JP3673857B2 - Method for detecting flat ground from shaded images - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method for detecting a flat surface of a subject surface from a shadow image acquired by a camera.For example, when an unmanned landing ship landes on an unexplored land such as a moon or a planet, the surface of a celestial body including a planned landing point is detected. The present invention relates to a flat ground detection method that can be used to find a horizontal flat ground as a planned landing point from a captured camera image.
[0002]
[Prior art]
When a lander lands on a celestial body such as the moon or planet, if it is a manned lander as in the Apollo program, the lander crew member will determine whether it is possible to land safely by looking at the planned landing point. can do. However, when it is an unmanned lander, such a judgment cannot be made. Also, if a detailed landing instruction is transmitted based on a video received at a remote place by transmitting a video of a planned landing point, signal transmission / reception is not possible. Takes time, and judgment in real time is impossible. Therefore, the unmanned landing ship will force landing regardless of the topography and obstacles. In order to increase the success rate of landing as much as possible, measures are taken to make the structure that the landing aircraft is difficult to fall over or can be dropped.
[0003]
As a method for measuring the shape of the ground surface and obstacles, a laser radar method (for example, Japanese Patent Laid-Open No. 2000-057323), an optical projection method (active stereo method: Japanese Patent Laid-Open No. 9-301300), 2 by improving the computer processing capability. A stereo method using two cameras (passive stereo method: Japanese Patent Application Laid-Open No. 11-259632, Japanese Patent Application No. 2000-349033 by the present applicant) has been proposed.
[0004]
In the laser radar method, altitude can be measured, but there is a problem that it cannot be used unless the altitude is about 100 m due to restrictions on laser light power. In addition, there are problems such as poor resolution and time taken to scan the planned landing point. In the active stereo method, it is necessary to accurately maintain the arrangement between the camera and the light source or the two cameras. Even if it is possible, sufficient measurement accuracy can be obtained unless the altitude is as low as about 100 m. There is no problem. In the passive stereo method, if there is no image feature on the surface of the planned landing point, the surface undulation cannot be detected, but the surface of the moon or planet is covered with sand called regolith, and the image feature is common Very few. At such low altitudes, the range in which landing safety can be confirmed is narrow, so there is a high possibility that a safe landing point cannot be found in the range.
[0005]
The “shape restoration from shadow” algorithm will be specifically described with reference to FIG. FIG. 4 is a perspective view schematically illustrating a “shape restoration from shadow” algorithm. An object surface 1 (hereinafter simply referred to as “surface 1”) such as the surface of the earth to be observed is illuminated by incident light 3 having an incident angle α from a light source 2 such as the sun. Assume that an image of the reflected observation light 4 is taken by a CCD camera 5 (hereinafter simply referred to as “camera 5”). The incident angle α and the observation angle β are defined as angles from a normal 6 standing on the surface 1. The brightness of the image taken by the camera 5 (surface 1 as the object) is (1) the luminance of the light source 2 (sun), (2) the distance R1 between the surface 1 and the light source 2, and (3) the surface 1 It depends on the reflectance characteristics, (4) the distance R2 between the surface 1 and the camera 5 (lander), and (5) the characteristics of the camera 5.
[0006]
(1) The luminance of the light source 2 and (5) the characteristics of the camera 5 can be regarded as constant. When the time is determined, the positions of the light source 2 and the camera 5 are determined, and (2) the distance R1 and (4) the distance R2 are also determined. Therefore, only the reflectance characteristics of the surface 1 are unknown. Since the reflectance characteristic that determines the brightness of the image is a function of the inclination of the surface 1 and the optical characteristic of the surface 1, if the optical characteristic of the surface 1 can be determined or estimated by some method, the inclination and brightness of the surface 1 are determined. (If the optical characteristic is constant, it is approximately proportional). Since the brightness of the surface 1 is obtained from an image captured by the camera 5, the inclination of the surface 1 can be obtained by solving this relational expression using the estimated optical characteristics of the surface 1. The method of integrating the slope of the surface 1 thus determined (accumulating the distance in the in-plane direction) and finally obtaining the surface height (shape) is the theory of “shape restoration from shadows”.
