CN116347247A - Linear array space remote sensing load automatic exposure method based on Bayesian optimization - Google Patents
Linear array space remote sensing load automatic exposure method based on Bayesian optimization Download PDFInfo
- Publication number
- CN116347247A CN116347247A CN202211666313.1A CN202211666313A CN116347247A CN 116347247 A CN116347247 A CN 116347247A CN 202211666313 A CN202211666313 A CN 202211666313A CN 116347247 A CN116347247 A CN 116347247A
- Authority
- CN
- China
- Prior art keywords
- optimal
- remote sensing
- load
- linear array
- automatic exposure
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Photometry And Measurement Of Optical Pulse Characteristics (AREA)
Abstract
A Bayesian optimization-based linear array space remote sensing load automatic exposure method relates to the technical field of space optical remote sensing imaging, and solves the problem that linear array space remote sensing load push broom imaging is difficult to realize automatic exposure. The invention provides an imaging quality evaluation method based on two-dimensional entropy, which comprises the steps of setting a detector to obtain seed image data under 1-level integral series and 1-time gain, establishing a numerical mapping model of irradiance of ground objects and image pixel gray scale under different continuous integral series and different gains by combining a load response function, and calculating image quality values under different integral series and different gains according to the seed image data and the established numerical mapping model; and fitting the calculated image quality value with an exposure quality curved surface by using Bayesian optimization, and selecting the optimal imaging parameters according to the exposure quality curved surface. The method has a certain engineering practice significance for improving the on-orbit application efficiency of the space remote sensing load.
Description
Technical Field
The invention relates to the technical field of space optical remote sensing imaging, in particular to a linear array space remote sensing load automatic exposure method based on Bayesian optimization.
Background
The aerospace optical remote sensing technology is one of important means for earth observation, and has important effects on disaster response, resource survey, urban and rural planning and the like. The optical remote sensing load can be divided into an area array and a linear array according to the detector, and corresponds to a staring imaging mode push-broom imaging mode. In order to realize the high-efficiency acquisition of clear remote sensing images, extensive researches on an aerospace optical load automatic exposure method are carried out at home and abroad. For the staring imaging mode of the area array detector, the multi-frame images in the same scene can be analyzed, and a mathematical model of imaging parameters and image quality is established to realize automatic exposure; because of the particularity of single imaging of the linear array detector in the push-broom imaging mode and the limitation of an on-board hardware system, real-time automatic exposure is difficult to realize, and domestic and foreign scientific research institutions currently widely adopt a load combination and constellation cooperation method to realize automatic exposure of the linear array space remote sensing load, and the multi-star and multi-load mode greatly improves the development cost and period.
Therefore, in order to solve the problem of automatic exposure in a push-broom imaging mode, the invention provides a linear array space remote sensing load automatic exposure method based on Bayesian optimization, so as to realize single-star single-load optimal integral series and gain selection, obtain a high-quality original remote sensing image and improve the space optical load on-orbit application efficiency.
Disclosure of Invention
The invention provides a linear array space remote sensing load automatic exposure method based on Bayesian optimization, which aims to solve the problem that the current space optical remote sensing field is difficult to realize the single-star single-load push-broom imaging mode automatic exposure.
A Bayesian optimization-based linear array space remote sensing load automatic exposure method comprises the following steps:
step one, setting a detector to obtain seed image data under 1-stage integration series and 1-time gain;
step two, combining load response functions to establish a numerical mapping model of irradiance of ground features and pixel gray level of images under different continuous integration series and different gains;
step three, calculating image quality values under different integration levels and different gains according to the seed image data obtained in the step one and the numerical mapping model established in the step two;
and step four, optimizing and fitting the image quality value calculated in the step three by using Bayesian, and selecting the optimal imaging parameters according to the exposure quality curved surface.
