Dynamic digital imaging simulation method for space-based space small target complex scene

Abstract

The invention relates to a space-based space small target complex scene dynamic digital imaging simulation method, and relates to the technical field of space imaging simulation. The method comprises the following steps: designing a target imaging model according to simulation system parameters; pixel-level plane drawing is carried out on the target digital library, and a target contour mask of the target under the specific ground pixel resolution is obtained; mapping a fixed star in the simulation view field into the target digital image; generating a target digital image of a diversity space illumination background; sequentially superposing a plurality of transient noises on the original imaging image, and carrying out fixed noise fusion processing to generate a target digital image containing a plurality of noises; and generating a spatial imaging simulation image sequential sequence through comprehensive motion simulation. By designing the target digital library construction method based on the two-dimensional shape feature point description method and the imaging analysis model of the target digital library construction method, general digital modeling can be carried out on various space targets, and the imaging process simulation of a space camera is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of space imaging simulation, and in particular to a method for simulating dynamic digital imaging of small space targets and complex scenes based on space. Background Art

[0002] In recent years, the number of spacecraft entering orbit has shown an upward trend, especially after the widespread application of small satellite technology, the number of spacecraft entering orbit has increased significantly. As a result, the space orbit occupancy rate has increased, the number of space debris and abandoned satellites has increased, and the risk of spacecraft collision in orbit has increased, posing a serious threat to the safe operation of normal satellites in orbit. Therefore, it is urgent for satellites and other spacecraft to have the ability to automatically perceive surrounding space targets in orbit, and image-based space target detection technology has great engineering application value.

[0003] Space target detection technology includes two technical approaches: artificial intelligence and traditional image processing. Both technical development and research rely heavily on the image data to be detected. In particular, artificial intelligence methods based on deep learning require a large amount of image data input to participate in model training. Since it is difficult to obtain space-based imaging images, and most of the image data of space-based observers involve certain military sensitivity and are not easy to be made public, there is a severe shortage of space-based space target imaging data sets that can be used for the design and verification of space target monitoring systems, which limits the optimization and development of detection algorithms for practical application needs.

[0004] Space target imaging simulation technology is one of the best ways to provide images to be inspected, and is often used for ground-side verification of equipment and its algorithms in technical research and engineering implementation. After years of technical research, the design framework of space target imaging simulation has gradually become clear, and the simulation effect is closer to the actual imaging. However, the existing simulation images still have problems such as single imaging simulation elements, limited imaging background, and insufficient dynamics of imaging simulation, and lack the ability to test the robustness of target detection algorithms. Summary of the invention

[0005] The present invention aims to solve the technical problems in the prior art and provide a method for simulating dynamic digital imaging of small space targets and complex scenes.

[0006] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0007] A space-based small target complex scene dynamic digital imaging simulation method comprises the following steps:

[0008] Step 1: Design the target imaging model according to the simulation system parameters, realize the theoretical mapping of the object position and brightness information in the three-dimensional coordinate space to the two-dimensional coordinate space of the image, and simulate the imaging process of the camera optical system;

[0009] Step 2: construct a target imaging digital library through a two-dimensional shape feature point description method, and use the target data imaging analytical model to perform pixel-level plane mapping on the target imaging digital library to obtain the target contour mask at a specific ground pixel resolution; then input it into the target imaging model to simulate the space camera to image the target and generate a target digital image;

[0010] Step 3: Select the Hipparcos star catalog as the data source for the space background star simulation, and map the stars in the simulated field of view to the target digital image through mutual conversion among the inertial coordinate system, image coordinate system, and pixel coordinate system, and obtain the real star data in the camera field of view. Finally, simulate the imaging of the star target in the field of view through the target imaging model, and add sub-pixel modeling to achieve an imaging effect that is more in line with the real position of the star point target.

[0011] Step 4: Generate a target digital image with a diverse spatial illumination background by using simulation modeling of a background with slowly changing edges, a background with steeply changing edges, and a background with linear non-uniform weak light;

[0012] Step 5: Multiple transient noises are sequentially superimposed on the original image, and then fixed noise fusion processing is performed to generate a target digital image containing multiple noises;

[0013] Step 6: Generate a spatial imaging simulation image time sequence through comprehensive motion simulation.

[0014] In the above technical solution, the transient noise in step five includes: readout noise, photon shot noise, dark current noise, hot pixel and cosmic ray noise.

[0015] In the above technical solution, the spatial imaging simulation image time series in step six includes: target, star background, background illumination and noise, and motion elements.

