Navigation information extraction method and system for optical autonomous navigation

Abstract

The application relates to a navigation information extraction method and system for optical autonomous navigation. The navigation information extraction method comprises the following steps: S1. edge detection is performed on a celestial body target image; S2. celestial body texture suppression is performed based on the edge detection result; and S3. a least square circle fitting method is used to extract a celestial body center position as navigation information. The application provides a center position information extraction method for a rich-texture celestial body target, can solve the false edges such as the morning and evening twilight lines and textures in the image processing process of the rich-texture celestial body target, is suitable for non-complete celestial body target images, has a high edge detection rate, has good edge continuity, and has a good contour fitting effect on a small-arc celestial body target.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of autonomous optical navigation, in particular to a navigation information extraction method and system for rich-texture celestial target and incomplete celestial image. BACKGROUND

[0002] Autonomous navigation of a spacecraft independent of ground stations and other equipment is an important trend in deep space exploration, and optical navigation technology for obtaining navigation information from celestial targets is an important part of autonomous navigation. The most common method for obtaining navigation information from celestial target images is to use a navigation camera to obtain the position of the center of a celestial target. The working principle is to use a navigation camera to take a celestial target image at a certain time, perform edge extraction, and then perform texture suppression on some rich-texture celestial target images to obtain the outline of the celestial target and calculate the center of the celestial target to complete the extraction of navigation information. With the development of deep space exploration in depth and breadth, many planar celestial targets can be used as objects for navigation information extraction. Because the surface texture of each celestial target is different, some celestial targets have rich texture information such as meteor craters and ring-shaped craters, and some celestial targets have cloud layers and other celestial features that interfere with navigation information extraction, increasing the difficulty of texture suppression. At different stages of spacecraft exploration, the captured images may contain complete clear outlines or only partial arcs of the celestial target, which directly affects the accuracy and success rate of navigation information extraction. The traditional navigation information extraction method cannot completely suppress rich texture or cannot be used for incomplete celestial arcs, which puts forward higher requirements for the navigation information extraction algorithm of the optical sensor in deep space exploration.

[0003] How to overcome the shortcomings of the above-mentioned prior art has become a topic that needs to be solved in this technical field. SUMMARY

[0004] To overcome the shortcomings of the prior art, the present application provides a navigation information extraction method, system and computer storage medium for optical autonomous navigation, which specifically adopts the following technical solutions.

[0005] A navigation information extraction method for optical autonomous navigation, the method comprising the following steps:

[0006] S1. performing edge detection on a celestial target image;

[0007] S2. performing celestial texture suppression based on the edge detection result;

[0008] S3. using a least squares circle fitting method to extract the center position of the celestial target as navigation information.

[0009] Further, the step S1 comprises:

[0010] S11. Smooth the image using Gaussian adaptive median filtering;

[0011] S12. Calculate the gradient magnitude and gradient direction for each pixel.

[0012] S13. Nonmaximum suppression;

[0013] S14. Improved threshold selection for weak edge detection.

[0014] Furthermore, step S2 includes:

[0015] S21. Perform texture removal based on edge detection results;

[0016] S22. Rich texture suppression.

[0017] Furthermore, step S3 includes:

[0018] S31. Define the conversion relationship between the standard equation and the general equation of a circle;

[0019] S32. Define the performance index function;

[0020] S33. Introduce the optimal expression form and solve for the optimal fitting result.