The invention has the beneficial effects that:
the automatic exposure method for the linear array space remote sensing load based on Bayesian optimization provided by the invention is used for rapidly and accurately determining the optimal imaging parameters of a push-broom mode, and avoiding revisit period caused by imaging failure. The method has good sensibility to the information quantity, brightness and edge sharpness of the image, has higher tolerance to noise level, can realize synchronous active control of imaging parameter sets, and has the characteristics of small operation quantity and high operation speed. The automatic exposure method is suitable for the linear array space remote sensing load, and can be popularized to the area array and new system space remote sensing load or other types of cameras.
Drawings
Fig. 1 is a schematic block diagram of a linear array space remote sensing load automatic exposure method based on Bayesian optimization.
Detailed Description
Referring to fig. 1, the embodiment of the linear array space remote sensing load automatic exposure method based on bayesian optimization is described, and specific steps of the method comprise low-level number, gain image acquisition 1, different integration levels, gain image synthesis 2, image prime number value evaluation 3, imaging quality curved surface bayesian fitting 4, optimal imaging parameter selection and setting 5, optimal image acquisition 6 and the like. The specific process of this embodiment is as follows:
step one, setting a detector to obtain seed image data under 1-stage integration series and 1-time gain;
firstly, exposing and reading out image data by using pixels of the first rows of the detector when the linear array space load is used for executing shooting task passing. Two-dimensional entropy is adopted as an important criterion of image quality, and pixels I in an image i,j The two-dimensional entropy at this point is:
wherein GS i,j Represents the gray level corresponding to pixel (i, j),
the probability that the corresponding gray level of pixel (i, j) is in the 9 x 9 domain is represented.
Carrying out normalization processing on the obtained two-dimensional entropy matrix, defining an entropy weight function to distinguish entropy layers, and simultaneously smoothing the influence of noise on image quality evaluation:
dividing the entropy matrix by maximum inter-class variance (OSTU) method, taking the part with small entropy as Saturated Region (SR), i.e. overexposed region and underexposed region, taking the ratio of SR pixels in the whole image as image quality evaluation index, and marking as Q I The smaller the value, the higher the image quality, and the stronger the post-processing operability.
Step two, combining a load response function CRF, and establishing a numerical mapping model of irradiance of ground features and image pixel gray scale under different continuous integration series and different gains;
step three, calculating image quality values under different integration levels and different gains according to the seed image data obtained in the step one and the numerical mapping model established in the step two;
and step four, optimizing and fitting the image quality value calculated in the step three by using Bayesian, and selecting the optimal imaging parameters according to the exposure quality curved surface.
In the present embodiment, the load response function CRF is:
DN＝G[f(EMΔt)+δ] (3)
DN represents the pixel gray level output by the system, G represents the gain, E is the irradiance of the entrance pupil, M is the number of integration stages, deltat is the integration time, deltais the bias term, and f is the detector responseThe response function. Δt is determined by the image shift speed, is usually constant, and δ is zeroed after correction. Synthesizing images I under different imaging parameters through CRF 0 ,……，I n Then, adopting Bayesian optimization to predict imaging quality distribution under different imaging parameters:
μ(x n+1 )＝k(x n ) T (K+σ 2 (x n )I) -1 y n (4)
σ 2 (x n+1 )＝k(x n ,x n )-k(x n ) T [K+σ 2 (x n )I] -1 k(x n ) (5)
wherein μ and σ are the mean and covariance of the imaging quality, respectively; x= (X 0 ,……,x n ) And y= (Y) 0 ,……,y n ) All are training points input, and any one of X is n ＝[M,G]M is the number of integration stages, G is the gain, y n For its corresponding imaging quality; k (x) n )＝k(x n ,x)，K＝k(x n ,x n+1 ) I is an identity matrix; where k (x, y) represents an exponential square kernel function, defined as
In the formula, the optimal integral series and gain are determined through the distributed global minimum points:
X optimal ＝arg minμ＝[M optimal ,G optimal ] (6)
wherein X is optimal Is the optimal value, M optimal For the optimal integral series, G optimal Optimum gain.
And adjusting imaging parameters of a load system, and acquiring an optimal exposure image by utilizing the number of stages which are not utilized by the detector in the load transit time, so as to realize the on-orbit automatic exposure of the linear array load.