[0016] In the above technical solution, the target imaging model design in step 1 includes the following steps:

[0017] Based on the physical imaging characteristics and parameters of the space camera, a theoretical simulation of the space camera imaging process that describes the target size and brightness characteristics is established through imaging resolution calculation, coordinate system conversion, pixel gray value calculation, and imaging diffusion operation.

[0018] In the above technical solution, the two-dimensional shape feature point description method in step 2 includes the following steps:

[0019] Decompose the rough outline of the object in the view into several basic shapes;

[0020] For each basic shape, a feature description matrix containing real size, position, shape and grouping information is designed;

[0021] Comprehensively construct a digital library of target imaging as target view profiles.

[0022] In the above technical solution, in step 2, the target data imaging parsing model performs pixel-level plane mapping on the target digital library to obtain a target contour mask at a specific ground pixel resolution, which specifically includes the following steps:

[0023] When parsing a target digital library, first parse the first row of the target digital library with serial number 1 to obtain the width and height of the target imaging container rectangle;

[0024] Calculate the coordinates of all points of other serial numbers in the target digital library to achieve the conversion from the feature point description coordinate system to the imaging container pixel coordinate system;

[0025] According to the attribute values ​​in the target digital library, the minimum traversal range of each attribute in the pixel coordinate system of the imaging container is determined;

[0026] Determine the mask generation criteria for each attribute, and finally obtain the target contour mask at a specific pixel resolution GSD.

[0027] In the above technical solution, the comprehensive motion simulation in step 6 includes the following steps:

[0028] The target position and overall grayscale of the target, various types of background illumination change parameters, transient and special spatial working condition noise parameters in the target imaging model are calculated and modeled in frame units. The motion models of target curve displacement, grayscale dynamic change, imaging tailing and platform jitter are added. The state information of image simulation elements is updated according to the simulation timestamp to construct dynamic digital images of space targets under different working conditions.

[0029] The present invention has the following beneficial effects:

[0030] The space-based small target complex scene dynamic digital imaging simulation method of the present invention can perform general digital modeling of various space targets by designing a target digital library construction method based on a two-dimensional shape feature point description method and its imaging analysis model; then, through imaging simulation methods such as target digital imaging, field of view star selection, imaging coordinate conversion, point target sub-pixel modeling, etc., the imaging process simulation of the space camera is realized.

[0031] The space-based small target complex scene dynamic digital imaging simulation method of the present invention constructs a simulation library of space complex lighting scenes by modeling a variety of lighting backgrounds and noises in the space imaging environment, which can meet the robustness test of the space target detection algorithm for multiple scenes.

[0032] The digital imaging simulation system established by the dynamic digital imaging simulation method of small space targets and complex scenes in the space-based space of the present invention includes modeling of target dynamic grayscale, curve motion, imaging tailing, and imaging platform jitter. The simulation system framework design is extensible and has good compatibility with the orbital coordinate model and target texture imaging model that may be connected subsequently.

[0033] The space digital imaging model and the designed simulation system established by the space-based space small target complex scene dynamic digital imaging simulation method of the present invention are dynamically adjustable, and dynamic digital images of space targets under different working conditions can be generated by setting the imaging simulation parameters. It can be used for ground-end development and verification of traditional space target detection algorithms, and can also provide data set support for network model training of artificial intelligence detection algorithms such as deep learning. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0035] Figure 1 The present invention is a design flow chart of the method for simulating dynamic digital imaging of small space targets and complex scenes in space.

[0036] Figure 2 It is a typical satellite orthographic projection rough outline decomposition diagram constructed by the two-dimensional shape feature point description method in the space-based space small target complex scene dynamic digital imaging simulation method of the present invention.

[0037] Figure 3 The target mask schematic diagram is obtained by calculating the target data imaging analytical model in the space-based small target complex scene dynamic digital imaging simulation method of the present invention.

[0038] Figure 4 It is a space target imaging simulation image obtained by integrating the target mask and the imaging model in the space-based space small target complex scene dynamic digital imaging simulation method of the present invention.

[0039] Figure 5 It is a digital image obtained by sub-pixel level modeling of point targets in the space-based small target complex scene dynamic digital imaging simulation method of the present invention.

[0040] Figure 6 It is an image obtained by modeling multiple space illumination backgrounds in the space-based small target complex scene dynamic digital imaging simulation method of the present invention.

