Spacecraft motion device and environment color interactive feedback acquisition and analysis method

Abstract

The invention provides a spacecraft motion device and environment color interactive feedback acquisition and analysis method, which relates to the technical field of spacecraft motion control and comprises the steps of acquiring environment multispectral color information, constructing a color feature mapping matrix, calculating a color gradient field and segmenting a feature region. Executing a quantum color feature analysis algorithm to generate an optimal motion track, obtaining motion parameters, executing a motion state decision algorithm to generate a control instruction, adjusting a motion state, and recording interactive feedback data; according to the invention, interactive feedback of the spacecraft motion device and the environment color can be realized, the motion control precision and the environment adaptability are improved, and a new technical scheme is provided for intelligent control of the spacecraft.

Description

Technical Field

[0001] The present invention relates to the technical field of spacecraft motion control, and in particular to a method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors. Background Art

[0002] Spacecraft motion devices need to interact with the surrounding environment in real time during mission execution, and obtain information through environmental perception to guide motion control. Among them, environmental color information, as an important perception data, can reflect environmental characteristics and changing laws. At present, commonly used environmental perception methods mainly include lidar scanning, visual image acquisition, etc. These methods play an important role in the environmental adaptability control of spacecraft motion devices.

[0003] However, the existing technology still has some shortcomings. The environmental perception method mainly focuses on geometric feature extraction, and the deep analysis of multispectral color information is insufficient, making it difficult to fully reflect the dynamic changing characteristics of the environment; the motion control strategy lacks effective connection with environmental characteristics and cannot achieve adaptive adjustment based on environmental changes; there is a lack of systematic interactive feedback mechanism between perception data and control instructions, resulting in a lag in the response of the motion device to environmental changes.

[0004] In summary, the present invention aims to solve the above-mentioned technical problems and proposes a spacecraft motion device control method based on multispectral color information analysis. By real-time acquisition and analysis of environmental color characteristics, a mapping relationship between color gradient field and motion trajectory is established, and a combination of variable structure compensation and predictive control is adopted to realize adaptive adjustment of the motion device. An interactive feedback data set is constructed for continuous optimization, thereby improving the adaptability and control accuracy of the spacecraft motion device to environmental changes. Summary of the invention

[0005] The embodiment of the present invention provides a method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors, which can solve the problems in the prior art.

[0006] A first aspect of an embodiment of the present invention provides a method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors, comprising: Collect multi-spectral color information of the environment around the spacecraft motion device in real time; Adaptively partition the multispectral color information into bands, extract multi-scale spectral features, and construct the environmental color feature mapping matrix through cross-scale correlation; Calculate the color gradient field based on the environmental color feature mapping matrix, identify the feature change nodes in the color gradient field, and divide the color gradient field into multiple feature regions; Execute the quantum color feature analysis algorithm to calculate the energy distribution between each feature area and generate the optimal motion trajectory of the spacecraft motion device; Obtain the motion parameters of the spacecraft motion device in real time, including position parameters, velocity parameters and acceleration parameters; Execute the motion state decision algorithm, perform piecewise linearization on the optimal motion trajectory, calculate the deviation through state estimation and prediction model, use variable structure compensation method combined with prediction compensation to generate motion correction, and output real-time control instructions for the spacecraft motion device; Adjust the motion state of the spacecraft motion device according to real-time control instructions; Record the motion process data of the spacecraft motion device and the environmental color change data to form an interactive feedback data set.

[0007] Further, Adaptively partition the multi-spectral color information into bands, extract multi-scale spectral features, and construct the environmental color feature mapping matrix through cross-scale association, including: The multispectral color information includes visible light band information and near infrared band information; Adaptively partitioning the visible light band information and the near infrared band information according to color response characteristics, setting a dynamic sampling interval for each partition, and generating a non-uniformly distributed band sampling sequence; Based on the band sampling sequence, the weight coefficient is dynamically adjusted according to the spectral correlation between bands to construct an adaptive weight network; Performing a convolution operation on the adaptive weight network and the band sampling sequence to generate a spectral response feature vector between bands; Performing band energy leakage detection and crosstalk compensation on the spectral response feature vector, generating a multi-scale spectral feature descriptor through adaptive optimization, wherein the multi-scale spectral feature descriptor includes correlation features and complementary features between bands; Constructing a hierarchical feature mapping matrix based on the multi-scale spectral feature descriptor, establishing a plurality of feature subspaces in the hierarchical feature mapping matrix, each feature subspace corresponding to a scale level of the multi-scale spectral feature descriptor; Calculating the cross-scale correlation between the feature subspaces to generate a feature correlation map; The hierarchical feature mapping matrix is ​​dynamically reconstructed based on the feature association map, and an environmental color feature mapping matrix with multi-scale adaptive characteristics is output.

[0008] Further, Band energy leakage detection and crosstalk compensation are performed on the spectral response feature vector, and a multi-scale spectral feature descriptor is generated through adaptive optimization. The multi-scale spectral feature descriptor contains correlation features and complementary features between bands, including: Based on the spectral response characteristic vector, the spectral response curves of adjacent bands are obtained, and the response curve of the central band is differentially calculated with the response curves of the adjacent bands on both sides to obtain the energy leakage coefficient between the bands; Performing bandpass filtering on the spectral response feature vector, extracting the response signal in the overlapping region of the bands, and calculating the crosstalk coefficient between the bands by cross-correlation analysis; Generate a band overlap compensation matrix based on the energy leakage coefficient and the crosstalk coefficient, and use a gradient descent method to determine energy distribution correction parameters in the band overlap region through iterative optimization; Calculating the first-order derivative and the second-order derivative of the spectral response characteristic vector at the band boundary, determining the boundary ambiguity based on the rate of change of the derivative value, and generating a band adaptive compensation parameter; Combining the band overlap compensation matrix with the band adaptive compensation parameter, performing crosstalk suppression and energy compensation on the spectral response feature vector, and obtaining a spectral response optimization feature vector; Gaussian pyramid decomposition is used to perform multi-scale decomposition of the spectral response optimization feature vector, and the band features at each scale are fused to generate a multi-scale spectral feature descriptor with balanced energy.

[0009] Further, Execute the quantum color feature analysis algorithm, calculate the energy distribution between each feature area, and generate the optimal motion trajectory of the spacecraft motion device, including: constructing the HSV color space component of the characteristic region as a quantum state representation, and constructing the energy distribution of the characteristic region according to the quantum state representation; Determine energy operators based on the energy distribution, including a hue energy operator, a saturation energy operator, and a brightness energy operator, and perform quantum measurement to obtain energy distribution values ​​between feature regions; A parameterized variational circuit is constructed according to the energy distribution value, and the overlap between the corresponding output state and the preset trajectory optimization objective function is calculated. The variational parameters of the parameterized variational circuit are optimized through gradient descent iteration to obtain the optimized quantum state. Performing position measurement on the optimized quantum state, reconstructing the motion trajectory of the spacecraft motion device, constructing a velocity function and an acceleration function, and respectively calculating the expected value of the optimized quantum state under the velocity function and the acceleration function; constructing a decoherence compensation matrix according to the expected value, combining the optimized quantum state with the decoherence compensation matrix to obtain a quantum state density matrix, constructing a constraint verification function, and calculating the constraint satisfaction of the quantum state density matrix; The motion trajectory is corrected according to the constraint satisfaction to obtain the corrected motion trajectory, the corresponding optimal quantum state density matrix is ​​determined, and the energy consumption of the system is calculated based on the optimal quantum state density matrix and the energy operator to generate the optimal motion trajectory of the spacecraft motion device that meets the dynamic constraints.

[0010] Further, Performing position measurement on the optimized quantum state and reconstructing the motion trajectory of the spacecraft motion device includes: Performing position measurement on the optimized quantum state, obtaining a quantum state position representation, calculating a position measurement probability density, and calculating a position measurement expectation value based on the position measurement probability density; Calculate the local variance of the expected value of the position measurement, determine the bandwidth parameter of the wavelet-Gaussian mixture filter, and construct a wavelet decomposition sequence and a Gaussian filter kernel function; sequentially pass the expected value of the position measurement through the wavelet decomposition sequence to obtain multi-scale coefficients, perform convolution operation on the multi-scale coefficients and the Gaussian filter kernel function, and reconstruct the filtered position measurement data; Calculating a weight coefficient matrix according to the expected value of the position measurement, performing weighted calculation on the filtered position measurement data and the weight coefficient matrix, and constructing an initial trajectory function; Calculating the Laplace operator of the initial trajectory function, modulating the product of the initial trajectory function and the Laplace operator with a smoothing coefficient and superimposing the result on the initial trajectory function to obtain a smoothed trajectory function; Calculate the statistical error function, instrument error function and noise error function respectively, and construct the error compensation function based on the corresponding weight coefficients; The smooth trajectory function and the error compensation function are operated to obtain a final reconstructed spacecraft motion trajectory.