Aiming at the problem of automatic exposure of the space remote sensing load of the linear array, the embodiment provides a grouping utilization strategy of the integral series of the linear array detector, and establishes a numerical mapping model of irradiance of ground objects and gray scales of image pixels under different continuous integral series and different gains by combining a load response function CRF, wherein the model has good sensitivity to information quantity, brightness and edge sharpness, and the noise level is effectively balanced through an entropy weight function in the model; an imaging parameter (integration series and gain) synchronous active control method is provided, and the two main parameters are usually independently or sequentially determined at present; the numerical mapping model based on CRF is established, the cost of shooting a real image in the image quality evaluation process is reduced, and the optimal imaging parameters of the load are found through the Bayesian optimization internal prediction step, so that automatic exposure is realized. High cost and long period caused by multiple stars and multiple loads are avoided, so that the linear array space remote sensing load has automatic exposure capability, and high-quality original remote sensing image acquisition in a push-broom mode is realized.
The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalent substitutions, improvements, etc. within the spirit and principles of the present invention should be included in the scope of the present invention.
Claims (3)
1. A Bayesian optimization-based linear array space remote sensing load automatic exposure method is characterized by comprising the following steps of: the exposure method is realized by the following steps:
step one, setting a detector to obtain seed image data under 1-stage integration series and 1-time gain;
step two, combining load response functions to establish a numerical mapping model of irradiance of ground features and pixel gray level of images under different continuous integration series and different gains;
step three, calculating image quality values under different integration levels and different gains according to the seed image data obtained in the step one and the numerical mapping model established in the step two;
and step four, optimizing and fitting the image quality value calculated in the step three by using Bayesian, and selecting the optimal imaging parameters according to the exposure quality curved surface.
2. The automatic exposure method for the linear array space remote sensing load based on Bayesian optimization according to claim 1, wherein the method is characterized by comprising the following steps: in the second step, the formula of the load response function CRF is expressed as follows:
DN＝G[f(EMΔt)+δ]
in the formula, DN is the pixel gray level output by the system, G is the gain, E is the irradiance of the entrance pupil, M is the number of integration stages, deltat is the integration time, deltais the bias term, and f is the response function of the detector.
3. The automatic exposure method for the linear array space remote sensing load based on Bayesian optimization according to claim 1, wherein the method is characterized by comprising the following steps: in the fourth step, the image I under different imaging parameters is synthesized through CRF 0 ，......，I n The imaging quality distribution under different imaging parameters is predicted through Bayesian optimization:
μ(x n+1 )＝k(x n ) T (K+σ 2 (x n )I) -1 y n
σ 2 (x n+1 )＝k(x n ，x n )-k(x n ) T [K+σ 2 (x n )I] -1 k(x n )
wherein μ and σ are the mean and covariance of the imaging quality, respectively; x= (X 0 ，..，x n ) And y= (Y) 0 ，......，y n ) All are training points input, and any one of X is n ＝[M，G]M is the number of integration stages, G is the gain, y n For its corresponding imaging quality, k=k (x n ，x n+1 ) I is an identity matrix;
determining the optimal integral series and gain through the distributed global minimum points:
X optimal ＝arg minμ＝[M optimal ，G optimal ]
wherein X is optimal Is the optimal value, M optimal For the optimal integral series, G optimal An optimal gain;
and adjusting imaging parameters of a load system, and acquiring an optimal exposure image by utilizing the number of stages and gain which are not utilized by the detector in the load transit time, so as to realize the on-orbit automatic exposure of the linear array load.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211666313.1A CN116347247A (en) | 2022-12-23 | 2022-12-23 | Linear array space remote sensing load automatic exposure method based on Bayesian optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211666313.1A CN116347247A (en) | 2022-12-23 | 2022-12-23 | Linear array space remote sensing load automatic exposure method based on Bayesian optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116347247A true CN116347247A (en) | 2023-06-27 |