[0007]
However, in this restoration theory, the algorithm for solving the relational expression is weak against noise contained in the image, and further has a problem that it takes a long calculation time, and it is considered that real-time processing is impossible. In order to solve these problems, the Pentland method using the fast Fourier transform (FFT) by linearizing the theory of “shape restoration from shadows” has been proposed. However, the Pentland method has a new problem that the accuracy of the restored shape is deteriorated. Further, in order to avoid the singularity problem in Fourier transform, the average value of the brightness of the image is set to 0, and such average value is used as a reference (that is, the portion indicating the average value of the brightness of the image is assumed to be horizontal). ) The shape of the surface 1 is restored. Even when the entire photographed terrain is tilted in a certain direction, the average value of the brightness of the image is considered to be horizontal, and therefore the restored terrain cannot find the tilt in the certain direction of the entire terrain. Therefore, it is dangerous to determine whether or not landing is possible only from the terrain restored by the Pentland method.
[0008]
By the way, the reflectance characteristic is generally a function of an incident angle of light on a target surface, an observation angle for observing light scattered and reflected on the surface, and an optical characteristic of the surface. If the function can be decomposed into a product of a function of only the incident angle and the observation angle and a function of only the optical characteristic, the optical characteristic becomes a reflection coefficient that does not depend on the incident angle and the observation angle. The inclination of the target surface can be obtained by obtaining the incident angle and the observation angle from the positions of the light source and the camera and the brightness of the image. However, since the reflectance characteristics of the surface 1 such as the moon or planet actually vary depending on the location, it is difficult to accurately estimate the reflectance characteristics. Strictly speaking, the optical characteristics also change depending on the inclination of the surface 1 (incident angle of light, observation angle). However, considering these factors, it takes too much calculation time for shape restoration, and shape restoration in real time. I can't. In many cases, as described above, it is assumed that the surface 1 is a completely diffusing surface, a so-called Lambertian surface, and the reflectance characteristic does not depend on the observation angle. However, if such an assumption is made, the accuracy of shape restoration is lowered.
[0009]
[Problems to be solved by the invention]
Therefore, when applying a shape restoration algorithm that obtains the shape of the subject surface based on the contrast of the image, even if the reflectance characteristics are not clarified in detail, the influence of the inclination in the whole fixed direction is eliminated on the restored surface. Thus, there is a problem to be solved in that the flat surface of the object surface can be obtained while applying the shape restoration algorithm.
[0010]
An object of the present invention is to provide a method for detecting a flat ground from a shadow image, which detects a flat ground on the surface of a subject in real time (real time) required to avoid an obstacle when determining the shape of the surface of the subject based on brightness and darkness. Is to provide.
[0011]
[Means for Solving the Problems]
In order to solve the above-described problem, the flat ground detection method from a shadow image according to the present invention is to obtain the shape of the subject surface based on the brightness of the image with respect to an image obtained by imaging the subject surface from the sky. And detecting the flat ground that can be landed from the surface of the subject in real time by applying the Pentland method that linearizes the shape restoration algorithm of the above and associates the brightness and the shape of the subject surface with Fourier transform. Yes.
[0012]
According to the method for detecting a flat ground from a shadow image, the shape restoration algorithm is not linearly applied to directly determine the shape of the subject surface based on the brightness of the image, but the shape restoration algorithm is linearized and the light and dark and the subject surface are detected. The Pentland method is used to associate the shape of the image with the shape by Fourier transform. The direct application of the shape restoration algorithm is difficult to detect in real time, and the restoration error of the concavo-convex shape becomes large. However, by applying the Pentland method to the restoration of the flat surface shape, the accuracy required for obstacle avoidance is obtained. The flat surface shape can be restored at high speed. For example, when the landing aircraft actually descends, it is possible to detect the land where the land surface can land in real time.
