of a plurality of pictures connected via tie points specified in the captured image. An aerial photogrammetry method is proposed in which a ground control point is used as a block adjustment point to perform simultaneous adjustment calculation including the direct localization data to obtain external orientation elements and point coordinates.

  Patent Document 2 (Japanese Patent Application Laid-Open No. 2004-28728) relates to a terrain recognition device that recognizes the undulating shape of the ground, and captures a scene including the ground and outputs a pair of image data, and the pair of images. The parallax is calculated by stereo matching on the basis of the stereo image, and the distance data in which the parallax group related to the image data corresponding to one frame is associated with the coordinate position on the image plane defined by the image data is output. In each of the divided sections, an image processing section, a dividing section that divides the distance data into a plurality of sections on a two-dimensional plane corresponding to a coordinate position on an image plane defined by at least the image data, and And calculating a height group for one section based on the distance data, and calculating a height for the calculated height group. A terrain recognition device has been proposed that includes a calculation unit that calculates a frequency distribution relating to a height, and a height specifying unit that specifies the height of the ground relating to the section for which the frequency distribution is calculated based on the calculated frequency distribution. .

  Further, in Patent Document 3 (Japanese Patent Laid-Open No. 2002-81008), an earthwork design support system using digital map data that uses digital map data to calculate an excavation line and an amount / area of earthwork design and create a design drawing. Has proposed. This earthwork design support system includes 3D geological model generation means for generating a 3D geological model from topographic data having 3D information and geological data having information on geology / material, and excavation based on the 3D geological model. Excavation line generating means for generating an excavation shape of earthwork design according to an input of conditions, calculation means for calculating excavation soil amount and excavation area based on the three-dimensional geological model and the excavation shape, the three-dimensional geological model, And a design drawing creating means for creating a design drawing based on the excavation shape.

Japanese Patent Application Laid-Open No. 2004-245741 JP 2004-28728 A JP 2002-81008 A

  As described above, in the past, various techniques have been proposed for taking aerial photographs using a flying object and performing surveys using these aerial photographs, and using a pair of stereo images, A terrain recognition device that calculates groups and recognizes the undulation shape of the ground has also been proposed.

  However, conventionally proposed aerial photogrammetry methods perform two-dimensional surveys that measure area, distance, etc., and do not recognize a three-dimensional solid shape including height and volume. In this regard, Patent Document 2 proposes a technique for calculating a ground height group using a stereo image, but the calculation result is not accurate enough to be used for surveying.

  Therefore, the present invention uses a plurality of images taken from a high place using a flying object, a crane, a lift, etc., and generates an image for generating point cloud data representing the three-dimensional shape of the target region with high accuracy. A first problem is to provide a photographing method and a method for generating the point cloud data.

  In addition, when performing construction in civil engineering work, etc., soil volume is often calculated to make the designed land, and in that case, surveying is necessary to grasp the shape, slope, unevenness, etc. of the land to be constructed. Has been done. This survey is actually carried out by light wave survey, three-dimensional laser survey on the ground, or triangulation. However, when these surveys are performed, labor costs increase and the number of work days becomes enormous. In addition, when conducting light wave surveying or 3D laser surveying on the ground and performing soil volume calculation based on the results, 3D coordinate data at surveying points acquired based on the survey results are used to perform soil volume calculation. Had to manually enter the computer.

In view of this, the present invention provides a three-dimensional shape of the target area so that the topographic survey can be performed with a small number of people in a short period of time, and the soil volume can be calculated using the data acquired by the survey. A second problem is to provide an image capturing method for generating point cloud data to be represented with high accuracy and a method for generating the point cloud data.

  In order to solve at least one of the above problems, the present invention uses an image capturing method capable of more accurately generating point cloud data representing the three-dimensional shape of the target region, and the captured image. It provides a method for generating point cloud data.

  That is, an image capturing method according to the present invention is an image capturing method used for generating point cloud data representing a three-dimensional shape of a target region from a plurality of captured images. A plurality of captured images to be overlapped with each other by 30% or more, and at least one of the plurality of captured images is set at a predetermined interval in a target area for capturing the plurality of captured images. The image is taken by including at least one or more reference signs.

  The present invention also provides a point cloud data generation method for generating point cloud data representing a three-dimensional shape of a target region using an image captured by the above imaging method. That is, a point cloud data generation method for generating point cloud data representing a three-dimensional shape of a target area from a plurality of captured images, wherein three or more reference signs whose distances are known are set in the target area As a plurality of captured images for setting and generating point cloud data representing the three-dimensional shape of the target region, they overlap each other by 30% or more, and at least one of the plurality of captured images Is a point cloud data generation method that uses a captured image capturing at least one of the reference signs.