Family
ID=86882866
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211666313.1A Pending CN116347247A (en) | 2022-12-23 | 2022-12-23 | Linear array space remote sensing load automatic exposure method based on Bayesian optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116347247A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113079323A (en) * | 2021-03-31 | 2021-07-06 | 中国科学院长春光学精密机械与物理研究所 | Space remote sensing load automatic exposure method based on two-dimensional entropy |
| CN114339064A (en) * | 2021-12-03 | 2022-04-12 | 南京仙电同圆信息科技有限公司 | Bayesian optimization exposure control method based on entropy weight image gradient |
-
2022
- 2022-12-23 CN CN202211666313.1A patent/CN116347247A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113079323A (en) * | 2021-03-31 | 2021-07-06 | 中国科学院长春光学精密机械与物理研究所 | Space remote sensing load automatic exposure method based on two-dimensional entropy |
| CN114339064A (en) * | 2021-12-03 | 2022-04-12 | 南京仙电同圆信息科技有限公司 | Bayesian optimization exposure control method based on entropy weight image gradient |
Non-Patent Citations (3)
| Title |
|---|
| 曹旗磊;赵明;董丽丽;张明波;许文海;: "航天遥感相机及参数优化方法", 激光与光电子学进展, no. 03 * |
| 王德江;匡海鹏;: "模拟增益对电荷耦合器件信噪比与动态范围影响的实验研究", 物理学报, no. 07 * |
| 郑亮亮;金光;曲宏松;张贵祥;: "多通道宽响应域TDI CCD成像系统的非均匀性校正", 光学学报, no. 11 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108038445B (en) | SAR automatic target identification method based on multi-view deep learning framework | |
| CN104065892B (en) | A kind of relative radiometric correction method for staring satellite area array CCD camera | |
| JP5470959B2 (en) | Multi-frame reconstruction method, system and software | |
| Zhong et al. | An adaptive subpixel mapping method based on MAP model and class determination strategy for hyperspectral remote sensing imagery | |
| CN112766199A (en) | Hyperspectral image classification method based on self-adaptive multi-scale feature extraction model | |
| CN101783888A (en) | Automatic exposure method based on class histogram | |
| CN106791508B (en) | A kind of method of adjustment and adjustment system of numeric field TDI camera imaging quality | |
| CN104899870A (en) | Depth estimation method based on light-field data distribution | |
| CN102891966A (en) | Focusing method and device for digital imaging device | |
| CN112613397B (en) | Method for constructing target recognition training sample set of multi-view optical satellite remote sensing image | |
| CN104486562B (en) | Embedded infrared image superframe processing method based on the fixed time of integration | |
| CN110647932B (en) | Planting crop structure remote sensing image classification method and device | |
| CN103489171A (en) | Wide-range remote-sensing image automatic dodging and color uniformizing method based on standard color library | |
| CN110120018A (en) | A kind of face battle array in-orbit relative radiometric calibration method of high dynamic range noctilucence imaging satellite | |
| CN101635782A (en) | Method and device for obtaining image based on dynamic time delay integral | |
| CN109120848A (en) | A kind of space camera integration stages number adjusting method | |
| CN118172499B (en) | Building height inversion method based on resource third-order remote sensing image | |
| CN116310852A (en) | Double-time-phase remote sensing image unsupervised classification and change detection method and system | |
| CN111583330A (en) | Multi-scale space-time Markov remote sensing image sub-pixel positioning method and system | |
| CN113079323B (en) | Space remote sensing load automatic exposure method based on two-dimensional entropy | |
| CN116347247A (en) | Linear array space remote sensing load automatic exposure method based on Bayesian optimization | |
| CN113160311A (en) | Solar cell panel adjusting method and device, sun tracking system and storage equipment | |
| CN113570523B (en) | Method for automatic generation and extrapolation of relative radiation correction coefficients for optical images | |
| CN113034555B (en) | Feature fine matching method based on minimum spanning tree and application | |
| CN114882385A (en) | Method for counting wheat ears in field based on unmanned aerial vehicle platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |