Deep space lander landing point selection method based on terrain feature quantization

Abstract

The invention discloses a deep space probe landing point selection method based on topographic feature quantification, and the method comprises the steps: constructing a multi-target topographic feature optimization model through analyzing the parameters of a landing region, such as the topographic slope, the surface roughness, the obstacle distribution and the regional area, and achieving the automatic selection of a safe landing point. The method comprises the following specific steps: firstly, detecting spatial distribution of potential obstacles such as rocks and meteorite craters, and secondly, quantitatively analyzing topographic feature parameters such as topographic roughness, gradient change rate and topographic relief; and finally, an optimal landing position is determined under multiple terrain constraint conditions by using an optimization search algorithm, so that the landing safety and task success rate of the detector are effectively improved. According to the method, the safe landing point of the detector is accurately obtained by innovatively applying terrain feature quantification and a search algorithm, so that effective detection of a safe position in a landing area is efficiently realized, and safe landing of the detector is practically guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous landing navigation of deep space probes, and in particular relates to a method for selecting a landing point for a deep space lander based on quantification of terrain features. Background Art

[0002] Deep space exploration, as an important means for mankind to explore the frontiers of the universe, is not only a key symbol of a country's scientific and technological strength and comprehensive national strength, but the autonomous soft landing of its probe during the landing and descent process is even more crucial. During the powered descent phase, after completing the selection of candidate landing areas, the terrain of the landing area needs to be detected and obstacles avoided. The terrain during the descent process is scanned by sensors carried by the satellite to perform rough obstacle avoidance, so as to assist the control system in avoiding possible large obstacles and ensure that the probe can safely approach the landing area. When approaching the landing area, the probe needs to hover and perform more fine-grained terrain detection, using equipment such as landing cameras and laser rangefinders to obtain higher-resolution terrain data to detect flat areas, obstacles or other features. These data will assist the control system in determining the final landing point and adjusting the attitude for a safe landing.

[0003] To ensure the probe's safe landing in the pre-selected landing zone and select a suitable location for landing, a deep space lander must first detect obstacles in the terrain, such as rocks and slopes. The landing area terrain is then inspected for craters, rocks, and slopes. Finally, a landing site selection algorithm is used to select a safe landing site for the probe under various terrain constraints.

[0004] In deep space exploration, selecting a landing site is both a critical and challenging task. Besides avoiding obstacles in the landing area, terrain and other factors must also be considered. The probe's inherent characteristics, obstacles, slope, roughness, and size of the landing area all affect landing safety and must therefore be fully considered during the landing process. This study addresses the multi-objective constraints of the landing area terrain, constructs a quantitative model of terrain characteristics, and uses a search algorithm to find the optimal landing site while satisfying these constraints. Summary of the Invention

[0005] This paper provides a method for selecting a landing site for a deep space lander based on quantified terrain features, aiming to ensure a safe landing within the landing zone during a deep space probe's soft landing. By using a quantitative model of the landing zone's terrain features and a landing zone search algorithm that satisfies multiple objective constraints, the method selects the maximum landing area corresponding to the optimal safe landing point within the landing zone, thereby improving the safety of the probe's landing process.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for selecting a landing site for a deep space lander based on quantification of terrain features comprises the following steps:

[0008] S1. Acquire 3D terrain data of the lander landing area and preprocess the landing area image;

[0009] S2. Use deep learning algorithms to accurately detect potential obstacles such as rocks and craters in the landing area;

[0010] S3. Analyze the terrain characteristics of the landing area, including terrain slope, surface roughness, obstacle distribution, and area parameters, and establish a quantitative model of the multi-objective terrain characteristics of the landing area;

[0011] S4. Obtain the optimal landing point coordinates based on the terrain multi-objective feature quantification model and the landing constraints of the landing area.

[0012] Furthermore, the step S1 is specifically as follows:

[0013] S11. Obtain the digital elevation map (DEM) of the Zhurong Mars rover landing area;

[0014] S12. Normalize the size of the landing area image and adjust it to 256×256 pixels.

[0015] Furthermore, the step S2 is specifically as follows:

[0016] S21. Fine-tune the deep Unet network to detect obstacles in the landing area of deep space probes, including rocks and crater obstacles in the lander terrain.

[0017] Furthermore, the step S3 is specifically as follows:

[0018] S31. Analyze the terrain characteristics of the landing area and establish a quantitative model of terrain characteristics that integrates multi-dimensional constraints;

[0019] S32. Represent the landing area terrain, including obstacles, terrain roughness, slope, undulation, and landing area constraint characteristics;

[0020] S33. Based on step S21, accurate extraction of the terrain obstacle feature O(x′) of the landing area is achieved;

[0021] S34. The roughness R of the terrain is used to describe the degree of unevenness of the object surface. The difference between the fitted plane and the real data is used to estimate the roughness of the terrain. The roughness R is expressed as Where: A, B, C and D are the parameters of the fitting plane, x0, y0, z0 are the three-dimensional coordinates of the real terrain data point; the numerator Ax0+By0+Cz0+D represents the absolute value of the directed distance from the point (x0, y0, z0) to the fitting plane, and the denominator is is the modulus of the normal vector of the fitted plane, which is used to normalize the distance; Ax+By+Cz=D is the local plane of the probe landing area map, where: x, y, z are coordinates in three-dimensional space;

[0022] S35. Slope S is the maximum inclination angle in all directions at a point on the surface of the earth, from the east-west slope and north-south slope In the digital elevation model (DEM) raster data, the local terrain features are calculated by the elevation difference between the target pixel and its adjacent pixels, and the east-west slope Characterizes the horizontal elevation change per unit distance in the east-west direction and the slope in the north-south direction Represents the elevation change per unit distance in the horizontal north-south direction; the slope S is calculated using the vector synthesis principle and is expressed as The maximum slope angle is solved by the modulus of the two-dimensional slope vector, and the steepest slope of the point on the surface is expressed in degrees; among them, tan -1 is the inverse tangent function, used to convert the composite slope into an angle value;

[0023] S36. The terrain undulation F represents the degree of ground undulation. The maximum value of the terrain undulation E is used. max and the minimum value E min The difference between the two measures the fluctuation, F is expressed as E max -E min ;

[0024] S37. Based on the constraints of the quantified landing area terrain characteristics, multi-dimensional terrain constraints are introduced and the range of the objective function is restricted, thereby finding the optimal solution while satisfying the constraints. The constraint model for selecting the optimal landing point based on the quantified landing area terrain characteristics is expressed as follows:

[0025] minimize g({R(x′),S(x′),F(x′),O(x′),A(x′)})

[0026] subject to

[0027] r′ <mean(R(x′))

[0028] s′ <mean(S(x′))

[0029] f′ <mean(F(x′))

[0030] o′ <mean(O(x′))

[0031] a′>mean(A(x′))

[0032] Where g(·) is the objective function that satisfies all constraints; R(x′), S(x′), F(x′), O(x′), and A(x′) represent the overall roughness distribution, slope distribution, undulation, obstacle density, and safe area characteristics of the landing area; x′ is the set of pixel parameters that make up the landing area of the terrain, and mean(·) is the statistical average operation of all pixels within the domain of x′; r′, s′, f′, o′, and a′ represent the local roughness, local slope, local undulation, local obstacle value of the optimal landing point and its neighborhood, and the safe area corresponding to the optimal landing point, respectively.

[0033] Furthermore, the step S4 is specifically as follows:

[0034] S41. In the digital elevation model S map , with a size of N×N pixels. According to the actual needs of the Zhurong landing process, obstacle recognition and distribution recognition are first performed based on S21, and then the digital elevation map S map The algorithm uses the map center coordinates P(x0, y0) as the initial search base point, performs spatial iterative calculations through the multi-constraint optimization function g(·), and finally determines the characteristic parameters of the maximum safe landing area: the center coordinates of the safe landing area P fit (x f ,y f ), optimal safe landing radius r, effective coverage area S fit ; Among them: (x f ,y f ) becomes the best landing target point that meets the terrain constraints after multi-objective optimization;

[0035] S42. Based on multi-dimensional terrain constraints, use the depth and breadth first search of the graph to obtain S near the landing point best The area within the range that meets the landing constraints has its corresponding center coordinates as the optimal landing point;

[0036] S43. During the landing point acquisition algorithm in the landing zone, the Visited set is used to store the coordinates that have been visited during the search process, and the Q queue stores the coordinates of the locations to be visited. The initial radius during the landing zone search process is R0, and the minimum safe landing area area is S min ;

[0037] S44. The algorithm starts at coordinate P and increases the radius in sequence, searching for a location area at that coordinate that satisfies the landing point constraint. First, it determines whether there are any obstacles within the area, i.e., whether the constraint O(x′) is satisfied. After eliminating any dangerous areas, it determines whether the surrounding points have been visited. If they have not been visited and are not in queue Q, the point is added to the queue to be visited, reducing repeated point scanning, until the queue is empty.

[0038] S45. In an area without obstacles, determine whether other constraints in the area near the point are met. If so, return the nearest landing point and area that meet the constraints;

[0039] If not, calculate the safety factor using the Safe(·) function. Divide the image into four areas based on the center of the map. Compare the flat / dangerous area ratios to obtain the safety factors (s1, s2, s3, s4) for the four areas. Select the area with the highest safety factor as the movement direction.

[0040] S47. The final output result is: the center coordinates of the safe landing area P fit (x f ,y f ), the terrain characteristic parameters of the maximum feasible area include the average roughness mr, the average undulation mf and the average slope ms, the optimal safe landing radius r of the area, and the effective coverage area S of the area fit 4r 2 .

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] The present invention constructs a multi-objective terrain feature quantification model and applies it to the terrain of the deep space landing area, and uses a search algorithm to achieve the best safe landing point in the landing area. In this process, constraints such as the terrain slope, surface roughness, obstacle distribution, and landing area area of the landing area are taken into account at the same time, and an optimal landing point selection method based on multi-objective constraints of the landing area terrain is proposed to ensure the safe landing of the probe and the successful execution of the mission. The detection results of craters on the surfaces of the moon, Mars, and asteroids show that the method proposed in this paper has high crater detection accuracy and inference speed. In addition, the proposed landing point selection method can obtain the best selection under multiple constraints to ensure the safe landing of the probe. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Flowchart of the present invention.

[0044] Figure 2 This is a topographic map of the Zhurong landing area.

[0045] Figure 3 Result plots for landing point selection under different constraints. DETAILED DESCRIPTION

[0046] The present invention will be further described below in conjunction with the embodiments.

[0047] Example 1

[0048] like Figure 1As shown, a method for selecting a landing site for a deep space lander based on quantification of terrain features includes the following steps:

[0049] S1. Acquire 3D terrain data of the lander landing area and preprocess the landing area image;

[0050] S11. Figure 2 As shown, obtain the digital elevation map (DEM) of the landing area of the Zhurong Mars rover;

[0051] S12. Normalize the size of the landing area image and adjust it to 256×256 pixels.

[0052] S2. Use deep learning algorithms to accurately detect potential obstacles such as rocks and craters in the landing area;

[0053] S21. Fine-tune the deep Unet network to detect obstacles in the landing area of deep space probes, including rocks and crater obstacles in the lander terrain.

[0054] S3. Analyze the terrain characteristics of the landing area, including terrain slope, surface roughness, obstacle distribution, and area parameters, and establish a quantitative model of the multi-objective terrain characteristics of the landing area;

[0055] S31. Analyze the terrain characteristics of the landing area and establish a quantitative model of terrain characteristics that integrates multi-dimensional constraints;

[0056] S32. Represent the landing area terrain, including obstacles, terrain roughness, slope, undulation, and landing area constraint characteristics;

[0057] S33. Based on step S21, accurate extraction of the terrain obstacle feature O(x′) of the landing area is achieved;

[0058] S34. The roughness R of the terrain is used to describe the degree of unevenness of the object surface. The difference between the fitted plane and the real data is used to estimate the roughness of the terrain. The roughness R is expressed as Where: A, B, C and D are the parameters of the fitting plane, x0, y0, z0 are the three-dimensional coordinates of the real terrain data point; the numerator Ax0+By0+Cz0+D represents the absolute value of the directed distance from the point (x0, y0, z0) to the fitting plane, and the denominator is is the modulus of the normal vector of the fitted plane, which is used to normalize the distance; Ax+By+Cz=D is the local plane of the probe landing area map, where: x, y, z are coordinates in three-dimensional space;

[0059] S35. Slope S is the maximum inclination angle in all directions at a point on the surface of the earth, from the east-west slope and north-south slope In the digital elevation model (DEM) raster data, the local terrain features are calculated by the elevation difference between the target pixel and its adjacent pixels, and the east-west slope Characterizes the horizontal elevation change per unit distance in the east-west direction and the slope in the north-south direction Represents the elevation change per unit distance in the horizontal north-south direction; the slope S is calculated using the vector synthesis principle and is expressed as The maximum slope angle is solved by the modulus of the two-dimensional slope vector, and the steepest slope of the point on the surface is expressed in degrees; among them, tan -1 is the inverse tangent function, used to convert the composite slope into an angle value;

[0060] S36. The terrain undulation F represents the degree of ground undulation. The maximum value of the terrain undulation E is used. max and the minimum value E min The difference between the two measures the fluctuation, F is expressed as E max -E min ;

[0061] S37. Based on the constraints of the quantified landing area terrain characteristics, multi-dimensional terrain constraints are introduced and the range of the objective function is restricted, thereby finding the optimal solution while satisfying the constraints. The constraint model for selecting the optimal landing point based on the quantified landing area terrain characteristics is expressed as follows:

[0062] minimize g({R(x′),S(x′),F(x′),O(x′),A(x′)})

[0063] subject to

[0064] r′ <mean(R(x′))

[0065] s′ <mean(S(x′))

[0066] f′ <mean(F(x′))

[0067] o′ <mean(O(x′))

[0068] a′>mean(A(x′))

[0069] Where g(·) is the objective function that satisfies all constraints; R(x′), S(x′), F(x′), O(x′), and A(x′) represent the overall roughness distribution, slope distribution, undulation, obstacle density, and safe area characteristics of the landing area; x′ is the set of pixel parameters that make up the landing area of the terrain, and mean(·) is the statistical average operation of all pixels within the domain of x′; r′, s′, f′, o′, and a′ represent the local roughness, local slope, local undulation, local obstacle value of the optimal landing point and its neighborhood, and the safe area corresponding to the optimal landing point, respectively.

[0070] S4. Obtain the optimal landing point coordinates based on the terrain multi-objective feature quantification model and the landing constraints of the landing area;

[0071] S41. In the digital elevation model S map , with a size of N×N pixels. According to the actual needs of the Zhurong landing process, obstacle recognition and distribution recognition are first performed based on S21, and then the digital elevation map S map The algorithm uses the map center coordinates P(x0, y0) as the initial search base point, performs spatial iterative calculations through the multi-constraint optimization function g(·), and finally determines the characteristic parameters of the maximum safe landing area: the center coordinates of the safe landing area P fit (x f ,x f ), optimal safe landing radius r, effective coverage area S fit ; Among them: (x f ,y f ) becomes the best landing target point that meets the terrain constraints after multi-objective optimization; and meets the safe landing constraints g{r(x′),s(x′),f(x′),o(x′),a(x′)}.

[0072] S42. Based on multi-dimensional terrain constraints, use the depth and breadth first search of the graph to obtain S near the landing point best The area within the range that meets the landing constraints has its corresponding center coordinates as the optimal landing point;

[0073] S43. During the landing point acquisition algorithm in the landing zone, the Visited set is used to store the coordinates that have been visited during the search process, and the Q queue stores the coordinates of the locations to be visited. The initial radius during the landing zone search process is R0, and the minimum safe landing area area is S min ;

[0074] S44. The algorithm starts at coordinate P and increases the radius in sequence, searching for a location area at that coordinate that satisfies the landing point constraint. First, it determines whether there are any obstacles within the area, i.e., whether the constraint O(x′) is satisfied. After eliminating any dangerous areas, it determines whether the surrounding points have been visited. If they have not been visited and are not in queue Q, the point is added to the queue to be visited, reducing repeated point scanning, until the queue is empty.

[0075] S45. In an area without obstacles, determine whether other constraints in the area near the point are met. If so, return the nearest landing point and area that meet the constraints;

[0076] If not, calculate the safety factor using the Safe(·) function. Divide the image into four areas based on the center of the map. Compare the flat / dangerous area ratios to obtain the safety factors (s1, s2, s3, s4) for the four areas. Select the area with the highest safety factor as the movement direction.

[0077] S47. The final output result is: the center coordinates of the safe landing area P fit (x f ,y f ), the terrain characteristic parameters of the maximum feasible area include the average roughness mr, the average undulation mf and the average slope ms, the optimal safe landing radius r of the area, and the effective coverage area S of the area fit 4r 2 .

[0078] The landing point selection results of the landing area under different constraints are as follows: Figure 3 In the experiment, constraints were set for the area around the pre-selected landing point, including average roughness mr < 0.3, average undulation mf < 30, average slope ms < 3, and avoiding rocks and craters. The final output is the center coordinates of the safe landing area P fit (x f ,y f ), the values of the safe landing area mr, mf, ms, the optimal safe landing radius r of the area, and the effective coverage area S of the area fit 4r 2 .

[0079] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for selecting a landing site for a deep space lander based on quantification of terrain features, characterized in that: The following steps are involved: S1. Acquire 3D terrain data of the lander landing area and preprocess the landing area image; S2. Use deep learning algorithms to accurately detect potential obstacles such as rocks and craters in the landing area; S3. Analyze the terrain characteristics of the landing area, including terrain slope, surface roughness, obstacle distribution, and area parameters, and establish a quantitative model of the multi-objective terrain characteristics of the landing area; S4. Obtain the optimal landing point coordinates based on the terrain multi-objective feature quantification model and the landing constraints of the landing area.

2. The method for selecting a landing site for a deep space lander based on quantification of terrain features according to claim 1, wherein: The step S1 is specifically as follows: S11. Obtain the digital elevation map (DEM) of the Zhurong Mars rover landing area; S12. Normalize the size of the landing area image and adjust it to 256×256 pixels.

3. The method for selecting a landing site for a deep space lander based on quantification of terrain features according to claim 2, wherein: The step S2 is specifically as follows: S21. Fine-tune the deep Unet network to detect obstacles in the landing area of deep space probes, including rocks and crater obstacles in the lander terrain.

4. The method for selecting a landing site for a deep space lander based on quantification of terrain features according to claim 3, wherein: The step S3 is specifically as follows: S31. Analyze the terrain characteristics of the landing area and establish a quantitative model of terrain characteristics that integrates multi-dimensional constraints; S32. Represent the landing area terrain, including obstacles, terrain roughness, slope, undulation, and landing area constraint characteristics; S33. Accurately extract the terrain obstacle features O(x′) in the landing area based on step S21; S34. The roughness R of the terrain is used to describe the degree of unevenness of the object surface. The difference between the fitted plane and the real data is used to estimate the roughness of the terrain. The roughness R is expressed as Where: A, B, C and D are the parameters of the fitting plane, x0, y0, z0 are the three-dimensional coordinates of the real terrain data point; the numerator Ax0+By0+Cz0+D represents the absolute value of the directed distance from the point (x0, y0, z0) to the fitting plane, and the denominator is is the modulus of the normal vector of the fitted plane, which is used to normalize the distance; Ax+By+Cz=D is the local plane of the probe landing area map, where: x, y, z are coordinates in three-dimensional space; S35. Slope S is the maximum inclination angle in all directions at a point on the surface of the earth, from the east-west slope and north-south slope In the digital elevation model (DEM) raster data, the local terrain features are calculated by the elevation difference between the target pixel and its adjacent pixels, and the east-west slope Characterizes the horizontal elevation change per unit distance in the east-west direction and the slope in the north-south direction Represents the elevation change per unit distance in the horizontal north-south direction; the slope S is calculated using the vector synthesis principle and is expressed as The maximum slope angle is solved by the modulus of the two-dimensional slope vector, and the steepest slope of the point on the surface is expressed in degrees; among them, tan -1 is the inverse tangent function, used to convert the composite slope into an angle value; S36. The terrain undulation F represents the degree of ground undulation. The maximum value of the terrain undulation E is used. max and the minimum value E min The difference between the two measures the fluctuation, F is expressed as E max -E min ; S37. Based on the constraints of the quantified landing area terrain characteristics, multi-dimensional terrain constraints are introduced and the range of the objective function is restricted, thereby finding the optimal solution while satisfying the constraints. The constraint model for selecting the optimal landing point based on the quantified landing area terrain characteristics is expressed as follows: minimize g({R(x′),S(x′),F(x′),O(x′),A(x′)}) subject to r′ <mean(R(x′)) s′ <mean(S(x′)) f′ <mean(F(x′)) o′ <mean(O(x′)) a′>mean(A(x′)) Where g(·) is the objective function that satisfies all constraints; R(x′), S(x′), F(x′), O(x′), and A(x′) represent the overall roughness distribution, slope distribution, undulation, obstacle density, and safe area characteristics of the landing area; x′ is the set of pixel parameters that make up the landing area of the terrain, and mean(·) is the statistical average operation of all pixels within the domain of x′; r′, s′, f′, o′, and a′ represent the local roughness, local slope, local undulation, local obstacle value of the optimal landing point and its neighborhood, and the safe area corresponding to the optimal landing point, respectively.

5. The method for selecting a landing site for a deep space lander based on quantification of terrain features according to claim 4, wherein: The step S4 is specifically as follows: S41. In the digital elevation model S map , with a size of N×N pixels. According to the actual needs of the Zhurong landing process, obstacle recognition and distribution recognition are first performed based on S21, and then the digital elevation map S map The algorithm uses the map center coordinates P(x0, y0) as the initial search base point, performs spatial iterative calculations through the multi-constraint optimization function g(·), and finally determines the characteristic parameters of the maximum safe landing area: the center coordinates of the safe landing area P fit (x f ,y f ), optimal safe landing radius r, effective coverage area S fit ; Where: (x f ,y f ) becomes the best landing target point that meets the terrain constraints after multi-objective optimization; S42. Based on multi-dimensional terrain constraints, use the depth and breadth first search of the graph to obtain S near the landing point best The area within the range that meets the landing constraints has its corresponding center coordinates as the optimal landing point; S43. During the landing point acquisition algorithm in the landing zone, the Visited set is used to store the coordinates that have been visited during the search process, and the Q queue stores the coordinates of the locations to be visited. The initial radius during the landing zone search process is R0, and the minimum safe landing area area is S min ; S44. The algorithm starts at coordinate P and increases the radius in sequence, searching for a location area at that coordinate that satisfies the landing point constraint. First, it determines whether there are any obstacles within the area, i.e., whether the constraint O(x′) is satisfied. After eliminating any dangerous areas, it determines whether the surrounding points have been visited. If they have not been visited and are not in queue Q, the point is added to the queue to be visited, reducing repeated point scanning, until the queue is empty. S45. In an area without obstacles, determine whether other constraints in the area near the point are met. If so, return the nearest landing point and area that meet the constraints; If not, calculate the safety factor using the Safe(·) function. Divide the image into four areas based on the center of the map. Compare the flat / dangerous area ratios to obtain the safety factors (s1, s2, s3, s4) for the four areas. Select the area with the highest safety factor as the movement direction. S47. The final output result is: the center coordinates of the safe landing area P fit (x f ,y f ), the terrain characteristic parameters of the maximum feasible area include the average roughness mr, the average undulation mf and the average slope ms, the optimal safe landing radius r of the area, and the effective coverage area S of the area fit 4r 2 .

Images