  Three or more reference signs are photographed in the image photographed by the above photographing method and the photographed image used in the point cloud data generation method, and the reference sign is arbitrarily set in advance by an operator or the like. Each position information (latitude, longitude, three-dimensional information including altitude, etc.) and mutual distance are known. Therefore, as the three-dimensional position information of the reference sign appearing in the photographed image, set the known position information measured, or as the distance between the reference signs appearing in the photographed image, By setting the distance (actual value), it is possible to correct the captured image and the point cloud data generated from the captured image.

  Specifically, a plurality of reference signs are set in the target area in advance, and the position information of each reference sign is measured to obtain the position information, or the distance between the preset reference signs is measured. Then, as the position information of the reference signs in the point cloud data created from the photographed image, or as the distance between the reference signs, the position information (actually measured values) of the reference signs actually installed in the target area, or between the reference signs The distance (measured distance) measured between the gaps is used. As a result, the three-dimensional position information (X, Y, Z direction position, or latitude, longitude, altitude in the plane coordinates) in the point cloud data is determined by the position information of the reference sign or the measured distance between the reference signs. Since it is corrected, it can be used as an accurate value with higher reliability.

  In addition, it is possible to obtain accurate position information by actually surveying the reference signs, and since this can be used to correct the scale of the point cloud data, the point cloud data can also be made accurate. . Therefore, in the present invention, the accuracy can be ensured by using the position information actually measured as the reference for creating the point cloud data of the target area.

In addition, by installing the reference signs at three or more locations, the reference signs can have a three-dimensional relationship. Based on the three-dimensional relationship, the point cloud data generated from the shooting screen is corrected with respect to the width direction (XY direction, latitude-longitude direction) of the target area and the height method (Z direction, altitude direction). I can do it. As a result, accurate point cloud data can be acquired in three dimensions, and can be used for structure evaluation, land surveying, and soil measurement. Therefore, it is desirable that three or more reference signs should be provided so as to be ubiquitous in the entire target area, not arranged in a straight line, and be present at the edge and the center of the target area. Is more desirable. When the target area for creating the point cloud data is as wide as land, the interval between the reference signs is set within 100 m, particularly preferably within 50 m. Further, in the triangle having the reference sign as a vertex, it is preferable to set so that the distance of one side does not exceed 200 m, preferably does not exceed 150 m, and particularly preferably does not exceed 100 m. Further, it is preferable to set each side in a balanced manner so that the total of the three sides of the triangle does not exceed 650 m, preferably does not exceed 450 m, and particularly preferably does not exceed 300 m. Then, the area of the triangle is 13000 m ² or less, further 10000 m ² or less, particularly preferably disposed so as to be 7000 m ² or less. This is to further improve the accuracy of the point cloud data generated from the photographed image. The accuracy increases as the distance between the reference signs decreases. Moreover, it is desirable that any one or two or more of the distance of one side, the total of three sides, and the area in the triangle having the reference sign as a vertex satisfy the requirement.

  Further, when the above image is taken, a plurality of taken images are taken by overlapping each other by 30% or more, preferably 50% or more, particularly preferably 60% or more. By photographing multiple images in an overlapping manner, one measurement point can be stored in various images taken from different directions, and by doing stereo matching using parallax, etc. Three-dimensional coordinate data can be calculated. An aggregate of the three-dimensional coordinate data can be used as point cloud data.

  The target area to be imaged by the above-described imaging method according to the present invention may be a certain area on the land, or all or a partial area of the building or structure. Therefore, the image image | photographed with the imaging | photography method concerning this invention can be used in order to evaluate the shape, area, or volume of a building or a structure besides survey of the area of a land or undulation.

In carrying out the photographing method of the present invention, a photographing device such as a digital camera using various image sensors can be used, and it is desirable to photograph using a standard lens. If a telephoto lens is used as a lens used by the photographing apparatus, an advantage of obtaining an enlarged image can be obtained, but this point can also be dealt with by a standard lens if the photographing distance is shortened. If a wide-angle lens is used, a wide area can be photographed, but on the other hand, a large distortion due to aberration occurs in most areas including the edge side in the photographed image. Since it is physically impossible to remove all of this distortion, it is extremely difficult to ensure the accuracy of the 3D position information in the point cloud because the 3D position information will not be accurate even if a correction operation is performed. become. Therefore, in the photographing method according to the present invention, it is desirable to photograph using a standard lens having almost no aberration in order to reduce distortion at the edge portion of the photographed image as much as possible. The ground resolution at this time depends on the photographing device and the photographing altitude. For example, when the photographing altitude is 100 m, it is about 2.3 cm ² / pixel. Accuracy can be improved. Note that minute aberrations are present even when the image is taken with a standard lens, and parameters such as the focal length, principal point distance, and distortion coefficient related to the image are corrected, and the point cloud is created after minimizing distortion. Thereby, the point cloud which ensured further accuracy can be generated.

  Further, as described above, it is desirable that the image for creating the point cloud data has a constant angle at the time of photographing the target area, but it does not necessarily have to be a constant angle. For example, when surveying land with a flying object, it is desirable that the imaging device installed on the flying object always shoots the ground surface in a fixed direction in order to reduce distortion in the captured image. The orientation of the flying object may change due to such factors as Therefore, when the angle at the time of shooting is not constant for unavoidable reasons, image processing for correcting the shooting angle is performed for each shot image, and an image that is shot at a fixed angle can be created.

  In order to generate point cloud data for the entire target area using the plurality of captured images, a synthesis process for joining the captured images is performed to represent the three-dimensional shape of the entire target area. Create point cloud data. And about the created point cloud data, the measured value (three-dimensional positional information) measured by surveying etc. is used as the position information of the reference marker in the point cloud data, or the reference markers in the point cloud data are And the scale of the generated point cloud data is corrected based on the actual distance between the reference signs placed in the target area. Finally, parameters such as focal length / principal point distance and distortion coefficient which each image has as information are corrected, and distortion of the generated point group data is corrected. As a result, more accurate point cloud data can be generated.

  When generating the point cloud data, survey point creation points are extracted from the captured image. The survey point creation point is preferably a point that can be used as a mark in the captured image.For example, due to undulations, unevenness, or differences in geology or the presence of buildings, the color, brightness, It is possible to extract points with different tones. Then, for a plurality of extracted points, a normal analytical photogrammetry method such as executing a stereo matching process such as determining a position in a three-dimensional space based on the principle of triangulation using a plurality of photographed images. By performing orientation calculation using, three-dimensional data can be generated for the survey point (point).

  In particular, the above-described shooting method and the point cloud data generation method using the images shot by the shooting method are used in the design (especially soil volume calculation, etc.) for civil engineering construction such as creation when the target area is wide. The effect of improving the efficiency can be exhibited. Therefore, in the present invention, an earthwork design method for creating a design drawing by calculating the excavation line and the excavation soil amount / area of the earthwork design using survey data, the target area generated from a plurality of photographed images as the survey data An earthwork design method is provided in which point cloud data representing the three-dimensional shape is used and the point cloud data is generated by the point cloud data generation method.

  According to the image capturing method of the present invention described above, a plurality of points used when generating point cloud data representing a three-dimensional shape of a target region from a plurality of captured images captured using a standard lens with as little aberration as possible. The captured images overlap each other by 30% or more, and at least one of the plurality of captured images is a reference of three or more places set at predetermined intervals within a target area for capturing the plurality of captured images. Since shooting is performed including at least one of the signs, point cloud data representing the three-dimensional shape of the target region can be generated with high accuracy.

  Then, by using the image captured by this imaging method to generate point cloud data that represents the three-dimensional shape of the target area, point cloud data can also be generated with high precision in photogrammetry technology such as stereo matching processing. can do.

  Furthermore, when the image capturing method is performed and point cloud data is generated using the captured image, the scale is corrected using the values actually measured for the reference sign. The three-dimensional position information in the point cloud data is accurate. Therefore, the soil volume can be accurately calculated even when the soil volume is calculated in civil engineering work such as creation.

  Furthermore, since the parameters such as focal length, principal point distance, and distortion coefficient that each image has as information are corrected and distortion of the generated point cloud data is corrected, the three-dimensional position information in the point cloud data is more accurate. Therefore, the soil volume can be accurately calculated even when the soil volume is calculated in civil engineering work such as creation.

  Furthermore, when multiple photographic images used to generate point cloud data are taken from a high place using a flying object or a lift, many survey points can be set, and the survey points can be accurately set. Point cloud data can be generated. Since the extraction of surveying points can be performed based on a plurality of photographed images, it is possible to save the labor of having to perform surveying for each surveying point. Therefore, it is possible to greatly reduce labor and work time during surveying. Then, the point cloud data (three-dimensional coordinate data at the survey point) generated from the plurality of captured images is subjected to processing such as file conversion as necessary, and is directly input to a computer for soil volume calculation. Quantity calculation can be performed.

Therefore, according to the present invention, a point cloud representing a three-dimensional shape of a target area, which can perform topographic surveying in a short time and with a small number of people, and further can perform soil volume calculation using data acquired by surveying An image capturing method for generating data with high accuracy and a method for generating the point cloud data can be provided.

Schematic diagram showing a method of capturing an image according to the present embodiment Schematic diagram showing captured images Schematic showing the principle of stereo matching processing Plan view showing the installation of reference signs Flow chart showing the flow of work and processing from photography to soil volume calculation Computer screen showing image composition processing Drawing showing the location of reference signs in Tables 1-7 Computer screen showing noise cut processing Computer screen showing file format change status Drawing which shows the result of Example 2

  Embodiments of an image capturing method, a point cloud data generation method using the captured image, and a soil volume calculation method using the point cloud data according to the present invention will be described below with reference to the drawings.