[0041] Figure 7 It is a target digital image obtained by using multiple space imaging noise simulation modeling in the space-based small target complex scene dynamic digital imaging simulation method of the present invention.

[0042] Figure 8 It is a comprehensive imaging simulation image of a space target with a uniform weak light background output by a digital imaging simulation system established by the dynamic digital imaging simulation method for small space targets in complex scenes based on the present invention.

[0043] Fig. 9 The digital imaging simulation system established by the method for dynamic digital imaging simulation of small space targets and complex scenes based on the present invention outputs a simulated image of a space target with a background of slowly varying light spots. DETAILED DESCRIPTION

[0044] The inventive concept of the present invention is:

[0045] The space-based small target complex scene dynamic digital imaging simulation method of the present invention simulates the imaging effect of the space-based observation platform in the real space scene to the greatest extent, and provides data input for the design of space target monitoring system and detection algorithm research for practical applications.

[0046] The present invention conducts a simulation study on space target imaging for complex lighting scenes in space, using a visible light camera as the simulation object and a dynamic imaging image of a small target in a complex lighting space as the simulation target. A space small target imaging modeling method based on a two-dimensional shape feature point description method and an imaging analysis model is proposed; a high-precision background star imaging simulation is achieved through sub-pixel point target grayscale imaging position modeling; a variety of dynamic lighting background models are established and a noise calculation model is introduced to generate digital simulation sequence images of dynamic changes in the local signal-to-noise ratio of the target. In the comprehensive motion simulation, model designs such as target curve displacement, dynamic grayscale changes, imaging tailing and platform jitter are added to simulate the imaging effect of the space-based observation platform in the real space scene to the greatest extent, providing data input for the design of space target monitoring systems and detection algorithm research for practical applications.

[0047] Step 1: Design the target imaging model according to the simulation system parameters, realize the theoretical mapping of the object position and brightness information in the three-dimensional coordinate space to the two-dimensional coordinate space of the image, and simulate the imaging process of the camera optical system;

[0048] Step 2: Build a target imaging digital library through the two-dimensional shape feature point description method, use the target data imaging analysis model to perform pixel-level plane mapping on the target imaging digital library, and obtain the target contour mask at a specific ground pixel resolution; then input it into the target imaging model to simulate the space camera to image the target and generate a target digital image.

[0049] Step 3: Select the Hipparcos star catalog (Hipparcos-2) as the data source for space background star simulation. After mutual conversion among the inertial coordinate system, image coordinate system, and pixel coordinate system, map the stars in the simulated field of view to the target digital image, obtain the real star data in the camera field of view, and finally simulate the imaging of the star targets in the field of view through the target imaging model, and add sub-pixel modeling to achieve an imaging effect that is more in line with the actual position of the star point target.

[0050] Step 4: Generate a target digital image with a diverse spatial illumination background through simulation modeling of a background with slowly changing edges, a background with steeply changing edges, and a background with linear non-uniform weak light.

[0051] Step 5: Multiple transient noises are sequentially superimposed on the original image, and then fixed noise fusion processing is performed to generate a target digital image containing multiple noises.

[0052] Step 6: Generate a time sequence of space imaging simulation images through comprehensive motion simulation. The generated time sequence of space imaging simulation images includes: target, star background, background illumination and noise, and motion elements. The entire digital imaging simulation system is designed to be dynamically adjustable, and dynamic digital images of space targets under different working conditions can be generated by setting imaging simulation parameters.

[0053] The following is a detailed description of the space-based small target complex scene dynamic digital imaging simulation method of the present invention in conjunction with the accompanying drawings. Figure 1 As shown, the method comprises the following steps:

[0054] Step 1: Design the target imaging model based on the simulation system parameters, realize the theoretical mapping of the object position and brightness information in the three-dimensional coordinate space to the two-dimensional coordinate space of the image, and simulate the imaging process of the camera optical system.

[0055] The target imaging model in this step is based on the physical imaging characteristics and parameters of the space camera, describing the camera imaging process of the target size and brightness characteristics, and obtaining the target digital image transmitted by the simulated image sensor. The specific steps of the target imaging model design include: based on the physical imaging characteristics and parameters of the space camera, through imaging resolution calculation, coordinate system conversion, pixel gray value calculation, and imaging diffusion operation, a theoretical simulation of the space camera imaging process that describes the target size and brightness characteristics is established.