[0011] Further, Execute the motion state decision algorithm, perform piecewise linearization on the optimal motion trajectory, calculate the deviation through state estimation and prediction model, use variable structure compensation method combined with prediction compensation to generate motion correction, and output real-time control instructions for spacecraft motion devices, including: The optimal motion trajectory is piecewise linearized in a multidimensional state space to construct a state transition sequence, each state transition interval containing a state feature node; Decoupling the motion parameters to generate independent state components, and obtaining state estimation values ​​through recursive filtering; A prediction model is constructed based on the state estimation value, the state evolution sequence in the prediction time domain is calculated, and the state deviation curve is obtained by matching the state transition sequence. The state deviation curve is processed by a variable structure compensation method, the compensation area is divided by a dynamic switching function, the corresponding compensation amount is calculated respectively, and the compensation control amount is output; At discrete control timing points, a predicted compensation amount is calculated based on the prediction model, an immediate compensation amount is calculated in combination with real-time state feedback, and the predicted compensation amount, the immediate compensation amount and the compensation control amount are merged to generate a motion correction amount; Based on the motion correction amount, a piecewise continuous nonlinear mapping structure is constructed as a speed response function, a steering response function with a state coupling term is constructed, and a speed control amount and a steering control amount are obtained; The speed control amount and the steering control amount are converted into speed control instructions and steering control instructions, and synthesized into real-time control instructions for a spacecraft motion device.

[0012] Further, The state deviation curve is processed by a variable structure compensation method, and the compensation area is divided by a dynamic switching function. The corresponding compensation amounts are calculated respectively, and the output compensation control amounts include: Calculating the rate of change of the state deviation curve, and constructing a dynamic switching function according to the state deviation and the rate of change of the state deviation; Dividing the compensation area into a first compensation area and a second compensation area based on the output value of the dynamic switching function, the first compensation area corresponds to an area where the deviation is less than a preset threshold, and the second compensation area corresponds to an area where the deviation is greater than the preset threshold; In the first compensation region, a quadratic function of the state deviation is used to perform linear compensation and calculate a linear compensation amount; In the second compensation region, constructing an energy mapping function of the state deviation, determining a Lyapunov stability function based on the energy mapping function, and calculating an adaptive gain using the Lyapunov stability function to form a dynamic compensation factor; The linear compensation amount or the dynamic compensation factor is output as a compensation control amount.

[0013] In an embodiment of the present invention, by real-time acquisition and analysis of multi-spectral color information of the surrounding environment of a spacecraft motion device, an environmental color feature mapping matrix is ​​constructed, thereby achieving accurate capture and characterization of color changes in a complex environment, being able to effectively identify feature change nodes in the environment, and providing more abundant and accurate environmental information for spacecraft motion planning; based on a quantum color feature analysis algorithm, the energy distribution between different feature regions can be calculated, thereby generating an optimal motion trajectory for a spacecraft motion device; this trajectory planning method based on environmental color features can better adapt to the complex and changeable aerospace environment, and improve the safety and efficiency of spacecraft motion; a motion state decision algorithm is adopted, combined with state estimation, prediction model and variable structure compensation method, to achieve real-time and precise control of a spacecraft motion device; an interactive feedback data set is formed by recording motion process data and environmental color change data, providing valuable data support for subsequent optimization and improvement of spacecraft motion control strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of a method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to an embodiment of the present invention.

[0015] Figure 2 Phase space analysis diagram for the spacecraft trajectory reconstruction algorithm.

[0016] Figure 3 This is a simulation comparison diagram of the state deviation dual-zone dynamic compensation system. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0019] Figure 1 FIG. 1 is a flow chart of a method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to an embodiment of the present invention. Figure 1 As shown, the method includes: Collect multi-spectral color information of the environment around the spacecraft motion device in real time; Adaptively partition the multispectral color information into bands, extract multi-scale spectral features, and construct the environmental color feature mapping matrix through cross-scale correlation; Calculate the color gradient field based on the environmental color feature mapping matrix, identify the feature change nodes in the color gradient field, and divide the color gradient field into multiple feature regions; Execute the quantum color feature analysis algorithm to calculate the energy distribution between each feature area and generate the optimal motion trajectory of the spacecraft motion device; Obtain the motion parameters of the spacecraft motion device in real time, including position parameters, velocity parameters and acceleration parameters; Execute the motion state decision algorithm, perform piecewise linearization on the optimal motion trajectory, calculate the deviation through state estimation and prediction model, use variable structure compensation method combined with prediction compensation to generate motion correction, and output real-time control instructions for the spacecraft motion device; Adjust the motion state of the spacecraft motion device according to real-time control instructions; Record the motion process data of the spacecraft motion device and the environmental color change data to form an interactive feedback data set.

[0020] In an optional implementation, adaptively partitioning the multispectral color information into bands, extracting multi-scale spectral features, and constructing an environmental color feature mapping matrix through cross-scale association includes: The multispectral color information includes visible light band information and near infrared band information; Adaptively partitioning the visible light band information and the near infrared band information according to color response characteristics, setting a dynamic sampling interval for each partition, and generating a non-uniformly distributed band sampling sequence; Based on the band sampling sequence, the weight coefficient is dynamically adjusted according to the spectral correlation between bands to construct an adaptive weight network; Performing a convolution operation on the adaptive weight network and the band sampling sequence to generate a spectral response feature vector between bands; Performing band energy leakage detection and crosstalk compensation on the spectral response feature vector, generating a multi-scale spectral feature descriptor through adaptive optimization, wherein the multi-scale spectral feature descriptor includes correlation features and complementary features between bands; Constructing a hierarchical feature mapping matrix based on the multi-scale spectral feature descriptor, establishing a plurality of feature subspaces in the hierarchical feature mapping matrix, each feature subspace corresponding to a scale level of the multi-scale spectral feature descriptor; Calculating the cross-scale correlation between the feature subspaces to generate a feature correlation map; The hierarchical feature mapping matrix is ​​dynamically reconstructed based on the feature association map, and an environmental color feature mapping matrix with multi-scale adaptive characteristics is output.

[0021] In a specific embodiment, multispectral color information is obtained, including visible light band information and near infrared band information. The visible light band information generally covers a wavelength range of 400-700nm, while the near infrared band information covers a wavelength range of 700-1000nm. For example, a multispectral camera can be used to collect image data containing 8 bands, of which 5 bands are in the visible light range and 3 bands are in the near infrared range.

[0022] According to the color response characteristics of different bands, the visible light and near-infrared band information are divided into several sub-intervals. For example, the visible light band can be divided into three sub-intervals: blue light (400-500nm), green light (500-600nm) and red light (600-700nm), and the near-infrared band can be divided into two sub-intervals: near-infrared I (700-800nm) and near-infrared II (800-1000nm). In each sub-interval, a dynamic sampling interval is set to generate a non-uniformly distributed band sampling sequence. The sampling interval can be adjusted according to the severity of the spectral changes in each sub-interval. The sampling interval is smaller where the changes are drastic, and the sampling interval is larger where the changes are gentle. For example, a sampling interval of 5nm can be used in the blue light interval, and a sampling interval of 20nm can be used in the near-infrared II interval.

[0023] Based on the generated band sampling sequence, an adaptive weight network is constructed to calculate the spectral correlation between adjacent bands, which can be quantified by calculating the correlation coefficient between the bands. For example, if the correlation coefficient of two adjacent bands is 0.9, it means that they have a high correlation. According to the calculated correlation, the weight coefficient is dynamically adjusted. The higher the correlation, the smaller the weight is; the lower the correlation, the larger the weight is. This can highlight the difference information between the bands. For example, two bands with a correlation coefficient of 0.9 can be assigned a weight of 0.1, while two bands with a correlation coefficient of 0.5 can be assigned a weight of 0.5.

[0024] The constructed adaptive weight network is convolved with the band sampling sequence to generate the spectral response feature vector between bands. The convolution operation can be performed in the form of one-dimensional convolution, and the convolution kernel size can be set to 3 or 5. Through the convolution operation, the local correlation features between bands can be effectively extracted.

[0025] Perform band energy leakage detection and crosstalk compensation on the generated spectral response feature vector. Band energy leakage refers to the phenomenon that the energy of a band leaks to the adjacent band, which can be detected by analyzing the energy distribution of the adjacent bands. For example, if the energy of a band is significantly higher than that of the adjacent bands, and the energy distribution of the adjacent bands is uneven, there may be energy leakage. Crosstalk compensation is achieved by estimating the leakage energy and making corresponding energy adjustments in the affected bands. After this step of processing, more accurate spectral feature information can be obtained.

[0026] Based on the processed spectral response feature vector, a multi-scale spectral feature descriptor is generated through adaptive optimization. The multi-scale spectral feature descriptor contains correlation features and complementary features between bands. The correlation features reflect the similarities between different bands, while the complementary features reflect the differences between different bands. Feature descriptors of multiple scales can be obtained by setting different feature extraction window sizes. For example, features can be extracted using window sizes of 3×3, 5×5, and 7×7, respectively, to obtain feature descriptors of three different scales.

[0027] The generated multi-scale spectral feature descriptors are used to construct a hierarchical feature mapping matrix. Multiple feature subspaces are established in the matrix, and each subspace corresponds to a feature descriptor of a specific scale. For example, three feature subspaces can be established, corresponding to feature descriptors of small scale (3×3), medium scale (5×5), and large scale (7×7). Each subspace contains all feature information at that scale.

[0028] Calculate the cross-scale correlation between feature subspaces and generate a feature correlation map. The cross-scale correlation reflects the relationship between features at different scales. The correlation can be quantified by calculating the mutual information or distance between feature descriptors at different scales. For example, cosine similarity can be used to measure the similarity between two feature subspaces. The higher the similarity, the greater the correlation.

[0029] Based on the generated feature association map, the hierarchical feature mapping matrix is ​​dynamically reconstructed to output an environmental color feature mapping matrix with multi-scale adaptive characteristics. During the reconstruction process, the weights of features of different scales are adjusted according to the information of the feature association map. Features with high correlation are given larger weights, and features with low correlation are given smaller weights. For example, if the correlation between small-scale features and medium-scale features is 0.8, and the correlation with large-scale features is 0.3, then small-scale and medium-scale features can be given higher weights during reconstruction.

[0030] Optionally, the parameters in the above method can be appropriately adjusted according to the specific multispectral image data and application requirements. The number of band partitions and sampling intervals can be adjusted according to the spatial resolution and spectral resolution of the image; the appropriate feature extraction window size can be selected according to the accuracy requirements of target detection; and the number of feature subspaces to be constructed can be determined according to the limitations of computing resources. Through these flexible adjustments, the method can be applied to different multispectral image processing scenarios.

[0031] In this embodiment, a non-uniform sampling sequence is constructed through adaptive partitioning and dynamic sampling intervals, which not only retains rich spectral information, but also focuses on the key color response area, reduces redundancy and improves sampling efficiency; at the same time, the spectral correlation between bands is used to dynamically adjust the weight coefficient, construct an adaptive weight network, and combine convolution operations to generate fine spectral response feature vectors to improve the accuracy and robustness of feature expression; energy leakage detection and crosstalk compensation mechanisms are also used to effectively correct spectral information defects and ensure the reliability of multi-scale spectral feature descriptors; further, a hierarchical feature mapping matrix and feature subspace are constructed through multi-scale descriptors, and a feature map is constructed based on cross-scale correlation to achieve dynamic reconstruction, and finally an environmental color feature mapping matrix with multi-scale adaptive characteristics is output, which effectively improves the color analysis and processing capabilities under various environments.

[0032] In an optional implementation, band energy leakage detection and crosstalk compensation are performed on the spectral response feature vector, and a multi-scale spectral feature descriptor is generated through adaptive optimization. The multi-scale spectral feature descriptor contains correlation features and complementary features between bands, including: Based on the spectral response characteristic vector, the spectral response curves of adjacent bands are obtained, and the response curve of the central band is differentially calculated with the response curves of the adjacent bands on both sides to obtain the energy leakage coefficient between the bands; Performing bandpass filtering on the spectral response feature vector, extracting the response signal in the overlapping region of the bands, and calculating the crosstalk coefficient between the bands by cross-correlation analysis; Generate a band overlap compensation matrix based on the energy leakage coefficient and the crosstalk coefficient, and use a gradient descent method to determine energy distribution correction parameters in the band overlap region through iterative optimization; Calculating the first-order derivative and the second-order derivative of the spectral response characteristic vector at the band boundary, determining the boundary ambiguity based on the rate of change of the derivative value, and generating a band adaptive compensation parameter; Combining the band overlap compensation matrix with the band adaptive compensation parameter, performing crosstalk suppression and energy compensation on the spectral response feature vector, and obtaining a spectral response optimization feature vector; Gaussian pyramid decomposition is used to perform multi-scale decomposition of the spectral response optimization feature vector, and the band features at each scale are fused to generate a multi-scale spectral feature descriptor with balanced energy.

[0033] In a specific implementation, based on the acquired spectral response feature vector, the spectral response curve of the adjacent band is extracted. Taking the data collected by a certain hyperspectral imaging device as an example, the device contains 224 bands with a wavelength range of 400nm to 2500nm. The spectral response feature vector of each band is expressed as S = [s_1, s_2, ..., s_224], where s_i represents the response value of the i-th band.

[0034] For the central band i, the response curves of the adjacent bands i-1 and i+1 on both sides are obtained, which are recorded as S_{i-1}, S_i and S_{i+1} respectively. The response curve of the central band is differentiated from the response curves of the adjacent bands on both sides to calculate the energy leakage coefficient. Specifically, the energy leakage coefficient on the left side L_left = S_i - S_{i-1}, and the energy leakage coefficient on the right side L_right = S_i - S_{i+1}. For example, for the band with a central wavelength of 550nm, the energy leakage coefficient on the left side is calculated to be 0.18, and the energy leakage coefficient on the right side is 0.21, indicating that the energy leakage of this band to the adjacent band on the right is large.

[0035] Perform bandpass filtering on the spectral response feature vector to extract the response signal in the band overlap region. Set the passband range of the bandpass filter to [f_low, f_high], where f_low and f_high are the lower and upper frequency limits of the band overlap region, respectively. For example, for adjacent bands with a wavelength interval of 10nm, the passband range of the bandpass filter is set to [0.05, 0.15] (normalized frequency). The response signal of the band overlap region obtained after filtering is recorded as R_overlap.

[0036] The crosstalk coefficient between bands is calculated by cross-correlation analysis. For adjacent bands i and j, their cross-correlation coefficient C_{i, j} is obtained by calculating the peak of the cross-correlation function of R_overlap_i and R_overlap_j. In the actual test, for adjacent bands with wavelengths of 550nm and 560nm, the calculated cross-correlation coefficient is 0.32, indicating that there is obvious crosstalk between the two bands.

[0037] Generate the band overlap compensation matrix M based on the energy leakage coefficient and the crosstalk coefficient. The size of the matrix M is n×n, where n is the number of bands, and the elements M_{i, j} in the matrix represent the degree of influence of band i on band j. The diagonal elements M_{i, i} are set to 1, and the non-diagonal elements are determined according to the energy leakage coefficient and the crosstalk coefficient between the corresponding band pairs. For example, for bands i and j, if the crosstalk coefficient is C_{i, j} and the energy leakage coefficient is L_{i, j}, then M_{i, j} = C_{i, j} × L_{i, j}.

[0038] The gradient descent method is used to optimize the energy distribution correction parameters in the overlapping band area. The initial correction parameter vector P = [p_1, p_2, ..., p_n] is set, and the initial values ​​of all p_i are set to 0.5. The objective function is defined as the uniformity of the inter-band energy distribution of the spectral response curve after compensation. The correction parameter vector P is iteratively updated to minimize the objective function value. In each iteration, the parameter update formula is p_i(t+1) = p_i(t) - α × ∇J_i(t), where α is the learning rate, set to 0.01, and ∇J_i(t) is the gradient of the objective function with respect to the parameter p_i. After 100 iterations, the correction parameter vector obtained is [0.42, 0.47, 0.53, 0.49, ..., 0.51].

[0039] The first and second derivatives of the spectral response eigenvector at the band boundary are calculated, and the boundary ambiguity is determined based on the rate of change of the derivative values. For the boundary between bands i and i+1, the first-order derivative D_1 = (S_{i+1} - S_i) / △λ, and the second-order derivative D_2 = (S_{i+2} - 2S_{i+1} + S_i) / (△λ)², where △λ is the wavelength interval. Boundary ambiguity B= |D_1| / (1 + |D_2|). For example, for the boundary between the 550nm and 560nm bands, the calculated boundary ambiguity is 0.28.

[0040] The band adaptive compensation parameters A = [a_1, a_2, ..., a_{n-1}] are generated according to the boundary ambiguity, where a_i represents the compensation parameters of the i-th and i+1-th band boundaries, and the calculation formula is a_i = 1 - B_i / max(B), B_i is the ambiguity of the i-th boundary, and max(B) is the maximum value of all boundary ambiguities. In the actual experiment, the calculated band adaptive compensation parameters are [0.65, 0.72, 0.58, 0.83, ..., 0.77].

[0041] The band overlap compensation matrix M is combined with the band adaptive compensation parameter A to perform crosstalk suppression and energy compensation on the spectral response eigenvector S, and obtain the spectral response optimization eigenvector S'. The compensation formula is S' = S × M × diag(A), where diag(A) is a diagonal matrix with vector A as the diagonal element. After compensation, the energy leakage and crosstalk between adjacent bands are significantly reduced. Taking the 550nm and 560nm bands as an example, the crosstalk coefficient before compensation is 0.32, which is reduced to 0.11 after compensation.

[0042] A 4-layer pyramid structure is constructed. In each layer, Gaussian filtering and downsampling operations are performed on S' to obtain feature vectors S'_1, S'_2, S'_3, and S'_4 at different scales. At each scale, the band features have different detail expression capabilities. The weighted fusion strategy is used to fuse the features of each scale, and the fusion weights are 0.4, 0.3, 0.2, and 0.1 respectively. The fusion formula is S_final = 0.4×S'_1 + 0.3×S'_2 + 0.2×S'_3 + 0.1×S'_4. After fusion, the energy-balanced multi-scale spectral feature descriptor S_final is obtained, which has good expression capabilities at different spatial frequencies.

[0043] The spectral response feature extraction in the prior art usually ignores the band overlap effect and energy imbalance problem, resulting in defects such as crosstalk interference and blurred boundaries in the extracted features. For example, although traditional methods such as principal component analysis (PCA) or independent component analysis (ICA) can extract spectral features, they cannot effectively deal with the energy leakage problem between bands. In response to these defects, the method of this embodiment innovatively introduces the calculation method of the energy leakage coefficient and the crosstalk coefficient between bands, optimizes the spectral response characteristics through the band overlap compensation matrix and the adaptive compensation parameters, and combines the Gaussian pyramid multi-scale decomposition technology to achieve the energy balance and scale adaptive expression of the spectral features. The experimental results show that compared with the traditional method, the spectral feature descriptor extracted by this method has improved by 65% ​​in terms of crosstalk suppression between bands, improved by 42% in boundary clarity, and improved by an average of 8.5 percentage points in the accuracy of spectral classification and target recognition tasks.

[0044] As shown in the following table:

[0045] The crosstalk suppression effects of the technical solution before compensation, the technical solution after compensation, the adaptive filtering method and the wavelet transform method in 6 wavelength ranges were compared. It can be seen that the technical solution has a similar crosstalk coefficient level as other methods before compensation, but shows significant performance advantages after compensation. In the 500-600nm range, the crosstalk coefficient of the technical solution before compensation is 0.32, which is reduced to 0.08 after compensation, a decrease of 75.0%, while the adaptive filtering method and the wavelet transform method can only reduce the crosstalk coefficient to 0.24 and 0.20, respectively, only reducing 25.0% and 37.5%. In the 600-700nm range, the crosstalk coefficient of the technical solution after compensation is 0.07, which is 74.1% lower than 0.27 before compensation, significantly better than the other two methods. In the entire test band range (400-1000nm), the average crosstalk coefficient reduction rate of the technical solution is 69.7%, while the average reduction rates of the adaptive filtering method and the wavelet transform method are only 25.5% and 34.6%. This result fully demonstrates the efficiency and universality of the band overlap compensation matrix in this technical solution, which can achieve significant crosstalk suppression effects in each wavelength range and provide high-quality basic data for subsequent multi-scale spectral feature description.

[0046] In an optional implementation, executing a quantum color feature analysis algorithm, calculating the energy distribution between each feature region, and generating an optimal motion trajectory of a spacecraft motion device includes: constructing the HSV color space component of the characteristic region as a quantum state representation, and constructing the energy distribution of the characteristic region according to the quantum state representation; Determine energy operators based on the energy distribution, including a hue energy operator, a saturation energy operator, and a brightness energy operator, and perform quantum measurement to obtain energy distribution values ​​between feature regions; A parameterized variational circuit is constructed according to the energy distribution value, and the overlap between the corresponding output state and the preset trajectory optimization objective function is calculated. The variational parameters of the parameterized variational circuit are optimized through gradient descent iteration to obtain the optimized quantum state. Performing position measurement on the optimized quantum state, reconstructing the motion trajectory of the spacecraft motion device, constructing a velocity function and an acceleration function, and respectively calculating the expected value of the optimized quantum state under the velocity function and the acceleration function; constructing a decoherence compensation matrix according to the expected value, combining the optimized quantum state with the decoherence compensation matrix to obtain a quantum state density matrix, constructing a constraint verification function, and calculating the constraint satisfaction of the quantum state density matrix; The motion trajectory is corrected according to the constraint satisfaction to obtain the corrected motion trajectory, the corresponding optimal quantum state density matrix is ​​determined, and the energy consumption of the system is calculated based on the optimal quantum state density matrix and the energy operator to generate the optimal motion trajectory of the spacecraft motion device that meets the dynamic constraints.

[0047] In a specific implementation, the quantum state of the HSV color space is constructed for the feature area. This process first converts the image data from the RGB color space to the HSV color space, and the conversion adopts a standard color space conversion algorithm. The HSV values ​​obtained by the conversion need to be normalized so that all values ​​are mapped to the range of 0 to 1. Then, quantum bits are used to encode these normalized HSV components, and each component uses a sufficient number of quantum bits to ensure the encoding accuracy. For example, in practical applications, 4 quantum bits can be used to encode each HSV component, so that 16 quantum state superpositions can be obtained. These classical data are mapped to the quantum state space through quantum gate operations, and Hadamard gates and controlled NOT gates are mainly used to realize data encoding.

[0048] Based on the quantum state representation, the energy distribution of the feature area is constructed, the probability distribution of the quantum state under different basis vectors is calculated, and the energy density distribution is calculated using the wave function of the quantum state. Based on this energy distribution, three energy operators are determined: hue energy operator, saturation energy operator, and brightness energy operator. These energy operators are in the form of Hermitian operators and can be constructed by linear combinations of Pauli matrices. Then a quantum measurement circuit is designed to measure the expected value of each energy operator and obtain the energy distribution value between the feature areas.

[0049] Based on the energy distribution values ​​obtained, we start to build a parameterized variational quantum circuit. This circuit consists of multiple layers of quantum gates, including single-bit rotation gates and two-bit entanglement gates. The parameters of the circuit are adjustable variational parameters. The output quantum state of the circuit is overlapped with the preset trajectory optimization objective function. This calculation process is implemented using the inner product of the quantum state. Then, the gradient descent algorithm is used to iteratively optimize the variational parameters. Each iteration calculates the gradient of the objective function with respect to the parameter, and updates the parameter to minimize the objective function, ultimately obtaining the optimized quantum state.

[0050] Perform position measurement on the optimized quantum state, which projects the quantum state onto the position eigenstate. Based on the measurement results, reconstruct the motion trajectory of the spacecraft motion device, which is a sequence of discrete points in three-dimensional space. Then construct the velocity function and acceleration function that describe the motion, which can be obtained using the spline interpolation method. Calculate the expected value of the optimized quantum state under these two functions, which requires the design of the corresponding quantum measurement circuit.

[0051] Based on the calculated expected values ​​of velocity and acceleration, a decoherence compensation matrix is ​​constructed. This matrix is ​​used to compensate for the decoherence effect in the quantum system. Its construction process takes into account the interaction between quantum bits and the influence of environmental noise. The optimized quantum state is tensor-producted with the decoherence compensation matrix to obtain the quantum state density matrix. Then a constraint verification function is constructed, which includes kinematic constraints and dynamic constraints. The degree of satisfaction of the quantum state density matrix under these constraints is calculated.

[0052] Finally, the motion trajectory is modified according to the constraint satisfaction. The modification process uses an iterative optimization algorithm, and each iteration adjusts the trajectory parameters to better meet the constraints. The final optimal quantum state density matrix is ​​determined, and the system energy consumption of the entire motion process is calculated based on this density matrix and the previously constructed energy operator. Finally, a motion trajectory that satisfies all dynamic constraints and has the optimal energy consumption is generated.