  FIG. 1 is a schematic diagram illustrating an image capturing method according to the present embodiment. In the present embodiment, a case where an image for surveying the land 20 is specifically taken is shown, and a digital photographic image 21 taken by a photographing device 12 such as a digital camera is used as the image to be used. Therefore, in order to take the digital photographic image, a photographing device 12 such as a digital camera using an image sensor is used.

  In addition, the photographing apparatus is installed on the flying object 10 in order to photograph the photographed image from a high place. As the flying object 10, a device capable of flying at a constant speed is used, and in this embodiment, an unmanned flying object that can be operated by radio is used. By using such an unmanned air vehicle 10, it is possible to significantly reduce the cost and procedure required for flight compared to a manned airplane, and it is possible to easily photograph the target area without requiring a precise flight plan. I can do it. However, since the photographed image 21 only needs to be photographed from a certain height, it is not necessarily based on a flying object, and is photographed by installing it on a balloon or the like, or photographed on the tip side of an extended part of a lift or crane, etc. It is also possible to shoot with the device 12 installed.

  The photographing by the photographing device 12 needs to photograph the plurality of photographed images 21 so as to cover the entire target area, and thus the photographing device 12 needs to photograph while moving. Then, in order to generate three-dimensional point cloud data by stereo matching or the like using the photographed image 21, the image is photographed in any region as shown in the schematic diagram showing the photographed image in FIG. The photographed image 21a and the image 21b obtained by photographing the region in contact with the region are photographed so that they overlap each other by at least 30%, desirably 50% or more. Therefore, it is desirable that the image 21b to be photographed after photographing one area is photographed so as to overlap in 30% or more, and further 50% of the photographing area. However, since a plurality of captured images only need to overlap each other by 30% or more, after capturing a plurality of times such as capturing the target region again after capturing the entire target region. Alternatively, overlapping images may be extracted and used.

  In this way, by using overlapping captured images, the object 22 that is shown in common in each image is taken as a measurement point, and stereo matching using the parallax between each image at the measurement point Processing can be performed, and three-dimensional coordinate data at the measurement point can be calculated.

  FIG. 3 is a schematic diagram showing the principle of performing this stereo matching process. As shown in this figure, an imaging device 12 is installed on the flying object 10, and images are taken at predetermined intervals while the flying object 10 is moved. Although the interval at which this imaging is performed can be determined by the moving distance of the flying object 10, it is desirable that the imaging be performed at a constant time interval (several seconds). This is because if the moving speed of the flying object 10 is constant to some extent, the correlation with the moving distance of the flying object 10 can be maintained. Rather, when shooting according to the moving distance, the position of the flying object 10 must be specified by GPS or the like, and the apparatus itself becomes complicated. Therefore, in the present embodiment, the image capturing device 12 is configured to capture an image at regular time intervals using a timer function or the like provided in advance.

  By shooting while moving, the appearance of the undulating slope in the target region 20 varies depending on the shooting position. That is, while moving the flying object, a part of the target area 20 is imaged by the imaging device 12 attached thereto. At this time, the flying object moves substantially horizontally, and the imaging device 12 is attached to the flying object at a certain angle, so that the target area 20 is always photographed in a certain direction. As a result, even when the same object 22 (such as an inclined surface) in the target area 20 is imaged, how the object 22 (such as an inclined surface) appears on the captured image 21 due to the parallax based on the difference in the imaging position. (Especially length) will be different. Then, from the interrelationship between multiple objects in the captured image, 3D coordinate data is generated for the point P (point) of each object by the principle of triangulation, and the point cloud data as an aggregate Can be generated. Such a series of processes until generating point cloud data from a plurality of photographed images can be performed by, for example, a computer capable of executing a program for stereo matching processing.

  In addition, as the object (measurement point) used in the stereo matching process, a workpiece such as a road existing in the target area 20, an uneven part in the target area, a vegetation in the target area, or a target in the target area It is possible to use a portion such as a stone that is reflected in a photograph and having a different color or tone on the photographing screen. In particular, when performing soil volume calculation, not only the width of the target area but also the value of the height (Z-axis direction) of the measurement point P is important. It is desirable to use the starting point and ending point of the inclined part) or the inclined part as the measuring point. Note that since the point cloud data is created based on an object shown in an image such as a photograph, it is desirable that the photographed image has less distortion. Therefore, in the present embodiment, it is desirable to use a standard lens as the lens in the photographing apparatus 12, and it is desirable to select and use the object that exists at the center of the photographed image.

  In the present embodiment, since the three-dimensional coordinate data is calculated by stereo matching, it is desirable that the object 22 serving as the measurement point P is photographed in various images from various angles. This is because the coordinate data at the measurement point is accurately calculated. In addition, it is desirable that the plurality of captured images 21 have a wide overlapping area and use many overlapping images. This is to further improve the accuracy of the point cloud data.

  As described above, by shooting the entire target area, it is possible to generate point cloud data for the entire shooting area. The point cloud data obtained at this time is a relative value that identifies the positional relationship between the points (points), and by adding the distance to the flying object, the magnification of the lens at the time of shooting, and the focal length, , And can be calculated as the distance and length in the target area. However, since the calculated distance and length values are values calculated based on the shooting altitude (height) of the flying object, the magnification of the imaging device, etc., there is a considerable error between the measured values. It will be a thing.

Therefore, in the present embodiment, as shown in the plan view showing the installation state of the reference signs in FIG. 4, the target area 20 is provided with three or more preset reference signs 30. In addition to using a triangular point (reference point), the reference mark 30 can be installed at an arbitrary location after specifying the mutual positional relationship and distance. When setting the reference sign 30 arbitrarily, it is desirable to install it near the boundary (peripheral side) of the target area 20 to be measured, and each reference sign 30 can be connected to any three places It is desirable to arrange. Furthermore, it is desirable that the reference signs 30 are ubiquitous in the central portion of the target area 20 and the like, and it is desirable to install more reference marks 30 in order to improve accuracy. In addition, it is desirable that the reference mark 30 is installed in a convex portion or a concave portion where the boundary portion is curved in the target region 20. This is to generate accurate point cloud data even in deformed land. However, when installing the reference sign 30, it is necessary to survey the three-dimensional coordinate data of each reference sign 30 or to measure the distance from at least one of the reference signs 30. It is desirable to install an appropriate number of reference signs 30 in consideration of the above. As a rough guideline, each reference sign must be at least 3 places in order to establish coordinates on the point cloud, and in order to improve accuracy, the distance between one side of the triangle with each reference sign as the vertex is 200 m. In order not to exceed 150 m, preferably not to exceed 150 m, particularly preferably not to exceed 100 m, and / or the total of the three sides of the triangle should not exceed 650 m, preferably not exceed 450 m as particularly desirably placed so as not to exceed 300 meters, and / or, the area of the triangle 13000 m ² or less, as further 10000 m ² or less, and in particular 7000 m ² or less, the edge portion and the central portion of the target region It is desirable to make it unevenly distributed.

  In addition, the reference mark 30 needs to be formed in a size that can be recognized in the captured image 21, and even in an image captured from the air, the point (measurement point P) such as the center for creating point cloud data is recognized. It is necessary to form the size as possible. When installing such a reference sign 30, three-dimensional coordinate data is specified for a reference sign 30 that is arbitrarily set by surveying from a triangular point or the like, or at least the distance between any reference signs is measured in advance. Keep it. Then, as the three-dimensional position information of the reference sign 30 in the photographed image, the three-dimensional coordinate data actually obtained by surveying is used, or the distance between the reference signs calculated in the photographed image is actually measured. Correct the distance between the reference signs. The correction of the distance between the reference signs is, for example, when the reference signs are installed at intervals of 50 m in the measurement area, and the distance between the reference signs calculated from the captured image 21 is 51 m. Correction for multiplying the distance between the point cloud data generated in the vicinity of the reference sign by a value of “50/51” can be performed. At that time, the correction value can be corrected for all of the three-dimensional coordinate data of the point cloud data, that is, all of the X-axis, Y-axis, and Z-axis (or latitude, longitude, altitude). As a result, since the height (elevation) value on the ground surface can be corrected, a more accurate value can be obtained.

  In the present embodiment, if the size of the reference sign 30 is specified, the three-dimensional coordinate data of the point cloud data is corrected in relation to the size of the reference sign 30 shown in the captured image 21. You can also. That is, when the reference mark 30 is formed in a quadrangle, the length of one side or the length of the diagonal line is specified, and when the reference mark 30 is formed in a circle, if the diameter is specified, The three-dimensional coordinate data in the point cloud data can also be corrected based on the size of the reference sign 30 shown in the captured image 21. For example, when the reference sign 30 actually installed is a quadrangle with a side of 100 cm and the length of one side of the reference sign 30 calculated from the captured image 21 is 99 cm, “99/100” By performing the correction by multiplying the values of the three-dimensional coordinate data in the point cloud data, the accuracy can be improved.

  In addition, when installing the reference sign 30 described above, the reference sign 30 is arbitrarily set up at a place where measurement is possible from a triangular point serving as a reference for surveying by any surveying method, and surveying is performed for each reference sign 30 or any one of them. Install one reference sign 30 at any place, and place the next reference sign at a place of any distance measured using a tape measure or light wave rangefinder, etc. . By sequentially performing this operation, it is possible to install a plurality of reference signs whose distances are known.

A more accurate value can be obtained by giving a coordinate value to the reference sign reflected in the image. However, a minute aberration also exists in an image using a standard lens with little aberration. Therefore, parameters such as focal lengths fx and fy, principal point distances cx and cy, distortion coefficients k ₁ , k ₂ , k ₃ , k ₄ , p ₁ , and p ₂ for each image that forms the point cloud Correct. Such a correction operation can be performed by, for example, an image processing program installed in a computer. As a result, errors due to minute image aberrations can be minimized, and more accurate point cloud data can be created.

  As described above, by generating corrected correct point cloud data from a plurality of captured images 21, a three-dimensional point cloud model using this can be formed. Then, using this three-dimensional point cloud model, it is possible to calculate the amount of soil during creation.

  FIG. 5 is a flowchart showing the flow of work and processing from taking a picture to carrying out soil volume calculation. In this processing flow, in order to measure the area (ie, “target area 20”) to be subjected to soil volume calculation first, a reference sign setting operation for setting the reference mark 30 in the target area is performed ( S51). In addition to installing this reference sign while measuring the distance in advance, after setting the reference sign in any number of locations, survey the three-dimensional information (latitude, longitude, elevation) for each reference sign, Alternatively, the distance between the reference signs can be measured. For the survey of the reference sign, a known survey method such as triangulation, light wave survey, GPS survey or the like can be performed.

  Next, an imaging process is performed in which the flying object 10 to which the imaging device 12 is attached is caused to fly and the entire target area 20 is imaged (S52). In this photographing, a plurality of photographed images 21 in which a part of the target area 20 is photographed are generated by the photographing device 12 using a standard lens with little aberration, and each image is at least 30% or more over other photographed images. Shoot like a lap. Then, after the photographing of the entire target region is completed, an image composition process for composing a plurality of photographed images is performed (S53). This image composition processing is preferably performed while correcting the scale of each image according to the object (measurement point) shown in the overlapping images and the distance between the objects. Although the flying object 10 flies at a certain height, the flying height may slightly change due to natural factors such as wind due to outdoor work. This is because the scale of the change can be considered. Further, at the time of this image synthesis process, the stereo matching process described above can be performed simultaneously to calculate the three-dimensional coordinate data for the measurement point P and to correct the point cloud data based on the reference marker 30. Therefore, by this image composition processing, point cloud data composed of three-dimensional coordinate data for each measurement point can be formed from the captured two-dimensional image. Finally, in order to correct minute aberrations existing in the image, parameters such as the focal length and distortion coefficient of the image are corrected. By this processing, it is possible to generate a point cloud model in which the error is minimized. Such image composition processing can be performed by, for example, an image processing program installed in a computer.

  Next, in order to improve the accuracy of the point cloud data and to improve the processing speed by reducing the data size, a noise cut process for cutting noise is performed on the point cloud data created by the image synthesis process (S54). ). In this noise cut processing, there is heavy equipment in the target area, point cloud data is created for the heavy equipment, and point cloud data other than survey targets is deleted to prevent the ground from being calculated high I do.

  A file format that can be handled by applications (software) that perform point volume calculation using point cloud data that cuts noise and improves the accuracy of point cloud data and reduces the data size. A file conversion process to convert to (S55) is performed.

  After that, in the application for calculating the soil volume, the point volume data generated by the above process can be used to calculate the volume of the soil to carry in and out the soil to make the design surface ( S56).

  Through the above processing, it is possible to create point cloud data with higher accuracy using multiple images of the target area. By using this highly accurate point cloud data, soil volume calculation can be performed accurately. Can be done.

  In the following examples, the photographing device 12 was actually installed on the flying object 10, the land was photographed, coarse point cloud data was created from the photographed image 21, and the accuracy was evaluated by comparing the measurement points.

In this embodiment, the flying object 10 capable of flying at a speed of about 5 m / second above 80 m above the ground is used as the flying object. A standard lens with the following specifications was used as the imaging device 12 mounted on the flying object.
[Standard lens specifications]
・ Focal length 25mm
・ F value 1.8
・ Image sensor (length 13mm, width 17.3mm)
・ Shooting resolution (vertical 3456pixel, horizontal 4607pixel)
・ Resolution (vertical 0.0037616mm / pixel, horizontal 0.0037543mm / pixel)
-Angle of view 47 degrees

In order to prevent out-of-focus images with this photographing device, the shutter speed was set to 1/2000 second, and an image was photographed at a timing of about one in 3 seconds. While using a standard lens to reduce the distortion on the frame side, the image was taken from 100m above the ground in order to widen the range that can be captured in one image. By widening the shooting range for a single image, a large amount of information necessary for performing the image composition processing can be present for the image. The ground resolution at this time is 2.26 (cm ² / pixel).

  In this example, a grid of 30 m × 6 = 180 m and 50 m × 6 = 300 m was set (FIG. 7), and a reference sign 30 was installed at each intersection P. The coordinate values of each reference sign were surveyed in advance, and the coordinate information was grasped for all signs. When photographing, the images were photographed so as to overlap each other by 50% or more.