[0021] Furthermore, step S12 specifically includes calculating the gradient magnitude and gradient direction of each pixel according to the following formula:

[0022] (1)

[0023] In the formula, The x-axis of the celestial image. The vertical axis of the celestial image. for Orientation gradient, for Directional gradient, For gradient magnitude, The gradient direction;

[0024] Furthermore, step S14 specifically includes:

[0025] Get two thresholds and ,have less than the threshold The image part is Class, at the threshold and The image portion between is Class, greater than the threshold The image part is kind;

[0026] The class, class, class, respectively, are expressed as:

[0027]

[0028]

[0029] (2)

[0030] wherein, and are the current two thresholds, is the probability that the pixel gray level is , is the probability that the pixel is classified into class, is the probability that the pixel is classified into class, is the probability that the pixel is classified into class;

[0031] The class, class, class, respectively, are expressed as:

[0032]

[0033]

[0034] (3)

[0035] wherein, is the average gray value of class, is the average gray value of class, is the average gray value of class, is the probability that the pixel gray level is in class, is the probability that the pixel gray level is in class, is the probability that the pixel gray level is in class;

[0036] The average gray value of the whole image is:

[0037] (4)

[0038] ​In the formula, is the average gray value of the whole image;

[0039] The double-threshold inter-class variance is defined as:

[0040] (5)

[0041] In the formula, is the average gray value of the whole image;

[0042] The gray level at which the double-threshold inter-class variance is the maximum is the low threshold value of the image, and is the high threshold value of the image. , respectively, as the low threshold value and the high threshold value of the image;

[0043] The pixels between the low threshold value and the high threshold value are marked as weak edges, specifically including: after the double threshold is determined, the pixels greater than the high threshold value are marked as strong edges and set to 1, the pixels less than the low threshold value are marked as non-edges and set to 0, and the pixels between the low threshold value and the high threshold value are marked as weak edges, that is, suspected edges, so as to perform weak edge detection.

[0044] The isolated weak edges are suppressed, and isolation detection is performed, specifically including: judging whether the eight pixels around the pixel are in the same connected domain as the strong edges, if in the same connected domain, the point is retained; if not in the same connected domain, the point is removed.

[0045] Further, the step S21 specifically includes:

[0046] S211. Extracting the edge detection preprocessing result of the step S1;

[0047] S212. Extracting the skeleton of the preprocessed image and removing small structures;

[0048] S213. Removing track burrs;

[0049] S214. Marking the image to divide the connected domain.

[0050] Further, the step S22 specifically includes:

[0051] S221. Extracting the single connected domain marked in sequence;

[0052] S222. Calculating the tangent angle by sliding coordinates to perform track screening;

[0053] S223. Calculating the mean sequence variance of the track curvature;

[0054] S224. According to the minimum mean sequence variance of the real edge track curvature as the criterion, suppressing the texture. ​

[0055] The present application also relates to a navigation information extraction system using the navigation information extraction method as described above, the navigation information extraction system comprising:

[0056] an edge detection module for performing edge detection on the celestial target image;

[0057] a texture suppression module for performing celestial texture suppression based on the edge detection result;

[0058] a navigation information module for extracting the celestial center position as navigation information using a least square circle fitting method.

[0059] The present application also relates to a computer readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the navigation information extraction method as described above.

[0060] The technical scheme provided by the present application has the beneficial effects that: a center position information extraction method for rich-texture celestial target is provided, which can solve the problems of false edges such as the morning / evening twilight line and texture in the image processing of rich-texture celestial target, is suitable for non-complete celestial target images, has high edge detection rate and good edge continuity, and has good contour fitting effect on small-arc celestial target images. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 An example celestial image and its three-dimensional gray scale image of the present application.

[0062] Figure 2 A flowchart of the navigation information extraction method for optical autonomous navigation of the present application.

[0063] Figure 3 An example celestial target gray scale image and its edge detection result image of the present application.

[0064] Figure 4 A flowchart of the celestial texture suppression based on the edge detection result of the present application.

[0065] Figure 5 An index and tangent relationship transformation diagram of the present application.

[0066] Figure 6 A tangent angle transformation diagram of the present application.

[0067] Figure 7 A transformation process diagram of the celestial texture suppression based on the edge detection result of the present application.

[0068] Figure 8 A comparison diagram of the simulation image detection result of the present application and the canny operator detection.

[0069] Figure 9The figure is a schematic diagram of the navigation information extraction of the application compared with the Canny operator detection.

[0070] Figure 10 The figure is a schematic diagram of the edge detection rate of the application. DETAILED DESCRIPTION

[0071] The application will be further described below in conjunction with the accompanying drawings. The following examples are only used to more clearly illustrate the technical solutions of the application, and cannot be used to limit the protection scope of the application. It should be pointed out that the following detailed description is exemplary, and is intended to provide further description of the application.

[0072] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It must be noted that as used herein the terms are intended to be interpreted as being descriptive only and are not intended to be limiting subject matter according to the exemplary embodiments of the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising", when used in this specification, specify the presence of stated features, steps, operations, devices, components and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, devices, components, and / or groups thereof.

[0073] The specific embodiments of the application relate to a navigation information extraction method for optical autonomous navigation, which is directed to the inherent characteristics of complete celestial target images and incomplete celestial arc segments captured during the capture phase and the orbit phase of deep space exploration, and provides a navigation information extraction method capable of facing rich-textured celestial targets and incomplete celestial arc segments. The inherent characteristics include rich texture information such as craters, ring-shaped craters, and celestial twilight lines. Some celestial bodies also have cloud layers and other celestial features that interfere with navigation information extraction. Image processing is performed on the celestial target images. Due to the high gradient amplitude of the texture on the surface of the celestial body, the gradient amplitude of the celestial twilight line is lower than the real edge and higher than the deep space background. An example of a celestial image and its three-dimensional gray-scale image is shown in FIGS. 1 and 2. Figure 1 (a), and FIG. 2 Figure 1 (b). The required real edge of the celestial target cannot be directly obtained through edge detection only. The navigation information extraction method for optical autonomous navigation of the application is shown in FIG. 3, which includes the following steps: Figure 2

[0074] S1. Edge detection is performed on the celestial target image.

[0075] S2. Celestial texture suppression is performed based on the edge detection result.

[0076] S3. The least square circle fitting method is used to extract the center position of the celestial body as navigation information.

[0077] ​Regarding the step S1. edge detection on the celestial object image, according to the inherent characteristics of the celestial object image, the most commonly used canny edge detection operator in the improved engineering is used for edge detection, which can preliminarily eliminate the celestial surface texture with low gradient amplitude and segment the terminator. The step S1. edge detection on the celestial object image comprises the following steps:

[0078] S11. Smooth the image using a Gaussian-adaptive median filtering method. The Gaussian filter performs preliminary smoothing, and the adaptive median filter dynamically changes the filter size according to the preset conditions, selects an appropriate filter window for filtering, and takes into account the filtering of noise and the preservation of details.

[0079] S12. Calculate the gradient amplitude and gradient direction of each pixel according to the following formula:

[0080] (1)

[0081] In the formula, is the abscissa of the celestial image, is the ordinate of the celestial image, is the direction gradient, is the direction gradient, is the gradient amplitude, is the gradient direction.

[0082] S13. Non-maximum suppression. The gradient direction is used as the premise of non-maximum suppression, and the point with the maximum local gradient amplitude, i.e. the maximum value of the gradient amplitude, is saved.

[0083] S14. Weak edge detection with improved threshold selection. In the weak edge detection process, to solve the problem of retaining edge pixels due to local gradient value changes caused by noise and color changes, the celestial surface texture and the terminator are preliminarily suppressed, and the edge of the celestial region that is not directly illuminated by the sun but is still illuminated due to light scattering is extracted at the junction of the terminator. Filter the edge pixels with a threshold value and retain the edge pixels with high gradient values. Based on the central idea of the traditional Otsu algorithm for segmenting the target body and the background, the maximum inter-class variance method (Otsu algorithm) is improved for double threshold selection. Specifically, the following steps are included:

[0084] Divide the image into three parts, now assume that the two thresholds are and , there are , called the image part less than the threshold is class, called the image part between the threshold and is class, called the image part greater than the threshold the image portion of the image is classified as the class.