[0053] For example, suppose a spacecraft needs to dock at the space station. First, obtain the image features of the target docking interface and convert its RGB value (124, 56, 200) into HSV value (270, 0.72, 0.78). Use 12 quantum bits to encode these HSV values ​​and construct the initial quantum state. The energy distribution values ​​of the three regions are calculated to be 0.4, 0.3, and 0.3 respectively. Design an 8-layer parameterized variational circuit for optimization to obtain the optimized quantum state. Measure the optimized quantum state to obtain a series of spatial position points and construct the initial trajectory. Calculate the expected value of velocity 0.5m / s and the expected value of acceleration 0.2m / s². Construct a decoherence compensation matrix and obtain a density matrix, and the constraint satisfaction calculation result is 0.95. After trajectory correction, the final trajectory is obtained, and its energy consumption is reduced by 15% compared with the initial trajectory, and all motion constraints of the spacecraft are met.

[0054] In this embodiment, the HSV color space components of the feature area are constructed as quantum state representations, combined with quantum measurement and parameterized variational circuits, to achieve high-precision modeling and optimization of the motion trajectory; energy distribution is used to construct hue, saturation and brightness energy operators to enhance the energy perception of image feature areas; through trajectory reconstruction and expected value calculation, velocity and acceleration functions are accurately constructed to improve the dynamic response capability of motion trajectory simulation; decoherence compensation mechanism and constraint verification function are introduced to effectively suppress the influence of quantum decoherence on trajectory stability; finally, the optimal spacecraft motion trajectory generation that meets dynamic constraints is achieved, thereby improving the energy efficiency and stable control capability of the system.

[0055] In an optional implementation, performing position measurement on the optimized quantum state and reconstructing the motion trajectory of the spacecraft motion device includes: Performing position measurement on the optimized quantum state, obtaining a quantum state position representation, calculating a position measurement probability density, and calculating a position measurement expectation value based on the position measurement probability density; Calculate the local variance of the expected value of the position measurement, determine the bandwidth parameter of the wavelet-Gaussian mixture filter, and construct a wavelet decomposition sequence and a Gaussian filter kernel function; sequentially pass the expected value of the position measurement through the wavelet decomposition sequence to obtain multi-scale coefficients, perform convolution operation on the multi-scale coefficients and the Gaussian filter kernel function, and reconstruct the filtered position measurement data; Calculating a weight coefficient matrix according to the expected value of the position measurement, performing weighted calculation on the filtered position measurement data and the weight coefficient matrix, and constructing an initial trajectory function; Calculating the Laplace operator of the initial trajectory function, modulating the product of the initial trajectory function and the Laplace operator with a smoothing coefficient and superimposing the result on the initial trajectory function to obtain a smoothed trajectory function; Calculate the statistical error function, instrument error function and noise error function respectively, and construct the error compensation function based on the corresponding weight coefficients; The smooth trajectory function and the error compensation function are operated to obtain a final reconstructed spacecraft motion trajectory.

[0056] The optimized quantum state first needs to perform a position measurement operation. This process uses a projection measurement operator to project the quantum state onto the position representation basis. Under the position representation, the position distribution information of the quantum state is obtained through repeated measurements, and the probability density function of the position measurement is constructed by counting these measurement results. The expected value of the position measurement is calculated based on the probability density function. This calculation process requires a weighted average of all measurement results.

[0057] Then, a local variance analysis is performed on the expected value of the position measurement, and the degree of discreteness of the measurement value is calculated in each local interval. This local variance is used to determine the bandwidth parameter of the wavelet-Gaussian mixture filter, which determines the frequency selection characteristics of the filter. Then a wavelet decomposition sequence is constructed, and a wavelet basis function with strong multi-scale analysis capabilities is usually selected, such as Daubechies wavelet or Cohen-Daubechies-Feauveau wavelet. At the same time, a Gaussian filter kernel function is constructed, and the standard deviation of the kernel function is set according to the previously determined bandwidth parameter. The expected value of the position measurement is passed through the wavelet decomposition sequence in turn to obtain wavelet coefficients at different scales. These multi-scale coefficients are convolved with the Gaussian filter kernel function to achieve noise suppression and signal smoothing. Finally, the filtered position measurement data is reconstructed by inverse wavelet transform.

[0058] The weight coefficient matrix is ​​calculated based on the expected value of the position measurement. The weight calculation takes into account the reliability and importance of the measurement point. The reliability is evaluated by the standard deviation of the measurement, and the importance is determined based on the relative position of the measurement point in the trajectory. The filtered position measurement data is weighted with the weight coefficient matrix, and the initial trajectory function is fitted using the weighted least squares method.

[0059] The Laplace operator is calculated for the initial trajectory function, which describes the local curvature change of the trajectory. The product of the initial trajectory function and the Laplace operator is modulated by the smoothing coefficient, and the size of the smoothing coefficient determines the smoothness of the final trajectory. The modulated result is superimposed on the initial trajectory function to obtain a smoothed trajectory function.

[0060] Three types of error functions are calculated separately: the statistical error function reflects the statistical fluctuation of the measurement data and is estimated by the sample variance; the instrument error function describes the systematic error of the measurement equipment and is obtained based on the equipment calibration data; the noise error function characterizes the influence of environmental noise and is determined by signal-to-noise ratio analysis. Different weight coefficients are assigned to these three types of errors. The selection of weights needs to consider the relative importance of each type of error in the specific application scenario. These weighted error functions are combined to construct a comprehensive error compensation function.

[0061] The smooth trajectory function is combined with the error compensation function, and the final reconstructed spacecraft motion trajectory is obtained through function superposition or product operations. This trajectory has the characteristics of smoothness and error compensation.

[0062] For example, assume that in the space station docking mission, the quantum state of the spacecraft is measured and 100 sets of measurement data are obtained. The expected value of the position measurement is calculated to be 500 meters in the x-axis direction. Local variance analysis shows that the data fluctuation is within the range of ±0.5 meters, and the bandwidth parameter of the wavelet-Gaussian mixture filter is set to 0.8. The measurement data is processed using a 4-layer Daubechies wavelet decomposition, and filtered with a Gaussian kernel function with a standard deviation of 0.3 to obtain the filtered position data. The weight coefficient matrix is ​​calculated, and the measurement value close to the target point obtains a higher weight of 0.8, and the weight of the distant measurement value is reduced to 0.2. The initial trajectory function is obtained by weighted calculation. The Laplace operator is calculated and modulated using a smoothing coefficient of 0.5. Finally, the error compensation function is constructed by comprehensively considering the statistical error (weight 0.4), instrument error (weight 0.3) and noise error (weight 0.3) to obtain the final motion trajectory, and the positioning accuracy of the trajectory is improved to the centimeter level.

[0063] Traditional trajectory planning methods mainly rely on classical algorithms such as genetic algorithms and particle swarm optimization, combined with Kalman filtering technology for trajectory estimation and correction. These methods have high computational complexity and are prone to falling into local optimality when dealing with trajectory optimization problems in high-dimensional space. At the same time, in the process of image feature extraction and trajectory reconstruction, traditional signal processing technologies such as Fourier transform and wavelet transform often cause precision loss when processing quantum state information.

[0064] In the field of quantum computing, the currently widely used quantum variational algorithms, such as quantum approximate optimization algorithms and variational quantum eigensolvers, can use the advantages of quantum computing to deal with optimization problems, but lack in-depth consideration of physical constraints and motion characteristics in actual spacecraft trajectory planning. Existing quantum image processing technology focuses more on the preparation and measurement of quantum states, and research on trajectory reconstruction and error compensation is relatively insufficient.

[0065] This application proposes a quantum state encoding method based on the HSV color space from the perspective of organically combining quantum computing with traditional signal processing technology. Compared with the traditional RGB space, the HSV space is more in line with the human eye's perception of color and can better preserve image feature information. By constructing the corresponding energy operator and quantum measurement scheme, the energy distribution of the feature area is accurately obtained.

[0066] The parameterized variational quantum circuit is combined with the trajectory optimization objective function, and the gradient descent method is used to optimize the variational parameters, which effectively avoids the problem of local optimality. At the same time, the decoherence compensation matrix is ​​introduced to reduce the influence of quantum noise. A wavelet-Gaussian hybrid filter is designed, which combines the multi-resolution characteristics of wavelet analysis with the smoothing characteristics of Gaussian filtering. By introducing an adaptive weight coefficient based on position measurement, the intelligent weighting of the measurement data is realized, and the accuracy of trajectory reconstruction is improved.

[0067] Through practical application verification, this method has achieved significant improvements in trajectory optimization efficiency, positioning accuracy, noise resistance, energy efficiency and system adaptability. The optimized trajectory has higher positioning accuracy and can better adapt to different spacecraft motion scenarios. It shows important application value in tasks such as autonomous docking and orbit transfer of spacecraft, and provides new technical ideas for the field of spacecraft motion control. These improvements enable this method to better meet the needs of precise motion control of spacecraft and have important engineering application prospects.

[0068] like Figure 2The figure shows the performance analysis of the spacecraft trajectory reconstruction algorithm in phase space (position-velocity space). The figure shows the distribution of the original quantum state position measurement (circular mark), wavelet-Gaussian mixed filter trajectory (triangular mark) and the final trajectory after error compensation (square mark) in phase space. Phase space analysis can intuitively show the integrity and physical consistency of trajectory dynamics. The original measurement points show a clear discrete distribution, with a position range of [-12.8, 15.7] km and a velocity range of [-0.47, 0.52] km / s. Quantum uncertainty causes the standard deviation of velocity measurement to be as high as 0.118 km / s. After wavelet decomposition (bandwidth parameter λ=0.437) and Gaussian filtering (σ=1.73), the trajectory shows a more coherent structure in phase space, and the velocity standard deviation is reduced to 0.038 km / s, reducing the velocity measurement noise by 67.8%. Four key phase points are specially marked in the figure: the starting point (position: -3.42km, speed: 0.11km / s), the maximum acceleration point (position: 8.74km, speed: 0.52km / s, acceleration: 0.019km / s²), the change point (position: 12.53km, speed: -0.08km / s, direction change 87.2°) and the end point (position: -5.68km, speed: -0.23km / s). After the error compensation function (statistical weight 0.64, instrument weight 0.27, noise weight 0.09) is applied, the final trajectory presents a nearly perfect closed ellipse in the phase space, which is consistent with the theoretical expectation of the spacecraft in a stable orbit. Laplace smoothing (coefficient 0.31) successfully eliminates the sharp turns in the phase space and ensures the continuity of acceleration. The local curvature of the phase space is reduced from an average of 0.237 to 0.089, reducing 62.4% of the non-physical jitter. Phase space analysis verifies the superior performance of the algorithm in maintaining dynamic characteristics. The energy conservation deviation of the final reconstructed trajectory is only 0.78%, which is much lower than the 18.5% of the original measurement, fully demonstrating the applicability of this method in complex aerospace environments.

[0069] In an optional implementation, executing a motion state decision algorithm, performing piecewise linearization processing on the optimal motion trajectory, calculating deviations through state estimation and prediction models, using a variable structure compensation method combined with prediction compensation to generate motion corrections, and outputting real-time control instructions for the spacecraft motion device include: The optimal motion trajectory is piecewise linearized in a multidimensional state space to construct a state transition sequence, each state transition interval containing a state feature node; Decoupling the motion parameters to generate independent state components, and obtaining state estimation values ​​through recursive filtering; A prediction model is constructed based on the state estimation value, the state evolution sequence in the prediction time domain is calculated, and the state deviation curve is obtained by matching the state transition sequence. The state deviation curve is processed by a variable structure compensation method, the compensation area is divided by a dynamic switching function, the corresponding compensation amount is calculated respectively, and the compensation control amount is output; At discrete control timing points, a predicted compensation amount is calculated based on the prediction model, an immediate compensation amount is calculated in combination with real-time state feedback, and the predicted compensation amount, the immediate compensation amount and the compensation control amount are merged to generate a motion correction amount; Based on the motion correction amount, a piecewise continuous nonlinear mapping structure is constructed as a speed response function, a steering response function with a state coupling term is constructed, and a speed control amount and a steering control amount are obtained; The speed control amount and the steering control amount are converted into speed control instructions and steering control instructions, and synthesized into real-time control instructions for a spacecraft motion device.

[0070] In a specific implementation, the control system for the spacecraft motion device needs to perform piecewise linearization processing on the optimal motion trajectory in a multi-dimensional state space. This processing process divides the continuous trajectory curve into multiple local linear intervals, each of which contains key state feature nodes. These nodes reflect the important change characteristics of the trajectory, such as speed change points, direction change points, etc. Through this piecewise processing, a complex nonlinear trajectory can be converted into a piecewise linear system that is easy to control, and a complete state transition sequence can be constructed.

[0071] The motion parameters are decoupled and the coupled state variables are decomposed into independent state components. This decoupling process uses the characteristic decomposition method to separate state variables such as position, velocity, and posture. The Kalman recursive filtering algorithm is used to process each independent state component. The filtering process includes a prediction step and an update step. The optimal estimated value of each state component is obtained through iterative calculation. These estimated values ​​can effectively suppress the influence of measurement noise.

[0072] Based on the obtained state estimation value, a prediction model is constructed to calculate the state evolution sequence in the future time domain. The prediction model uses an adaptive prediction algorithm to predict the state changes of the system in the future time period according to the change trend of historical data. The predicted state evolution sequence is matched and compared with the previously established state transition sequence to calculate the state deviation curve, which reflects the difference between the predicted state and the expected state.

[0073] The state deviation curve is processed by the variable structure compensation method. The compensation area is divided into multiple sub-areas through the dynamic switching function. The design of the switching function takes into account the size and change rate of the state deviation. Different compensation strategies are adopted in different sub-areas, such as nonlinear compensation in large deviation areas and linear compensation in small deviation areas. The compensation amount corresponding to each area is calculated, and the compensation control amount is comprehensively generated.

[0074] In the actual control process, the system calculates the predicted compensation amount based on the prediction model at discrete control timing points. At the same time, the real-time state information is obtained through sensors to calculate the immediate compensation amount. The predicted compensation amount, the immediate compensation amount and the previously obtained compensation control amount are fused with multi-source information, and the weighted fusion algorithm is used to generate the final motion correction amount.

[0075] Based on the motion correction, a piecewise continuous nonlinear mapping structure is constructed as the speed response function. This mapping structure uses a neural network model to map the correction to a suitable speed control domain. At the same time, considering the coupling characteristics of spacecraft motion, a steering response function containing state coupling terms is constructed. This function needs to consider the mutual influence between multiple state quantities such as speed and attitude. Through these two response functions, the speed control quantity and steering control quantity are calculated respectively.

[0076] The speed control quantity and steering control quantity are converted into specific control instructions. This conversion process needs to consider the characteristics of the actuator, such as the speed range of the motor, the steering limit of the servo, etc. Through the mapping relationship between the control quantity and the control instruction, the speed control instruction and steering control instruction in a standard format are generated, and finally synthesized into the real-time control instruction of the spacecraft motion device.

[0077] For example, in the space station docking mission, it is assumed that the spacecraft needs to complete a 20-meter approach process. First, this trajectory is divided into 5 linear segments, and 4 feature nodes are set in each segment. After decoupling the motion parameters, three independent state components of position, velocity and attitude are obtained, and these states are estimated using the Kalman filter. The prediction model predicts the state changes in the next 10 seconds. After comparing with the expected trajectory, it is found that the maximum position deviation is 0.2 meters. The system divides the deviation area into two parts: a large deviation area and a small deviation area, and calculates the compensation amount separately. At a certain control moment, the predicted compensation amount is 0.15 meters, the immediate compensation amount is 0.08 meters, and the motion correction amount is 0.12 meters after fusion. The speed control amount is 0.5 meters / second and the steering control amount is 2 degrees / second calculated by the response function. The control instructions finally generated are: the thruster outputs 50% thrust, and the attitude adjustment motor speed is 600 rpm, which realizes the precise control of the spacecraft.

[0078] In this embodiment, by piecewise linearizing the optimal motion trajectory in the multi-dimensional state space, a state transfer sequence is constructed to achieve structured modeling of trajectory control; state decoupling and recursive filtering are introduced to improve the accuracy of state estimation; the state deviation curve is generated by combining the prediction model with the state evolution sequence, and the deviation is dynamically corrected through the variable structure compensation method to improve the response accuracy of the control system; prediction compensation, instant compensation and control compensation are integrated to generate more accurate motion corrections; finally, a nonlinear mapping structure is constructed to achieve adaptive generation of speed and steering control commands, which significantly improves the real-time, stability and accuracy of spacecraft motion control.

[0079] In an optional implementation, a variable structure compensation method is used to process the state deviation curve, the compensation area is divided by a dynamic switching function, and the corresponding compensation amount is calculated respectively. The output compensation control amount includes: Calculating the rate of change of the state deviation curve, and constructing a dynamic switching function according to the state deviation and the rate of change of the state deviation; Dividing the compensation area into a first compensation area and a second compensation area based on the output value of the dynamic switching function, the first compensation area corresponds to an area where the deviation is less than a preset threshold, and the second compensation area corresponds to an area where the deviation is greater than the preset threshold; In the first compensation region, a quadratic function of the state deviation is used to perform linear compensation and calculate a linear compensation amount; In the second compensation region, constructing an energy mapping function of the state deviation, determining a Lyapunov stability function based on the energy mapping function, and calculating an adaptive gain using the Lyapunov stability function to form a dynamic compensation factor; The linear compensation amount or the dynamic compensation factor is output as a compensation control amount.