[0056] The imaging mapping of the target size is mainly related to the camera spatial resolution GSD: L is the distance from the target to the camera optical lens; P is the pixel size of the camera image sensor; f is the focal length of the camera;

[0057] According to the input target contour mask, a mapping model from target brightness to target imaging grayscale value is established. For small space targets at a long distance, the present invention uses magnitude to describe their brightness. Let the target apparent magnitude be m , the irradiance E of the target reaching the outside of the camera optical system m (kw / m 2 ) is: E m =E s 10 0.4(-26.73-m) , E s is the solar irradiance outside the Earth's atmosphere, corresponding to the apparent magnitude of the Sun at this time being -26.73;

[0058] Assume that the aperture of the optical system is D, the transmittance is τ, the camera simulation exposure time is T, and the total energy Q(J) generated by the target received by the camera optical system is:

[0059] Assume that the energy of a single photon is E ph for: Planck constant h = 6.6260693 × 10 -34 J·s, speed of light c = 3×10 8 m / s, the average wavelength of visible light is λ is 550nm. During the camera simulation exposure time T, the total number of photons received by the image sensor is N ph for:

[0060] Assume that the quantum efficiency of the image sensor at λ is η, and the gain of the image sensor is K(e - / DN), the total digital quantity D e for:

[0061] Where η represents the quantum efficiency of the image sensor, which is the ratio of the average number of photoelectrons generated per unit time to the number of incident photons at a specified wavelength.

[0062] At long distances, the digital image of a space target usually occupies only a few to dozens of pixel units, showing the characteristics of a small target. Therefore, the digital quantity generated by a single target pixel can be approximately represented by the average digital quantity. Assuming that the digital image of the target on the image plane has a total of n pixels, the digital quantity D generated by a single pixel is i for:

[0063] Due to the influence of circular aperture diffraction, the image of the target on the focal plane of the camera will produce a diffusion phenomenon. The present invention uses a two-dimensional Gaussian filter to perform a convolution operation on the original image to achieve the diffusion effect of optical system imaging.

[0064] Step 2: Build a target imaging digital library through the two-dimensional shape feature point description method, use the target data imaging analysis model to perform pixel-level plane mapping on the target digital library, and obtain the target contour mask at a specific ground pixel resolution; then input it into the target imaging model to simulate the space camera to image the target and generate a target digital image.

[0065] In this step, the space target is first described by two-dimensional shape feature point method to construct a target imaging digital library, and then the target data imaging analysis model is used to obtain the target contour mask at a specific ground pixel resolution. Finally, the target contour mask is used as input to generate a target digital image through the target imaging model.

[0066] In this step, the two-dimensional shape feature point description method includes the following steps:

[0067] The rough outline of the target in the view is decomposed into several basic shapes; a feature description matrix containing real size, position, shape and grouping information is designed for each basic shape; and a digital library of the target view outline is comprehensively constructed.

[0068] The description information of each feature point in the digital library is represented by a line, that is, each serial number points to a specific feature point information. The architecture of the digital library is described in Table 1.

[0069] Table 1 Digital library architecture based on two-dimensional shape feature point description

[0070]

[0071] The target data imaging parsing model performs pixel-level planar mapping on the target digital library, including the following steps:

[0072] When parsing a target digital library, first parse the first row of the target digital library with serial number 1 to obtain the width W and height H of the target imaging container rectangle;

[0073] Calculate the coordinates of all points of other serial numbers in the target digital library to achieve the conversion from the feature point description coordinate system to the imaging container pixel coordinate system;

[0074] According to the attribute values ​​in the target digital library, the minimum traversal range of each attribute in the pixel coordinate system of the imaging container is determined;

[0075] The mask generation criteria of each attribute are determined, and finally the target contour mask of the target at a specific pixel resolution GSD is obtained, that is, a two-dimensional pixel matrix with pixel values ​​of 1 in the target area and 0 in the rest.

[0076] In this step, the space target takes a typical satellite as an example, and its front view with obvious shape features is digitally modeled and designed. Figure 2As shown in Figure 2, the rough outline of the target in the view is decomposed into 9 basic shapes, such as Figure 2 As shown in the identification numbers 1-9 in the figure, there are 2 circles, 3 rectangles, and 4 parallelograms. According to Table 1, a feature description matrix containing real size, position, shape, and grouping information is established for each basic shape. Finally, a digital library of the target view outline is constructed. The description information of each feature point in the digital library is represented by a line, that is, each serial number points to a specific feature point information.

[0077] After obtaining the target digital library, parse the first row of the target digital library with serial number 1 to obtain the width W and height of the target imaging container rectangle H ;

[0078] Calculate the coordinates of all points of other serial numbers in the target digital library to achieve the conversion from the feature point description coordinate system to the imaging container pixel coordinate system. The mathematical model of the imaging container pixel coordinate coefficient is: and

[0079] Among them, C k , R k They are the column and row coordinate values ​​of the imaging container pixel coordinate system respectively; k ,y k are the XY coordinate values ​​of the feature points in the feature point description coordinate system; N offset is the edge offset value; r c It is the radius value of the basic circular shape in the feature point description coordinate system. UpInt(·) means rounding up.

[0080] In this embodiment, the attributes in the target digital library include circle, rectangle and parallelogram. The minimum traversal range of circle, rectangle and parallelogram in the imaging container pixel coordinate system can be determined according to the calculation formula. The mathematical model is:

[0081] Round

[0082] rectangle

[0083] Parallelogram

[0084] Then the mask generation criteria for circles, rectangles and parallelograms are determined:

[0085] Round

[0086] Rectangular T model (i,j)=1R min ≤i≤R max ,C min ≤j≤C max ,

[0087] Parallelogram

[0088] When GSD is 0.5m, we get Figure 2 The target digital library constructed is analyzed and obtained Figure 3 Target mask shown.

[0089] Comprehensive target mask and imaging model, set the target magnitude m They are 3rd and 5th respectively. When the distance L is 100km, the target digital image obtained by imaging simulation is as follows Figure 4 As shown. Among them, P is 2.5μm, f is 70mm, T is 10ms, η is 0.65, and K is 2e - / DN, τ is 0.8, D is 90mm, the image quantization level is 8bit, the Gaussian convolution kernel size is 5×5, and the W×H of the target imaging container rectangle is 58m×26m.

[0090] Step 3: Select the Hipparcos star catalog (Hipparcos-2) as the data source for space background star simulation. After mutual conversion among the inertial coordinate system, image coordinate system, and pixel coordinate system, map the stars in the simulated field of view to the target digital image, obtain the real star data in the camera field of view, and finally simulate the imaging of the star targets in the field of view through the target imaging model, and add sub-pixel modeling to achieve an imaging effect that is more in line with the actual position of the star point target.

[0091] In this step, the formula The spherical coordinates of the star chart represented by right ascension and declination are converted into rectangular coordinates represented by the XYZ axis, where the XY plane of the rectangular coordinate system coincides with the celestial equatorial plane, the X axis passes through the intersection of 0° right ascension and 0° declination (i.e. the vernal equinox), the Z axis points to the north celestial pole, and the Y axis, X axis and Z axis form a right-handed rectangular coordinate system, thereby obtaining the inertial coordinates representing the position of the star. And the camera optical axis is taken as Z C Axis, the center of the optical system is the origin O C Establish the simulation camera coordinate system O C -X C Y C Z C . The position coordinates [XYZ] in the inertial system are transformed by the rotation transformation matrix T Convert to the position coordinates [X C Y C Z C ] T , giving the star position in the camera coordinate system with the optical axis pointing downward.

[0092] Re-pass Convert the rectangular coordinates of the camera coordinate system to spherical coordinates. The diagonal field of view of the camera is: Where P is the pixel size of the image sensor, N×M is the number of pixels of the image sensor, and f is the focal length of the camera. According to θ'≤FOV / 2, the star data to be simulated and imaged within the field of view of the simulation camera is determined. The star coordinates in the camera coordinate system that satisfies θ'≤FOV / 2 are marked as [X C ',Y C ',Z C '] T , according to the conversion formula from camera coordinate system to image coordinate system The coordinates of stars in the rectangular area inscribed in the circular field of view of the simulated camera can be obtained in the image coordinate system [X I ,Y I ] data, and then through the formula Transform to obtain the star coordinates in the pixel coordinate system [X P ,Y P ]data.

[0093] Since the star is far away from the camera, it appears as a point in the image, and becomes a point light spot after being diffused by the camera optical system. When performing star imaging simulation modeling, the energy distribution function of the diffuse light spot can be approximated by the Gaussian distribution function. Assume that the star numerical coordinates are (9.9556, 10.1257), the total digital quantity generated is 1024, and the diffuse neighborhood size used in sub-pixel modeling is 7×7. Take the star coordinates [X P ,Y P ] data as the center point, and the center coordinates of all pixels in the diffuse neighborhood as the coordinates of the diffuse processing point. The pixel digital quantity data is obtained through the imaging model, and then the Gaussian distribution function is used for diffusion processing to achieve the following Figure 5 The sub-pixel modeling of star point targets is shown in FIG. The pixel positions in the neighborhood are described by pixel coordinates in the pixel coordinate system, and the pixel center coordinates and star coordinate data in the neighborhood are described by numerical coordinates in the pixel coordinate system.

[0094] Step 4: Generate digital images of diverse spatial lighting backgrounds through simulation modeling of slowly changing edge spot backgrounds, simulation modeling of steeply changing edge spot backgrounds, and simulation modeling of linear non-uniform weak light backgrounds.

[0095] In this step, the modeling of the light spot is based on a two-dimensional Gaussian function model. In order to simulate the grayscale of the background light spot slowly extending from the center to the periphery during spatial imaging, a mathematical model for the simulation imaging of a time-varying edge slowly varying light spot is established:

[0096]

[0097] In the formula, K G (t) is the time-varying total energy of the light spot; σ Gis the time-varying radius of the light spot; x0(t) and y0(t) are the center coordinates of the light spot that change with time.

[0098] In order to simulate the special illumination background with concentrated energy at the edge of the light spot and large grayscale gradient, a mathematical model of light spot simulation imaging with steep time variation is established:

[0099] G(x,y,t)=F(x,y,t)*w(s,τ)

[0100]

[0101] Where * represents convolution; w(s,τ) represents the Gaussian filter kernel, H gray (t) represents the gray value of the spot imaging that changes with time; H offset Bias gray value for image background.

[0102] To simulate the non-uniform weak light background, the gray gradient direction is set to be horizontal, and the initial gray value of the left edge of the image is set to be H L,0 , the initial gray value of the left edge is H R,0 , the mathematical model of time-varying linear non-uniform background is established as:

[0103]

[0104] In the formula, K L , K R They represent the coefficients of change of grayscale values ​​of the left and right edges of the image over time; M represents the number of image columns.

[0105] According to actual needs, you can generate Figure 6 Simulated digital images of various types of spatial lighting backgrounds, such as the background simulation image of the slowly changing light spot at the edge, the background simulation image of the suddenly changing light spot at the edge, and the background simulation image of the linear non-uniform weak light.

[0106] Step 5: The transient noises such as readout noise, photon shot noise, dark current noise, hot pixel and cosmic ray noise are superimposed on the original image in sequence, and then a digital image containing multiple noises is generated through fixed noise fusion processing.

[0107] The order of noise modeling superposition is: readout noise, photon shot noise, dark current noise, hot pixel and cosmic ray noise, fixed noise. The specific modeling steps are as follows:

[0108] Based on the camera grayscale bias, Get the biased electron number, establish the mean as the biased electron number, and the variance is Gaussian distribution probability model of pixel electron number

[0109]

[0110] Then by Simulate camera readout noise after converting grayscale values;

[0111] Calculate the average signal of each non-zero grayscale pixel in the original image at exposure time T (expressed in grayscale value), establish the mean and variance are Poisson distribution model Simulate photon shot noise;

[0112] Set n e (x,y), given by Established within the exposure time T, with n dark is the Poisson probability distribution model of mean and variance, substitute After conversion to grayscale values, dark current noise is simulated;

[0113] The hot pixel and cosmic ray noise models are represented by discrete impulse functions;

[0114] Fixed noise is noise with fixed spatial position. Pixel-distributed FPN is similar to hot pixel noise but with limited maximum grayscale. Row-column distributed FPN establishes position noise with random intensity in rows and columns respectively.

[0115] The additive noise is sequentially added to the original image, and the multiplicative noise is finally fused to generate the final noisy digital image.

[0116] Set the parameter values: K is 2e - / DN, image quantization level is 8bit, I offset Take 50 (i.e. 100e - ),σ read Take 20e - ; Calculated by imaging the original image Here we take 8.0957, that is Take 8.0957; camera exposure time T is 10ms, n e (x,y) average 500e - / pixel / s; the maximum number of FPN electrons is 10e - , using the row distribution type; the maximum PRNU is 2%. The final generated noise imaging simulation image is as follows Figure 7 As shown. Step 6, generate a time sequence of space imaging simulation images through comprehensive motion simulation. Specifically, a time sequence of space imaging simulation images containing elements such as target, star background, background illumination and noise, and motion is generated through comprehensive motion simulation. The entire digital imaging simulation system is designed to be dynamically adjustable, and dynamic digital images of space targets under different working conditions can be generated by setting imaging simulation parameters.

[0117] In this step, the comprehensive motion simulation includes the following steps: frame-based calculation modeling of all time-varying variables such as the target position and overall target grayscale in the target imaging model, various types of background illumination change parameters, transient and special space working condition noise parameters, etc., and motion models such as target curve displacement, grayscale dynamic change, imaging tailing and platform jitter are added. The image simulation element status information is updated according to the simulation timestamp, and dynamic digital images of space targets under different working conditions can be constructed.

[0118] Specifically, for target curve displacement modeling, the method of the present invention designs a target motion model based on increments of a specified target initial position.

[0119] Assume that the origin of the imaging container pixel coordinate system is at the initial position of the pixel coordinate system (x c ,y r ), the target moving speed is v T (unit: pixel / ms), the camera exposure time is T (unit: ms), then the coordinates (x(k), y(k)) of the target in the pixel coordinate system at the kth frame satisfy:

[0120]

[0121] Where x(0) = x c , y(0) = y c , θ(k) satisfies:

[0122] Where θ(k) represents the velocity angle of the target at the kth frame; w f is the target starting velocity angle; w e The velocity angle at the end of the target simulation; the velocity angle is in pixel coordinate system X P The axis is the starting point, and the clockwise direction around the coordinate origin is the increasing direction.

[0123] For the modeling of grayscale dynamic changes, according to the total digital quantity D of the target imaging e , let the target grayscale change range be A e (Unit: %), the frequency of change is w e (Unit: ° / frame), then the total digital quantity D of the target k-th frame imaging e (k) is:

[0124] For target motion tail modeling, the method of the present invention divides the target displacement in a single exposure into units of 1 pixel displacement to obtain the number of displacement pixel grids. Then, the number of displacement pixel grids is taken as the number of microframes, and 1 pixel / microframe is taken as the target movement speed, and microframe imaging modeling is performed on the target. Among them, the speed angle increment and target grayscale of each microframe are the average values ​​of the same amount of the frame for the number of microframes.

[0125] For imaging platform jitter modeling, when performing imaging simulation, set the edge extension value to expand the edge of the original image size (N*M). At this time, the imaging target resolution remains unchanged, only the field of view is enlarged, and the single frame image generated by substituting it into the imaging model is recorded as the cache frame I buff , set the image cutting window to the original image size (N*M) and record it as I out , use I out to I buff Slide along the frame sequence to take pictures, and the output image sequence will produce a jitter effect. out The origin of the pixel coordinate system is at I buff The pixel coordinates in (x d ,y d ), the maximum jitter offset range radius is l max (Unit: pixel), the jitter frequency is w d , the sliding image acquisition model is established as:

[0126]

[0127] The target, background stars, illumination background, noise and dynamic motion imaging model are integrated to simulate the star gaze mode imaging in the sky area where the camera optical axis points (declination 52.358°, right ascension 175.128°). The total imaging simulation time is T total The camera exposure time T is 10ms. Figure 8 Target magnitude shown m is level 7, the distance L is 500km, the background illumination is set to uniform weak light, and the target moving speed is v T These are the 100th, 300th, 500th, 700th, and 900th imaging simulation images at 0.08 pixel / ms. Fig. 9 The imaging simulation image is shown with the illumination background set to a background strong light edge slowly changing spot. Among them, P is 2.5μm, f is 70mm, N is 1024, M is 2048, and the other imaging simulation parameters are the same as the previous settings.

[0128] The space-based small target complex scene dynamic digital imaging simulation method of the present invention can perform general digital modeling of various space targets by designing a target digital library construction method based on a two-dimensional shape feature point description method and its imaging analysis model; then, through imaging simulation methods such as target digital imaging, field of view star selection, imaging coordinate conversion, point target sub-pixel modeling, etc., the imaging process simulation of the space camera is realized.

[0129] The space-based small target complex scene dynamic digital imaging simulation method of the present invention constructs a simulation library of space complex lighting scenes by modeling a variety of lighting backgrounds and noises in the space imaging environment, which can meet the robustness test of the space target detection algorithm for multiple scenes.

[0130] The digital imaging simulation system established by the dynamic digital imaging simulation method of small space targets and complex scenes in the space-based space of the present invention includes modeling of target dynamic grayscale, curve motion, imaging tailing, and imaging platform jitter. The simulation system framework design is extensible and has good compatibility with the orbital coordinate model and target texture imaging model that may be connected subsequently.