  A plurality of images photographed as described above were processed by software for performing image composition processing as shown in FIG. 6 to create coarse point cloud data. The computer screen of FIG. 6 includes a display area 61 that displays a three-dimensional point cloud model composed of point cloud data, and a display area 62 that displays a captured image that is the basis of the three-dimensional point cloud model. A plurality of arbitrary marks were selected from a total of 49 reference marks, which were used as reference marks to correct the point cloud.

  The coordinate values of the reference signs that were not adopted on the point cloud data created as described above were treated as survey points, and the same reference signs were compared with the coordinate values previously measured by the optical wave survey. The results are shown in Tables 1 to 7 below.

  In Tables 1 to 7 below, “error from base coordinate” is the difference between the calculated coordinate value of the survey point and the coordinate value previously measured by the light wave survey, and “z value evaluation” is the z value If the error from the base coordinate is within ± 0.05, the sign “◯” is indicated. If the error exceeds ± 0.05 and within ± 0.1, the sign “A” is indicated. If the error is within ± 0.1, the error exceeds ± 0.1 and ± 0.15. If the error exceeds ± 0.15, the code “B” is attached.

  In Tables 1 to 7, “distance to the nearest target” is the distance from the survey point to the three closest reference signs (however, 0 when the survey point is the reference sign). In addition, “the three sides of the smallest triangle to be included” is a triangle in which the survey point is included (if the survey point is a reference sign, the sign is present), and the reference sign is connected so that it is the smallest triangle. The distance between the reference signs. The “minimum triangle area” is the area of the minimum triangle. In addition, the item “on coordinate connection” is given a symbol “◯” when the survey point exists on a line connecting any of the reference signs. In addition, the item “closed figure” is marked with “◯” when the survey point is in the smallest triangle.

  Table 1 below uses the points A1, A3, A5, A7, C1, C3, C5, C7, E1, E3, E5, and G7 on the coordinate axes shown in FIG. This is a result of comparing the coordinate value of another reference sign calculated based on the coordinate value of the survey point with the coordinate value previously measured by the light wave survey.

  As is clear from this result, when the reference signs are set as described above, it is confirmed that all the survey points have an accuracy in the range of -0.05 to 0.05 m and have a high survey accuracy. It was.

  Table 2 below shows other points calculated based on the points A1, A4, A7, D1, D4, D7, G1, G4, and G7 on the coordinate axes shown in FIG. 7 at the intersection of the grids. This is a result of comparing the coordinate value of the reference sign with the coordinate value of the survey point and comparing it with the coordinate value previously measured by the light wave survey.

  As is clear from this result, when the reference signs are set as described above, it is confirmed that all the survey points have an accuracy in the range of -0.05 to 0.05 m and have a high survey accuracy. It was.

  Table 3 below shows other reference signs calculated based on the points A1, A3, A5, A7, G1, G3, G5, and G7 on the coordinate axes shown in FIG. 7 at the intersections of the grids. This coordinate value is the coordinate value of the survey point, and this is compared with the coordinate value previously measured by the light wave survey.

  As is clear from this result, when the reference signs are set as described above, the error range of all survey points is an accuracy within a range of ± 0.1 to ± 0.05 m, which is a high survey accuracy. It was confirmed.

  Table 4 below shows the points A1, A7, D4, G1, and G7 on the coordinate axes shown in FIG. 7 as the reference signs at the intersections of the grids, and the coordinate values of other reference signs calculated based on these points are the survey points. This is a result of comparison with coordinate values previously measured by light wave surveying.

  As is clear from this result, when the reference signs are set as described above, it is confirmed that all the survey points have an accuracy in the range of -0.05 to 0.05 m and have a high survey accuracy. It was.

  Table 5 below shows the points A1, A7, B5, G1, and G7 on the coordinate axes shown in FIG. 7 as reference marks at the intersections of the grids, and the coordinate values of other reference signs calculated based on these points are survey points. This is a result of comparison with coordinate values previously measured by light wave surveying.

  As is clear from this result, when the reference signs are set as described above, the error range of all survey points is an accuracy within a range of ± 0.1 to ± 0.05 m, which is a high survey accuracy. It was confirmed.

  Table 6 below shows the points of A1, A7, G1, and G7 on the coordinate axes shown in FIG. 7 as reference marks at the intersections of the grids, and the coordinate values of other reference signs calculated based on these points are the coordinates of the survey points. This is a result of comparing with a coordinate value previously measured by light wave surveying.

  As is clear from this result, when the reference sign was set as described above, there were also survey points with an error range exceeding ± 0.15, so the reference sign was determined according to the required survey accuracy. It became clear that it was necessary to adjust the arrangement and installation distance of the.

  Table 7 below shows the points of B2, B6, F2, and F6 on the coordinate axes shown in FIG. 7 as the reference signs at the intersections of the grids, and the coordinate values of the other reference signs calculated based on these points are the coordinates of the survey points. This is a result of comparing with a coordinate value previously measured by light wave surveying.

As is clear from this result, when the reference sign was set as described above and the survey point was outside the closed figure, there were also survey points with an error range exceeding ± 0.15. It became clear that it was necessary to adjust the placement and installation distance of the reference signs according to the required survey accuracy.

  In the following examples, the photographing device 12 was actually installed on the flying object 10, the land was photographed, point cloud data was created from the photographed image 21, and the accuracy was evaluated by comparison with the measurement surface.

In this embodiment, the flying object 10 that can fly over 80 m above the ground at a speed of about 5 m / s is used as the flying object, and the standard lens of the specification of the first example is used as the photographing device 12 to prevent blurring. In order to do this, the shutter speed was set to 1/2000 second, and an image was taken at a timing of about one in 3 seconds. Although a standard lens was used to reduce distortion on the frame side, the image was taken from 80 meters above the ground in order to widen the range that can be captured in one image. By widening the shooting range for a single image, a large amount of information necessary for performing the image composition processing can be present for the image. The ground resolution at this time is 1.45 (cm ² / pixel).

  In this example, the distance between the reference signs was set so that the distance between the reference signs closest to each other was within 100 m. When photographing, the images were photographed so as to overlap each other by 50% or more.

  A plurality of images photographed as described above were processed by software for performing image composition processing to create point cloud data. In order to remove noise in the point cloud data, processing was performed with software for cutting noise as shown in the computer screen 61 of FIG. The screen of FIG. 8 includes a display area 63 for displaying a 3D point cloud model composed of point cloud data, and a display area 64 for showing a cross-sectional shape at an arbitrary point in the 3D point cloud model. Yes. Then, in order to convert the noise-cut data into a file format that can be handled by software for calculating soil volume, the file format was changed as shown in the computer screen 70 of FIG.

The value of the point cloud data created as described above was compared with the result of surveying the same target area by ground three-dimensional laser measurement. The result is shown in FIG. As is apparent from FIG. 10, in the survey using the point cloud data obtained in this example, the area of −0.05 to +0.05 m is 8054.49 m ^{2 with} respect to the target area 8153.56 m ² . I know that there is. That is, it was confirmed that most point groups (98.79%) had a precision in the range of -0.05 to 0.05 m and high surveying precision.

  The above-described photography method according to the present invention can be widely used in photogrammetry, and the point cloud data generation method according to the present invention can be carried out in land surveying in the constructed land and soil calculation at the time of creation. It can also be used to measure the size, volume and position of buildings and structures.

10 Unmanned air vehicle
12 Imaging device
20 Target area
21 Images taken
22 Object
30 Reference sign
P Measurement point (intersection point)

Claims (7)

An image capturing method used to generate point cloud data representing a three-dimensional shape of a target region from a plurality of captured images,
A plurality of captured images for generating the point cloud data overlap each other by 30% or more, and at least one of the plurality of captured images is a predetermined area within a target region for capturing the plurality of captured images. A method for photographing an image, comprising photographing at least one of three or more reference signs set at intervals.
A point cloud data generation method for generating point cloud data representing a three-dimensional shape of a target region from a plurality of captured images,
Set three or more reference signs with known distances in the target area,
As a plurality of captured images for generating point cloud data representing the three-dimensional shape of the target region, they overlap each other by 30% or more, and at least one of the plurality of captured images is A point cloud data generation method characterized by using a captured image in which at least one of the reference signs is captured.
The reference signs are such that the distance between one side of a triangle having each reference sign as a vertex does not exceed 200 m, the total of the three sides of the triangle does not exceed 650 m, and the area of the triangle is 13000 m ² or less. The point cloud data generation method according to claim 2, wherein the point cloud data is arranged so as to be unevenly distributed at an edge portion and a center portion of the target region.
By providing known coordinates of the reference sign to the reference sign on the image, the scale of the plurality of captured images is corrected to generate point cloud data representing the three-dimensional shape of the entire target area, and the point cloud The point cloud data generation method according to claim 2, wherein the generated point cloud data is corrected based on a ratio between a distance between the reference markers in the data and a distance between actual reference markers.
The point cloud data generation method according to claim 2, wherein the photographed image is photographed using a standard lens to use an image having as little aberration as possible.
The photographed image is photographed by a photographing device using a standard lens,
The scale of the plurality of captured images is corrected to generate point cloud data representing the three-dimensional shape of the entire target area, and the distance between the reference markers in the point cloud data and the actual distance between the reference markers 3. The point cloud data generation method according to claim 2, wherein the generated point cloud data is corrected based on the ratio, and the shooting parameters including the focal length, principal point distance, and distortion coefficient of the shot image are corrected.
An earthwork design method that uses a survey data to calculate the excavation line, excavation soil volume / area, and create a design drawing.
The point cloud data generating method according to any one of claims 2 to 6, wherein point cloud data representing a three-dimensional shape of a target area generated from a plurality of captured images is used as the survey data. An earthwork design method characterized by being generated by