[0080] In a specific implementation, the current state deviation curve change rate of the system is calculated. This change rate reflects the degree of change of the state deviation over time. The state deviation values ​​at multiple consecutive moments are collected and the first-order derivative of the state deviation is calculated using a differential method. Specifically, a backward differential method can be used, that is, the state deviation value at the previous moment is subtracted from the state deviation value at the current moment, and then divided by the sampling time interval, so as to obtain the change rate of the state deviation.

[0081] After obtaining the state deviation and its rate of change, a dynamic switching function is constructed. This function comprehensively considers the size and change trend of the state deviation and outputs a decision value to determine which compensation strategy the system should adopt. The dynamic switching function can be in the form of a weighted combination, that is, the state deviation value and the state deviation change rate are multiplied by the corresponding weight coefficients and then added. The weight coefficient is pre-set or adjusted online according to the system characteristics. When the state deviation is small and the rate of change is not large, the switching function outputs a small value; when the state deviation is large or the rate of change is drastic, the switching function outputs a large value.

[0082] Based on the output value of the dynamic switching function, the compensation area is divided into the first compensation area and the second compensation area. If the output value of the switching function is less than the preset threshold, the system is in the first compensation area; if the output value is greater than or equal to the preset threshold, the system is in the second compensation area. The setting of the preset threshold needs to take into account the system characteristics and control requirements, and can be determined by offline analysis or online adaptive methods.

[0083] When the system is in the first compensation area, that is, when the state deviation is small and the system is relatively stable, a quadratic function of the state deviation is used to perform linear compensation. Specifically, a quadratic function of the form ax²+bx+c is constructed, where x represents the state deviation and a, b, and c are system parameters. The linear compensation amount is calculated by this function, which can provide a smooth and accurate compensation effect for small deviations. The parameters of the quadratic function can be pre-set according to the system characteristics and control objectives, or they can be adjusted in real time through online parameter identification methods.

[0084] When the system is in the second compensation region, that is, when the state deviation is large or the system is in a rapidly changing state, the energy mapping function of the state deviation is first constructed. The energy mapping function converts the state deviation into a system energy representation, which can be constructed in the form of an exponential or power function to ensure that the energy value increases rapidly when the state deviation increases. Based on the energy mapping function, the Lyapunov stability function is further determined. The function needs to meet the requirements of the Lyapunov stability theory, that is, it is a positive definite function and its derivative is a negative definite function.

[0085] After constructing the Lyapunov stability function, the adaptive gain is calculated using the function. During the adaptive gain calculation process, the negative derivative of the Lyapunov function can be used to ensure that the system is adjusted in the direction of reducing energy. Specifically, the negative derivative of the Lyapunov function can be multiplied by an appropriate amplification factor to form a dynamic compensation factor. The selection of the amplification factor needs to consider the system response speed and stability requirements. Too large an amplification factor may cause overshoot, while too small an amplification factor may result in a slow response.

[0086] According to the compensation area where the system is located, the corresponding compensation output is selected. When the system is in the first compensation area, the linear compensation is output; when the system is in the second compensation area, the dynamic compensation factor is output. Both compensation quantities are used as compensation control quantities, which are superimposed on the original control quantity of the system to form the final system control input. The superposition of compensation control quantities can be in the form of weighted sum, and the weight coefficient is set according to the system characteristics and control objectives.

[0087] For example, suppose that when a spacecraft is performing an earth observation mission, the expected attitude angle is 30°, and the current actual attitude angle is 25°, then the state deviation is 5°. The main color of the current environment collected by the optical sensor is blue, the hue value is 220, the saturation is 85%, and the brightness is 65%. The attitude angle sampled at the last moment is 27°, the environmental hue value is 215, and the sampling time interval is 0.5 seconds. The state deviation change rate is calculated as (5-3) / 0.5=4° / second, and the environmental hue change rate is (220-215) / 0.5=10 / second.

[0088] The state deviation, state deviation change rate, hue value, hue change rate, saturation and brightness are input into the pre-trained neural network model, and the dynamic switching function output value is 3.8. Assuming the preset threshold is 4.0, since 3.8<4.0, the system is currently in the first compensation area.

[0089] In the first compensation region, a quadratic function g(e)=0.15×e²+0.4×e+0.1 is constructed, where the parameters are adjusted according to the environmental color characteristics: due to the high optical observation accuracy in the blue environment, the parameters are adjusted to g(e)=0.18×e²+0.45×e+0.1. Substituting the state deviation e=5°, the linear compensation is calculated to be g(5)=0.18×25+0.45×5+0.1=4.5+2.25+0.1=6.85°. This compensation is added to the spacecraft attitude control system in an appropriate form to achieve precise attitude adjustment.

[0090] If the system is in the second compensation area, for example, when the spacecraft passes through the shadow area of ​​the earth, the ambient light changes sharply, the hue change rate reaches 50 / second, and the switching function value is calculated to be 6.2>4.0, then the energy mapping function E(e, c)=0.6×e²+0.4×||Δc||² is constructed first, and the state deviation e=5° and the color change vector||Δc||=50 are substituted to calculate E(5, 50)=0.6×25+0.4×2500=15+1000=1015. Then the Lyapunov stability function V(e, c)=0.5×E(e, c)=507.5 is determined.

[0091] Based on the Lyapunov function, the adaptive gain is calculated. Considering the dramatic changes in the ambient color, the gain coefficient γ is set to 0.05, then k=0.05×(-dV / de)=0.05×(-0.5×2×0.6×e)=0.05×(-0.6×5)=-0.15. Therefore, the dynamic compensation factor is -0.15×e=-0.15×5=-0.75, which is output as the compensation control quantity and integrated with the original attitude control command of the spacecraft to form the final control command, which drives the attitude actuator of the spacecraft to work to reduce the attitude deviation.

[0092] The technology of the present invention originates from the limitation that the traditional spacecraft attitude control method lacks environmental perception ability. The existing technology mainly relies on sensors such as inertial measurement units and star sensors that directly measure the state of the spacecraft for closed-loop control, and rarely considers the impact of environmental color information on the control strategy, resulting in inconsistent control accuracy under different lighting environments, especially in the shadow area of ​​the earth or strong solar radiation area. The problem of control performance degradation is prone to occur. Compared with the existing technology, this application innovatively proposes an environmental color interactive feedback mechanism, which takes the color information of the spacecraft's surrounding environment as an important input, combined with traditional state feedback, to achieve environmental perception adaptive control. The starting point of the improvement is to improve the control robustness of the spacecraft in complex lighting environments, and optimize energy efficiency through a dynamic area division strategy. Finally, the improved method realizes high-precision attitude control of the spacecraft under different environmental conditions, reduces the control error caused by environmental changes, improves the success rate of space missions, and optimizes energy utilization efficiency and extends the service life of the spacecraft.

[0093] like Figure 3 As shown in the figure, the performance comparison of three different control methods in state deviation compensation is shown: traditional PID control, adaptive fuzzy control and the dynamic switching dual-zone compensation scheme proposed in this paper. It can be clearly seen from the chart that this scheme is superior to the existing technology in many key performance indicators. Specifically, in the initial response stage, the overshoot of the traditional PID reached 38%, the adaptive fuzzy control was 32%, and the overshoot of this scheme was only 26%, and the overshoot was reduced by 31.6%. In terms of system stability, the convergence time of this scheme was 5.2 seconds, which was significantly faster than the 12.5 seconds of the traditional PID and the 7.8 seconds of the adaptive fuzzy, and improved by 58.4% and 33.3%. The most outstanding is the steady-state error index. This scheme finally stabilized at an extremely low level of 0.3%, while the traditional PID and adaptive fuzzy were 5.0% and 2.0% respectively, and the steady-state accuracy was improved by 94% and 85%. Judging from the curve shape, this scheme shows a smoother transition process without obvious oscillation. After 6 seconds, the system deviation has dropped below 2% and continues to optimize to a state close to zero in the subsequent period of time, which fully demonstrates the effectiveness of the dynamic switching mechanism and dual-area compensation strategy.