[0013]
In this method for detecting a flat area from a shaded image, a surface that can be regarded as a horizontal plane of the subject surface based on the average value of the brightness obtained for a wide range of high-altitude images obtained by imaging the subject surface from a high altitude. The Pentland method can be applied to an image obtained by imaging the surface of the subject with reference to the reflectance characteristic of the horizontal plane.
[0014]
In a wide range of images obtained when the subject surface is photographed from a high altitude, the partial tilt is canceled and the captured area can be regarded as horizontal as a whole. Based on this, it is possible to determine the reflectance characteristics of a surface that can be regarded as a horizontal surface of the object surface. By using the reflectivity characteristics of the horizontal plane considered as such, it is possible to remove the inclination in a certain direction on the surface of the subject that may appear in an image captured thereafter. That is, the flatness of the subject surface can be obtained without the influence of the inclination in the whole fixed direction on the restored surface. Based on the flatness of the subject surface, it is possible to detect a flat surface having a certain spread on the subject surface. In addition, shape restoration from shadows may cause problems of shape restoration error and reflectance characteristic estimation error, but these errors can be avoided by detecting horizontal flat ground instead of obstacle detection. It becomes possible.
[0015]
In this method of detecting a flat area from a shadow image, for a low altitude image obtained by imaging a part of the subject surface from a low altitude, the brightness and darkness of the subject surface is different from the brightness and darkness of the portion in the high altitude image. The Pentland method can be applied to the low-altitude image after the brightness adjustment by performing brightness adjustment so as to match. A high altitude image obtained by imaging from a high altitude image and a low altitude image obtained by imaging from a low altitude are the same surface parts taken under the same conditions such as a light source, Light and dark are different. Therefore, by performing brightness adjustment that matches the light and darkness in the high altitude image and low altitude image with the same part, the reflectance characteristics of the horizontal surface of the subject surface can be obtained even for low altitude images after brightness adjustment. While maintaining the reference, the Pentland method can be applied in the same manner as in the case of a high altitude image.
[0016]
In this method for detecting a flat ground from a shadow image, the flat ground on the surface of the subject can be determined based on the magnitude of a Fourier coefficient obtained in the solution process of the Pentland method. Since the Fourier coefficient is interpreted as an amplitude corresponding to the spatial frequency of the subject surface, the flat surface of the surface can be determined based on the magnitude of the obtained Fourier coefficient.
[0017]
In the method for detecting a flat ground from the shaded image, the object surface is a celestial surface including a planned landing point of the lander, and the detection of the flat ground on the surface of the celestial body is performed by selecting the planned landing point of the lander. Can be done in real time. In other words, when an unmanned lander attempts to land on the surface of a celestial object on the moon or planet, it detects the flat surface of the celestial surface by applying the flat surface detection method using the celestial surface including the planned landing point as the subject surface. Can be done accurately in time. Therefore, even if the surface has no image features and there are obstacles such as rocks on the surface, it is not an individual obstacle but a surface including the obstacle (in this case, a flat surface). Therefore, it is possible to autonomously select a flat land suitable for landing without control by radio waves from a remote location.
[0018]
When applied to flat surface detection of a celestial body surface including a planned landing point of a lander, for the image obtained by photographing the subject surface, the image is divided into a predetermined number, and the divided image obtained by dividing the image is obtained. By executing the Pentland method, image division / scheduled landing point selection processing is performed for selecting the planned landing point that can be landed most safely based on the flat ground detected for the surface of the subject captured in the divided image. can do. Then, the image division / scheduled landing point selection processing is performed for each predetermined altitude according to the landing aircraft descent until the planned landing point is selected as the final landing planned point that can be landed most safely in the final landing decision image. It is preferable to execute the image on the subject surface. Even if altitude information about the topography of the celestial surface is obtained by some method, it takes time to detect a safe landing point from such altitude information. Therefore, a flat surface is detected by dividing the image obtained by photographing the subject surface by a predetermined number and applying the Pentland method to the subject surface photographed for each divided divided image, and based on the detected flat ground. Thus, the planned landing point that can land safely is selected. As the altitude of the lander decreases, the area of the celestial body surface where the divided images are projected becomes a narrow range, so by finding the flat land for the divided images, you can confirm the point suitable for landing or The selection of the most suitable point can be changed. The above image segmentation / landing point selection process is performed on the surface of the celestial body including the planned landing point at every predetermined altitude as the landing gear descends. This is executed until the final landing scheduled point having the flatness that can be landed most safely is selected in the final landing determination image for determining the point.
[0019]
In the method for detecting a flat ground from the shadow image, the same point on the subject surface is photographed at different times, and an actual value of the inclination of the subject surface is obtained based on the plurality of images having different incident angles at the point, Based on the actual value of the inclination, the actual shape height of the subject surface can be obtained. For example, when detecting the flat surface of the celestial surface including the landing point with a camera mounted on the landing aircraft, the same point on the celestial surface including the planned landing point is taken during the orbit or descent from the light source. By obtaining at least two images having different incident angles, it is possible to know the actual value of the inclination of the celestial surface.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of a method for detecting a flat ground from a shadow image according to the present invention will be described with reference to the accompanying drawings. FIG. 1 is a flowchart showing an outline of a method for detecting a flat ground from a shadow image according to the present invention. The Pentland method, which uses the Fourier transform by linearizing the “shape restoration from shadow” algorithm, cannot accurately perform the shape restoration of obstacles, but it can detect horizontal and horizontal flat surfaces using the method described below. is there.
[0021]
FIG. 1 shows a flow chart for detecting a flat surface suitable for landing when a lander lands on a celestial body surface in the method for detecting a flat ground from a shadow image according to the present invention. As a premise for performing image processing, unless otherwise stated, it is assumed that the position of the light source 2 does not change in a short time range in which landing is performed, and the landing aircraft is lowered under the condition of a constant descent angle (including vertical descent). According to this flowchart, the ground surface including the planned landing point is imaged from the high altitude by the camera 5 as the surface 1 (step 1, hereinafter abbreviated as “S1”, and the like). At high altitude, the field of view of the image is, for example, several tens km square. Next, the average value of the brightness of the high altitude image obtained by photographing from the high altitude is calculated (S2). When the photographed image is from a high altitude, a wide range of the earth's surface is captured, so that the inclination of the captured terrain can be regarded as horizontal on average. That is, the average brightness of the captured image from a high altitude can be regarded as the brightness of the horizontal plane (reflectance characteristics). If the reflectance characteristics of the horizontal plane are obtained, the relative inclination (which of the two points is relatively steeper) can be obtained, but the absolute value of the inclination (actual magnitude value) can be obtained. I don't know. At this stage, since it is a high-altitude image, there is a possibility that it will be an obstacle to landing among those that can be seen for the first time at a low altitude. In order to finally select a safe landing point, it is necessary to obtain not only the relative inclination but also the surface height. In order to obtain the surface height, an absolute value of inclination and an absolute value of distance (actual distance value) are required.
[0022]
Next, an image of a planned landing point is acquired from a position (on the moon / planetary orbit or descending) at a different incident angle (S3). In this case, even when the same photographing object is the same spot, the photographing time is different, so that the relative position with respect to the position of the light source 2 is different, and images having different incident angles are obtained. Images with different incident angles are used to determine the absolute value of the slope of the surface of the planned landing point. That is, the incident angle changes even at the same horizontal plane at different times. Since the direction (incidence angle) of the sun with respect to the horizontal plane can be estimated, reflectance characteristics with respect to the incident angle of the surface can be measured from images taken at different times. If the measurement time is almost the same and the position of the sun is constant, the incident angle changes due to the inclination of the surface according to the topography, and the brightness of the surface changes. By using the reflectance characteristic measured previously, the absolute value of the inclination of the plane can be obtained by the “shape restoration from shadow” algorithm. In this way, by observing the same point at different times, it is possible to know the intensity distribution of the reflected light with respect to the incident angle at the same point as the reflectance characteristic from the change in brightness, and as a result, the absolute value of the slope is calculated. I can know.
[0023]
In accordance with the descent of the landing aircraft, the ground surface including the planned landing point is photographed by the camera 5 at a low altitude (S4). At an altitude of 100 m, the field of view of the image is narrowed to several tens of meters. The average value of images at high altitude obtained in S2 and S4 By comparing the brightness of the same surface area of the image taken at low altitude obtained in step 1, the brightness of the image taken at low altitude and the image taken at high altitude will be the same The brightness is adjusted (S5). The Pentland method is applied to the image that has been subjected to the brightness adjustment in S5 (S6). By applying the Pentland method, it is possible to restore a terrain with a horizontal plane from an image taken at a low altitude. Furthermore, the absolute value of the inclination of the ground surface from the horizontal plane can be calculated from the images of the planned landing points obtained from the positions with different incident angles obtained in S3 (S7).
[0024]
Next, detection of flat ground by the Pentland method will be described. The advantage of the Pentland method is that the shape can be restored without giving boundary conditions. FIG. 2 shows a flowchart when a flat ground in the image processing method according to the present invention is detected using the Pentland method. If the surface irregularities as the surface 1 are considered as a kind of waveform, that is, the surface irregularities are a combination of a number of waves having different frequencies, the coefficient obtained by Fourier transforming the waves is the spatial frequency of the surface 1 ( This corresponds to an amplitude (difference in height) corresponding to a fundamental frequency determined from the area of the divided screen and a frequency that is an integer multiple thereof. It is considered that the flatness of the surface 1 is higher as the coefficient is smaller. In the Pentland method, linear approximation is performed, and as a first step, the lightness and darkness of the image is Fourier transformed. As a second step, a Fourier transformed image is obtained by performing arithmetic processing on the Fourier transformed image. The reason is that the brightness of the image is proportional to the inclination of the surface, and is expressed by the differentiation of the terrain (altitude). Since the differential operation is converted into a product in the Fourier transform, the terrain Fourier transform is obtained by dividing the light-dark Fourier transform of the image by a certain number. As the third stage, the topography (altitude) can be obtained by inverse Fourier transforming the Fourier transform of the topography obtained in the second stage.
[0025]
According to the flowchart shown in FIG. 2, first, the threshold value of the Fourier coefficient is determined in advance from the flatness necessary for the landing machine to land safely without falling (S10). Next, an image of the ground surface is taken by the camera 5 while the lander is descending (S11). The captured image of the ground surface is divided into tiles in the size of an area that can be landed (S12). As an example, the number of image divisions is 10 × 10 (= 100). For each divided image, the flatness of each divided image is obtained from the Fourier transform coefficient obtained in the process of the Pentland method as follows. That is, since the brightness is obtained as a function of the coordinates of the ground surface photographed as an image for each divided image, the above-described calculation using the Fourier transform is performed based on the brightness (S13). A method of detecting flat land by restoring the topography obtained in the third stage is common, but it is also possible to detect flat ground using the Fourier coefficient obtained in the second stage with a short calculation time. It is determined whether the Fourier coefficient obtained in the second stage is smaller than the threshold value, and a landing candidate point is selected by determining that the Fourier coefficient is smaller than the threshold value (S14). The area of the surface 1 corresponding to the divided image is registered and updated as a planned landing point on the assumption that the lander can land (S15). After S15, the analysis image Of a given altitude It is determined whether it is a final landing image (S16). If it is determined in S16 that the image is the final landing image, the analysis is terminated. When it is determined in S16 that the analysis image is not the final landing image, the process returns to S11 to take an image of the ground surface, and the flatness analysis based on brightness is performed on the next divided image. By determining the Fourier coefficient, it is determined whether or not the landing aircraft can land for each divided image. When there are a plurality of landing candidate sites registered in S15, the analysis is further advanced. For example, the landing candidate site corresponding to the screen having the smallest Fourier coefficient can be selected as the final landing scheduled point. The Fourier coefficient for the higher-order spatial frequency can be ignored because the contribution to the flatness is small, so that the flatness determination process can be speeded up.
[0026]
Next, detection from high altitude will be described. The “shape restoration from shadow” algorithm is a detection method for obtaining the shape of the surface from the brightness of the surface, so that even an altitude difference of 1/10 or less of the resolution of the image can be detected. Since it is relatively easy to obtain an image obtained by photographing the surface of a planned landing point from an altitude of several kilometers, it is relatively easy to obtain an object having a resolution of about 1 m. Therefore, if the flat land detection method according to the present invention is applied to such an image, It is possible to detect a flatness of about 0.1 m. In other words, the Pentland method, which linearizes the “shape restoration from shadows” algorithm, cannot accurately perform the shape restoration of obstacles, but considers obstacles such as isolated rocks as surface undulations, It becomes possible to detect a flat land in real time. In addition, even on unexplored lands such as the moon and planets, it is sufficient for the lander to land if it is possible to find a flat, flat surface with no inclination in the target range including the planned landing point.
[0027]
FIG. 3 shows an example of Fourier coefficients obtained in the solution process of the Pentland method in the method for detecting a flat ground from a shadow image according to the present invention (for the sake of simplicity, the image size is 32 × 32 pixels, a small number of each coefficient). The following is a diagram showing the truncation). In FIG. 3, Fourier coefficients for each two-dimensional spatial frequency are displayed in a matrix. (A) is an example of a Fourier coefficient when an obstacle (for example, a crater) exists, and (b) is an example of a Fourier coefficient of a relatively flat terrain. The topography is considered to be a synthesis of waves with two-dimensional spatial frequencies. In the figure, the origin is at the upper left, and the closer to the origin, the lower the spatial frequency, the higher the coefficient corresponding to the lower order. The coefficient corresponding to the spatial frequency is represented. In either case, the Fourier coefficient corresponding to the higher-order spatial frequency is zero, which corresponds to the fact that there is no terrain where the height difference changes rapidly in a narrow range, as in the actual terrain. . Non-zero Fourier coefficients are relatively concentrated in the vicinity of the origin, but when the obstacle (a) exists, the value of the Fourier coefficient is smaller than in the case of the relatively flat topography (b). It can be seen that the amplitude of the waveform component such as the swell of the terrain corresponding to the low frequency is large. Thus, the flatness of the terrain appearing in the image can be determined based on the magnitude of the Fourier coefficient distributed in two dimensions.
[0028]
In the following, an embodiment as a system to which the flat ground detection method using a shadow image according to the present invention is applied will be described. The “shape restoration from shadow” algorithm that restores the 3D shape from a single camera image can obtain the relative height difference of the terrain based on the size of the image, but the absolute height difference value is I don't understand. Since the absolute value of the slope is found by using the above method, the height can be known if the absolute value of the distance is known. The following three methods can be considered to know the absolute value of the distance.
(1) Using only one camera image
This method is effective when the terrain shown in the camera contains a terrain whose size is known, and the size of the area where the image is shown can be determined from the known terrain. Based on the size of the image range, the absolute value of the terrain height difference can be obtained.
(2) Method of using altitude information from measuring equipment such as laser radar and radio altimeter
If altitude information up to the planned landing point can be obtained by a measuring device such as a laser radar or radio altimeter, the camera's viewing angle is known, so the range shown in the image is known, and the absolute value of the altitude difference of the topography is calculated. be able to.