[0131] The space digital imaging model and the designed simulation system established by the space-based space small target complex scene dynamic digital imaging simulation method of the present invention are dynamically adjustable, and dynamic digital images of space targets under different working conditions can be generated by setting the imaging simulation parameters. It can be used for ground-end development and verification of traditional space target detection algorithms, and can also provide data set support for network model training of artificial intelligence detection algorithms such as deep learning.

[0132] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A space-based small target complex scene dynamic digital imaging simulation method, characterized in that: The steps include: Step 1: Design the target imaging model according to the simulation system parameters, realize the theoretical mapping of the object position and brightness information in the three-dimensional coordinate space to the two-dimensional coordinate space of the image, and simulate the imaging process of the camera optical system; Step 2: construct a target imaging digital library through a two-dimensional shape feature point description method, and use the target data imaging analytical model to perform pixel-level plane mapping on the target imaging digital library to obtain the target contour mask at a specific ground pixel resolution; then input it into the target imaging model to simulate the space camera to image the target and generate a target digital image; Step 3: Select the Hipparcos star catalog as the data source for the space background star simulation, and map the stars in the simulated field of view to the target digital image through mutual conversion among the inertial coordinate system, image coordinate system, and pixel coordinate system, and obtain the real star data in the camera field of view. Finally, simulate the imaging of the star target in the field of view through the target imaging model, and add sub-pixel modeling to achieve an imaging effect that is more in line with the real position of the star point target. Step 4: Generate a target digital image with a diverse spatial illumination background by using simulation modeling of a background with slowly changing edges, a background with steeply changing edges, and a background with linear non-uniform weak light; Step 5: Multiple transient noises are sequentially superimposed on the original image, and then fixed noise fusion processing is performed to generate a target digital image containing multiple noises; Step 6: Generate a spatial imaging simulation image time sequence through comprehensive motion simulation.

2. The space-based small target complex scene dynamic digital imaging simulation method according to claim 1 is characterized in that: The transient noise in step 5 includes: readout noise, photon shot noise, dark current noise, hot pixel and cosmic ray noise.

3. The space-based small target complex scene dynamic digital imaging simulation method according to claim 1 is characterized in that: The spatial imaging simulation image time series in step six includes: target, star background, background illumination and noise, and motion elements.

4. The space-based small target complex scene dynamic digital imaging simulation method according to any one of claims 1 to 3, characterized in that: Designing the target imaging model in step 1 includes the following steps: Based on the physical imaging characteristics and parameters of the space camera, a theoretical simulation of the space camera imaging process that describes the target size and brightness characteristics is established through imaging resolution calculation, coordinate system conversion, pixel gray value calculation, and imaging diffusion operation.

5. The space-based small target complex scene dynamic digital imaging simulation method according to any one of claims 1 to 3, characterized in that: The two-dimensional shape feature point description method in step 2 includes the following steps: Decompose the rough outline of the object in the view into several basic shapes; For each basic shape, a feature description matrix containing real size, position, shape and grouping information is designed; Comprehensively construct a digital library of target imaging as target view profiles.

6. The space-based small target complex scene dynamic digital imaging simulation method according to any one of claims 1 to 3, characterized in that: In step 2, the target data imaging parsing model performs pixel-level plane mapping on the target digital library to obtain the target contour mask at a specific ground pixel resolution, which specifically includes the following steps: When parsing a target digital library, first parse the first row of the target digital library with serial number 1 to obtain the width and height of the target imaging container rectangle; Calculate the coordinates of all points of other serial numbers in the target digital library to achieve the conversion from the feature point description coordinate system to the imaging container pixel coordinate system; According to the attribute values ​​in the target digital library, the minimum traversal range of each attribute in the pixel coordinate system of the imaging container is determined; Determine the mask generation criteria for each attribute, and finally obtain the target contour mask at a specific pixel resolution GSD.

7. The space-based small target complex scene dynamic digital imaging simulation method according to any one of claims 1 to 3, characterized in that: The comprehensive motion simulation in step six includes the following steps: The target position and overall grayscale of the target, various types of background illumination change parameters, transient and special spatial working condition noise parameters in the target imaging model are calculated and modeled in frame units. The motion models of target curve displacement, grayscale dynamic change, imaging tailing and platform jitter are added. The state information of image simulation elements is updated according to the simulation timestamp to construct dynamic digital images of space targets under different working conditions.

Images