[0085] the class, class, class are respectively represented as:

[0086]

[0087]

[0088] (2)

[0089] wherein, and are the current two thresholds, is the probability that the pixel gray level is , is the probability that the pixel is classified into class, is the probability that the pixel is classified into class, is the probability that the pixel is classified into class.

[0090] the class, class, class are respectively represented as:

[0091]

[0092] (3)

[0093] wherein, is the average gray value of class, is the average gray value of class, is the average gray value of class, is the probability that the pixel gray level is in class, is the probability that the pixel gray level is in class, is the probability that the pixel gray level is in class.

[0094] the average gray value of the whole image is:

[0095] ​(4)

[0096] wherein, is the average gray value of the whole image.

[0097] The inter-class variance is defined as:

[0098] (5)

[0099] wherein, is the average gray value of the whole image.

[0100] The above formula shows that the inter-class variance is a measure of the separability between classes, the greater the inter-class variance, the higher the degree of distinction, so the gray level is taken when the inter-class variance of the above formula is maximum The optimal gray threshold of the image is recorded as , respectively, as the low threshold and high threshold of the image.

[0101] In the above double threshold selection process, the pixels between the high and low thresholds are marked as weak edges, that is, suspected edges, so as to perform weak edge detection, and isolated weak edges are suppressed, and isolation detection is performed, specifically including marking the pixels greater than the high threshold as strong edges and setting 1, marking the pixels less than the low threshold as non-edges and setting 0, and marking the pixels between the high and low thresholds as weak edges, that is, suspected edges, so as to perform weak edge detection. Isolated weak edges are suppressed, and isolation detection is performed, and it is judged whether the eight pixels around the pixel are in the same connected domain as the strong edge, if they are in the same connected domain, the point is retained; if they are not in the same connected domain, the point is removed. Thus, the weak edge detection result, that is, the celestial object image edge detection result, is obtained. As shown in Figs. 2(a) and 2(b). Figure 3 (a) is a celestial object gray scale image original drawing, and Fig. 2(b) is an edge detection result of the present application. Figure 3

[0102] Regarding the step S2. performing celestial texture suppression based on the edge detection result, the arrangement of the real celestial object edge pixels after edge detection presents a regular situation, the real celestial edge is relatively smooth, and the celestial surface texture presents an irregular pixel arrangement mode. Therefore, based on the curvature factor, a constraint condition for real celestial object edge detection is increased, and the contour edge, surface texture and noise are screened and distinguished. As shown in Figs. 3(a) and 3(b). Figure 4 , the step S2. performing celestial texture suppression based on the edge detection result, includes the following steps:

[0103] S21. performing texture removal based on the edge detection result.

[0104] First, the skeleton of the object is extracted, and the pre-processing result is marked as a matrix ​​​​. From the target to the target center of the outer corrosion refinement, until no longer change (single layer pixel width), the extraction of the target in the image on the center of the pixel profile, the results recorded as a matrix .

[0105] Due to the small structure of the small and closed curve, or small and short detection results, which are collectively referred to as small structures, statistics connected domain, for the connected domain less than 15 pixels of the object is deleted, and the Extract all connected domains and mark as , where is a variable, the maximum value is There are connected domain number, if means The first connected domain in the new matrix is extracted. The connected domain of the celestial surface texture bifurcation point, endpoint, in the processing process, the Using ring-shaped template convolution to get the convolution result .

[0106] (6)

[0107] In the formula, is the current coordinate of the matrix coordinate.

[0108] The end point of the connected domain is marked, because the connected domain end point is adjacent to only one point, the end point of the connected point is , the end point is stored, if there are too many end points on a connected domain, it means that the connected domain has many burrs, which is the trace of the celestial surface texture, then remove it, and mark the processed image as matrix .