(3) Using two camera images
By using in combination with the passive stereo method using two camera images, absolute altitude information can be obtained. When the reflectance characteristics of the surface of the planned landing point change even in a narrow range (that is, there is an image feature), the reliability of the shape restoration accuracy obtained by the “shape restoration from shadow” algorithm decreases, The passive stereo method makes it easy to solve the corresponding point problem and improves the distance measurement accuracy. On the other hand, when the reflectance characteristics of the surface of the planned landing point are uniform (ie, there are few image features), the distance measurement accuracy of passive stereo is reduced, but it is obtained by the “shape restoration from shadow” algorithm The shape restoration accuracy is improved. Similar to the human visual system, the “shape reconstruction from shadow” algorithm and the passive stereo method function in a complementary manner.
[0029]
In the above example, the landing of an unmanned lander was explained. However, even a manned lander should be equipped with a system to which the flatness detection method according to the present invention is applied as a system that assists in finding a landing site. Is preferred.
[0030]
【The invention's effect】
According to the flat land detection method from the shadow image according to the present invention, it is possible to accurately find a land that can be landed in real time without knowing the details of the brightness (reflectance characteristics) of the horizontal plane. Even if the landing aircraft is descending to an unexplored surface such as the surface of a celestial body, it is possible to find the optimal point for landing from a wide range including the predetermined landing point in real time Become.
[Brief description of the drawings]
FIG. 1 is a flowchart showing an outline of a method for detecting a flat land when the method for detecting a flat ground from a shaded image according to the present invention is applied to detection of a flat surface suitable for landing when a lander lands on a celestial body surface. It is.
FIG. 2 is a flowchart when a flat ground is detected using the Pentland method in the image processing method according to the present invention.
FIG. 3 is a diagram showing an example of a Fourier coefficient obtained in the solution process of the Pentland method in the flat ground detection method from a shadow image according to the present invention.
FIG. 4 is a perspective view schematically illustrating a “shape restoration from shadow” algorithm.
[Explanation of symbols]
1 Subject surface 2 Light source
3 Incident light 4 Observation light
5 CCD camera 6 Normal
α Incident angle β Observation angle
R1 Distance between surface 1 and light source 2 R2 Distance between surface 1 and camera 5

Claims (6)

A reflectance characteristic of a surface that can be regarded as a horizontal plane of the subject surface is determined based on the average value of the brightness obtained for a wide range of high altitude images obtained by imaging the subject surface from a high altitude, With respect to a low altitude image obtained by imaging a part from a low altitude, brightness adjustment is performed so that the brightness of the subject surface matches the brightness of the portion of the high altitude image, and the brightness after dark adjustment is performed. A method for detecting a flat ground from a shadow image, wherein the horizontal plane is detected by applying the Pentland method to a low-altitude image. The method of claim 1 , wherein the flatness of the object surface is determined based on a magnitude of a Fourier coefficient obtained in a solution process of the Pentland method. The object surface is astronomical surface comprising landing scheduled point lander, detection of the flat land of the celestial body surface, according to claim, characterized in that it is carried out for the selection of the landing scheduled point of the lander A flat ground detection method from the shaded image according to 1 or 2 . For the image obtained by photographing the subject surface, the image is divided into a predetermined number, the Pentland method is applied to the divided image obtained by the division, and the flatness is determined from the magnitude of the Fourier coefficient at that time. 4. The method for detecting a flat ground from a shadow image according to claim 3 , wherein the planned landing point with the highest flatness and the safest landing is selected from the divided images . The image segmentation / scheduled landing point selection process is performed in such a manner that the subject at every predetermined altitude is lowered according to the landing aircraft descent until the planned landing point is selected as a final landing planned point that can be landed most safely in the final landing determination image. The method according to claim 4, wherein the method is performed on the image obtained by photographing a surface. Photographing the same point on the subject surface at different times, obtaining an actual value of the tilt of the subject surface based on the plurality of images having different incident angles at the point, and determining the actual surface of the subject based on the actual value of the tilt The method for detecting a flat ground from a shadow image according to any one of claims 1 to 5, wherein an actual shape height is obtained.

Images