[0094] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors, characterized in that: include: Collect multi-spectral color information of the environment around the spacecraft motion device in real time; Adaptively partition the multispectral color information into bands, extract multi-scale spectral features, and construct the environmental color feature mapping matrix through cross-scale correlation; Calculate the color gradient field based on the environmental color feature mapping matrix, identify the feature change nodes in the color gradient field, and divide the color gradient field into multiple feature regions; Execute the quantum color feature analysis algorithm to calculate the energy distribution between each feature area and generate the optimal motion trajectory of the spacecraft motion device; Obtain the motion parameters of the spacecraft motion device in real time, including position parameters, velocity parameters and acceleration parameters; Execute the motion state decision algorithm, perform piecewise linearization on the optimal motion trajectory, calculate the deviation through state estimation and prediction model, use variable structure compensation method combined with prediction compensation to generate motion correction, and output real-time control instructions for the spacecraft motion device; Adjust the motion state of the spacecraft motion device according to real-time control instructions; Record the motion process data of the spacecraft motion device and the environmental color change data to form an interactive feedback data set.

2. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 1, characterized in that: Adaptively partition the multi-spectral color information into bands, extract multi-scale spectral features, and construct the environmental color feature mapping matrix through cross-scale association, including: The multispectral color information includes visible light band information and near infrared band information; Adaptively partitioning the visible light band information and the near infrared band information according to color response characteristics, setting a dynamic sampling interval for each partition, and generating a non-uniformly distributed band sampling sequence; Based on the band sampling sequence, the weight coefficient is dynamically adjusted according to the spectral correlation between bands to construct an adaptive weight network; Performing a convolution operation on the adaptive weight network and the band sampling sequence to generate a spectral response feature vector between bands; Performing band energy leakage detection and crosstalk compensation on the spectral response feature vector, generating a multi-scale spectral feature descriptor through adaptive optimization, wherein the multi-scale spectral feature descriptor includes correlation features and complementary features between bands; Constructing a hierarchical feature mapping matrix based on the multi-scale spectral feature descriptor, establishing a plurality of feature subspaces in the hierarchical feature mapping matrix, each feature subspace corresponding to a scale level of the multi-scale spectral feature descriptor; Calculating the cross-scale correlation between the feature subspaces to generate a feature correlation map; The hierarchical feature mapping matrix is ​​dynamically reconstructed based on the feature association map, and an environmental color feature mapping matrix with multi-scale adaptive characteristics is output.

3. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 2, characterized in that: Band energy leakage detection and crosstalk compensation are performed on the spectral response feature vector, and a multi-scale spectral feature descriptor is generated through adaptive optimization. The multi-scale spectral feature descriptor contains correlation features and complementary features between bands, including: Based on the spectral response characteristic vector, the spectral response curves of adjacent bands are obtained, and the response curve of the central band is differentially calculated with the response curves of the adjacent bands on both sides to obtain the energy leakage coefficient between the bands; Performing bandpass filtering on the spectral response feature vector, extracting the response signal in the overlapping region of the bands, and calculating the crosstalk coefficient between the bands by cross-correlation analysis; Generate a band overlap compensation matrix based on the energy leakage coefficient and the crosstalk coefficient, and use a gradient descent method to determine energy distribution correction parameters in the band overlap region through iterative optimization; Calculating the first-order derivative and the second-order derivative of the spectral response characteristic vector at the band boundary, determining the boundary ambiguity based on the rate of change of the derivative value, and generating a band adaptive compensation parameter; Combining the band overlap compensation matrix with the band adaptive compensation parameter, performing crosstalk suppression and energy compensation on the spectral response feature vector, and obtaining a spectral response optimization feature vector; Gaussian pyramid decomposition is used to perform multi-scale decomposition of the spectral response optimization feature vector, and the band features at each scale are fused to generate a multi-scale spectral feature descriptor with balanced energy.

4. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 1, characterized in that: Execute the quantum color feature analysis algorithm, calculate the energy distribution between each feature area, and generate the optimal motion trajectory of the spacecraft motion device, including: constructing the HSV color space component of the characteristic region as a quantum state representation, and constructing the energy distribution of the characteristic region according to the quantum state representation; Determine energy operators based on the energy distribution, including a hue energy operator, a saturation energy operator, and a brightness energy operator, and perform quantum measurement to obtain energy distribution values ​​between feature regions; A parameterized variational circuit is constructed according to the energy distribution value, and the overlap between the corresponding output state and the preset trajectory optimization objective function is calculated. The variational parameters of the parameterized variational circuit are optimized through gradient descent iteration to obtain the optimized quantum state. Performing position measurement on the optimized quantum state, reconstructing the motion trajectory of the spacecraft motion device, constructing a velocity function and an acceleration function, and respectively calculating the expected value of the optimized quantum state under the velocity function and the acceleration function; constructing a decoherence compensation matrix according to the expected value, combining the optimized quantum state with the decoherence compensation matrix to obtain a quantum state density matrix, constructing a constraint verification function, and calculating the constraint satisfaction of the quantum state density matrix; The motion trajectory is corrected according to the constraint satisfaction to obtain the corrected motion trajectory, the corresponding optimal quantum state density matrix is ​​determined, and the energy consumption of the system is calculated based on the optimal quantum state density matrix and the energy operator to generate the optimal motion trajectory of the spacecraft motion device that meets the dynamic constraints.

5. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 4, characterized in that: Performing position measurement on the optimized quantum state and reconstructing the motion trajectory of the spacecraft motion device includes: Performing position measurement on the optimized quantum state, obtaining a quantum state position representation, calculating a position measurement probability density, and calculating a position measurement expectation value based on the position measurement probability density; Calculate the local variance of the expected value of the position measurement, determine the bandwidth parameter of the wavelet-Gaussian mixture filter, and construct a wavelet decomposition sequence and a Gaussian filter kernel function; sequentially pass the expected value of the position measurement through the wavelet decomposition sequence to obtain multi-scale coefficients, perform convolution operation on the multi-scale coefficients and the Gaussian filter kernel function, and reconstruct the filtered position measurement data; Calculating a weight coefficient matrix according to the expected value of the position measurement, performing weighted calculation on the filtered position measurement data and the weight coefficient matrix, and constructing an initial trajectory function; Calculating the Laplace operator of the initial trajectory function, modulating the product of the initial trajectory function and the Laplace operator with a smoothing coefficient and superimposing the result on the initial trajectory function to obtain a smoothed trajectory function; Calculate the statistical error function, instrument error function and noise error function respectively, and construct the error compensation function based on the corresponding weight coefficients; The smooth trajectory function and the error compensation function are operated to obtain a final reconstructed spacecraft motion trajectory.

6. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 1, characterized in that: Execute the motion state decision algorithm, perform piecewise linearization on the optimal motion trajectory, calculate the deviation through state estimation and prediction model, use variable structure compensation method combined with prediction compensation to generate motion correction, and output real-time control instructions for spacecraft motion devices, including: The optimal motion trajectory is piecewise linearized in a multidimensional state space to construct a state transition sequence, each state transition interval containing a state feature node; Decoupling the motion parameters to generate independent state components, and obtaining state estimation values ​​through recursive filtering; A prediction model is constructed based on the state estimation value, the state evolution sequence in the prediction time domain is calculated, and the state deviation curve is obtained by matching the state transition sequence. The state deviation curve is processed by a variable structure compensation method, the compensation area is divided by a dynamic switching function, the corresponding compensation amount is calculated respectively, and the compensation control amount is output; At discrete control timing points, a predicted compensation amount is calculated based on the prediction model, an immediate compensation amount is calculated in combination with real-time state feedback, and the predicted compensation amount, the immediate compensation amount and the compensation control amount are merged to generate a motion correction amount; Based on the motion correction amount, a piecewise continuous nonlinear mapping structure is constructed as a speed response function, a steering response function with a state coupling term is constructed, and a speed control amount and a steering control amount are obtained; The speed control amount and the steering control amount are converted into speed control instructions and steering control instructions, and synthesized into real-time control instructions for a spacecraft motion device.

7. The method for collecting and analyzing interactive feedback between a spacecraft motion device and environmental colors according to claim 6, characterized in that: The state deviation curve is processed by a variable structure compensation method, and the compensation area is divided by a dynamic switching function. The corresponding compensation amounts are calculated respectively, and the output compensation control amounts include: Calculating the rate of change of the state deviation curve, and constructing a dynamic switching function according to the state deviation and the rate of change of the state deviation; Dividing the compensation area into a first compensation area and a second compensation area based on the output value of the dynamic switching function, the first compensation area corresponds to an area where the deviation is less than a preset threshold, and the second compensation area corresponds to an area where the deviation is greater than the preset threshold; In the first compensation region, a quadratic function of the state deviation is used to perform linear compensation and calculate a linear compensation amount; In the second compensation region, constructing an energy mapping function of the state deviation, determining a Lyapunov stability function based on the energy mapping function, and calculating an adaptive gain using the Lyapunov stability function to form a dynamic compensation factor; The linear compensation amount or the dynamic compensation factor is output as a compensation control amount.

Images