[0109] The bifurcation point in a single connected domain is extracted, the remaining small isolated connected domain is removed, and finally the bifurcation point in the connected domain is restored to connect, and all connected domains are summarized to a storage space, the bifurcation point is recorded as a multi-connected domain, and the image is stored as matrix .

[0110] Secondary division of data, re-finding of image connected domain, and storage of each connected domain end point, single connected domain has two end points, the calculation of end point is still to Using ring-shaped template convolution, the result is marked as matrix .

[0111] The obtained connected domain end point and connected domain combination are stored. Each connected domain is recorded as a track , where is a variable, the maximum value is There are connected domain number, if Meaning Extract the first connected component to form a new matrix.

[0112] See appendix Figure 5 As shown, the trajectory is processed. Taking a single trajectory as an example, the pixel information of each point on the trajectory is found. Processing begins from the trajectory endpoint, and the current endpoint is recorded as the current point of the trajectory, which is also the first point. The stored information is set to 0, and then the 3x3 neighborhood of the current point is...

[0113] Find the index of the non-zero element within a 3x3 neighborhood. Search the trajectory by index, and denoted as the tangent direction from the current pixel to the next pixel on the trajectory. Store it.

[0114] See appendix Figure 6 As shown, the current point coordinates The sum of the coordinates of the next point on the trajectory and the tangent direction is the coordinates of the next point on the trajectory. In the following calculations, the position of the next point on the trajectory is recorded as the current position and the calculations are performed in a loop until the entire trajectory is traversed, so that the required information for the current trajectory can be obtained.

[0115] Because the arc segments on the surface of celestial bodies are almost perfectly smooth curves with gently changing tangents, the current connected domain... Inner Tangent line L1 at each point: With the fourth pixel separated by four pixels Tangent line L2 at each point: The angle between the tangents is:

[0116] (7)

[0117] Since there is a certain angle between the tangent directions of the real edge pixels of the celestial target, the angle is restricted. By using the sliding coordinate method, the tangents L1 and L2 between every two pixels of the trajectory are calculated, and the angle between the tangents of every two pixels is recorded. If the angle between the two tangents is greater than or equal to 90°, it means that the trajectory of the curve has changed abruptly, so the trajectory is discarded, and then the next trajectory is determined, until the traversal is completed.

[0118] S22. Rich Texture Suppression

[0119] This represents both the tangent direction at the current point and the first-order gradient calculation result at the current point, stored as an n x 2 matrix. Continuing the calculation, the forward difference yields the second-order gradient calculation result. Let the current point be the... One point, for No. OK The value of the column is then for the first connected component. The first and second difference results for each point are:

[0120] (8)

[0121] (9)

[0122] (10)

[0123] (11)

[0124] (12)

[0125] In the formula, For the current connected component, the first points First-order difference results in the direction, For the current connected component, the first points First-order difference results in the direction, For the current connected component, the first points Directional second-order difference results, For the current connected component, the first points Directional second-order difference results, For the current connected component, the first Curvature at each point.

[0126] Based on the curvature of the trajectory The trajectory is constrained by using a sliding coordinate system. The curvature of every five pixels along the trajectory is averaged sequentially. The pixel corresponding to the maximum average curvature of each trajectory is set to empty. The trajectory curvature is then recalculated to find the average curvature. variance of a sequence Mean curvature of the true edge trajectory of a celestial object The rate of change is gradual, while the mean curvature of the texture trajectory is... The variation is large, uneven, and prone to abrupt changes; the variance of the mean curvature of the surface texture trajectory of celestial bodies. The variance of the curvature at the edge of a real celestial object is much larger than the mean. The image of the actual celestial target has the smallest edge variance, and texture suppression is performed based on this criterion. Let the mean curvature sequence of the current trajectory be... common The matrix containing n elements stores the actual edges. .

[0127] (13)

[0128] (14)

[0129] Therefore, the true edge for:

[0130] (15)

[0131] In the formula, The mean of the trajectory curvature, For the first The mean and variance of the curvature series of the trajectories. For point The value at that location.

[0132] Appendix Figure 7 The transformation process of step S2 of the present invention is shown.

[0133] Regarding step S3, extracting the celestial body center position as navigation information using the least squares circle fitting method. To ensure that the fitting effect of the small arc segment celestial body meets the requirements for navigation information extraction, this invention abandons the previous least squares ellipse fitting method during the contour fitting process and adopts the least squares circle fitting method for extracting navigation information. Step S3, extracting the celestial body center position as navigation information using the least squares circle fitting method, specifically includes:

[0134] Let the coordinates of the center of the circle be... The radius of the circle is Then the standard equation of a circle is:

[0135] (16)

[0136] The general equation for transforming it into a circle has the following form:

[0137] (17)

[0138] Contains three unknowns , , There is a transformation relationship involved:

[0139] (18)

[0140] The performance index functions are defined as follows:

[0141] (19)

[0142] In the formula The parameters to be estimated are represented as those of a circle. Different edge detection points The observed value where the detection point is That is, the observation point is The model function of the circle to be estimated is

[0143] Therefore, the objective function is introduced The optimization solution expression form is as follows:

[0144] (20)

[0145] The meaning of the optimization solution expression form of the objective function is that The performance index function is minimized when the parameter value corresponding to the optimal estimation of the circle parameter p on which it depends is taken as the optimal fitting parameter, that is, the extreme point coordinates are taken as the optimal fitting result.

[0146] According to the definition of the extreme value, if the performance index function is to be minimized, it needs to be taken at the place where the first-order derivative of the function is zero, and there is:

[0147] (21)

[0148] Then, the value of the coefficient is determined by taking the minimum value of the objective function

[0149] The partial derivative value is:

[0150] (22)

[0151] Similarly, we have:

[0152] (23)

[0153] From this, the following equation set can be obtained:

[0154] (24)

[0155] If

[0156] (25)

[0157] Then the equation set can be simplified to the following form:

[0158] (26)

[0159] The equation is a three-element first-order non-homogeneous linear equation set, and the unique solution can be obtained, and the general equation parameter of the circle is:

[0160] (27)​

[0161] According to the equation, the standard equation of the circle can be calculated, and the center coordinates of the circle are obtained, which are the celestial body center coordinate positions and the circle radius is the celestial target image apparent radius information:

[0162] (28)

[0163] The navigation information extraction effect of the present application will be described below. The simulation image detection results of the present application are shown in Figs. 1 to 3. Figure 8 (a) shows the simulation image detection results of the present application, and Figs. 2 and 3 show the simulation image detection results of the present application. Figure 8 (b) shows the simulation image detection results of the present application.

[0164] The simulation image is quantitatively analyzed, and the fitting error of the simulation image is given, as shown in the following table:

[0165] Parameter error Method of the invention Center coordinate x0 error / pixel -0.8081 Center coordinate y0 error / pixel 0.7451 Center coordinate error / pixel 1.0992 LOS vector error / rad 1.9014*10 -4 ]]> LOS vector error / ° 0.0109

[0166] The simulation image detection results of the present application are shown in Figs. 1 to 3. Figure 9 (a) shows the celestial Celestia image, and Figs. 2 and 3 show the simulation image detection results of the present application. Figure 9 (b) shows the simulation image detection results of the present application. Figure 9 (c) shows the simulation image detection results of the present application.

[0167] The Celestia image is quantitatively analyzed, and the fitting error of the simulation image is given, as shown in the following table: Figure 9 (a) the fitting error of the celestial Celestia image, as shown in the following table:

[0168] Parameter error Method of the invention Center coordinate x0 error / pixel -0.3553 Center coordinate y0 error / pixel 0.4855 Center coordinate error / pixel 0.6016

[0169] The algorithm in this paper basically completes the detection of the edges of the celestial target. For the part of the celestial body that is not directly illuminated by sunlight but is illuminated by the scattering of sunlight, the edge detection performance is better. Therefore, even if the celestial image is photographed in a small arc segment during the shooting process, the non-intact celestial arc segment can be detected, and the fitting effect meets the requirements of navigation information extraction. For the simulation image, the real edge of the celestial target is recorded during the simulation process, and the real edge detection rate of the celestial target is obtained, as shown in Fig. 4. Figure 10 The average edge detection rate is 92.4996% (indicated by the horizontal line).

[0170] The starting point of the arc segment is marked as a starting point, and 30° is added as an arc segment fitting calculation error; from the starting point, every 0.5° is taken as a new starting point, and every time 30° is added, the corresponding arc segment fitting calculation error is added, until the entire arc segment is traversed. The arc segment fitting result is affected by the arc segment selection position. On the entire detected arc segment, although the fitting accuracies of different arc segments are different, they all meet the navigation requirements. The minimum value of the arc segment of the non-complete celestial body that can be detected for navigation is 30° of the central angle, and the corresponding arc segment percentage is 8.333%. The average error of the navigation information extraction is 2.2207 pixels.

[0171] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and the necessary general hardware platform. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the method described in each embodiment or some parts of the embodiments of the present application.

[0172] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device or system embodiments, since they are basically similar to the method embodiments, they are described more simply, and the related parts can be referred to the part of the method embodiments. The above-described device and system embodiments are only illustrative, and the units described as separate components can be or can not be physically separated, and the components displayed as units can be or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to the actual needs. Those skilled in the art can understand and implement it without creative labor.

[0173] As described above, the preferred embodiments of the present application are described, and it should be noted that those skilled in the art can make several improvements and modifications without departing from the technical principles of the present application, and these improvements and modifications should be considered as the protection scope of the present application.

Claims

1. A method for navigation information extraction for optical autonomous navigation, characterized in that, The method comprises the following steps: S1. performing edge detection on the celestial object image; S2. performing celestial texture suppression based on the edge detection result; S3. extracting the celestial center position as navigation information using a least square circle fitting method. The step S1 comprises: S11. smoothing the image using a Gaussian-adaptive median filtering method; S12. calculating the gradient amplitude and gradient direction of each pixel S13. non-maximum suppression; S14. improved weak edge detection of threshold selection; The step S2 comprises: S21. performing texture removal based on the edge detection result; S22. rich texture suppression; The step S3 comprises: S31. defining the conversion relationship between the standard equation and the general equation of a circle; S32. defining a performance index function; S33. introducing an optimal expression form and solving the optimal fitting result; The step S12 specifically comprises calculating the gradient amplitude and gradient direction of each pixel according to the following formula: （1） wherein is an image coordinate of the celestial body, is an image coordinate of the celestial body, is is a directional gradient, is is a directional gradient, is a gradient magnitude, is a gradient direction; The step S21 specifically comprises: S211. extracting the edge detection preprocessing result of the step S1; S212. extracting the skeleton from the preprocessed image and removing small structures; S213. removing trajectory burrs; S214. marking image division connected domains; The step S22 specifically comprises: S221. sequentially extracting the marked single connected domain; S222. sliding coordinate calculation of tangent angle for trajectory screening; S223. calculating the mean value series variance of trajectory curvature; S224. suppressing texture according to the criterion that the mean value series variance of the real edge trajectory curvature is the smallest.

2. A navigation information extraction system employing the navigation information extraction method according to claim 1, characterized by, The navigation information extraction system comprises: an edge detection module, configured to perform edge detection on the celestial object image; a texture suppression module, configured to perform celestial texture suppression based on the edge detection result; a navigation information module, configured to extract the celestial center position as navigation information using a least square circle fitting method.

3. A computer readable storage medium having stored thereon a computer program, characterized in that: The program is executed by the processor to implement the navigation information extraction method of claim 1.

Images