-
为了验证仿真模型的真实性和准确性,选取几组不同拍摄时间、探测方向、虚拟目标位置的仿真参数,模拟不同日照条件下的仿真图像。算法运行环境为Win10 64位操作系统,Intel i7-9700CPU,主存32 GB,软件平台为MATLAB 2012b。模型中的探测器像素尺寸为320×256,像元间距为20 μm,探测波段为8~14 μm。卫星处于东经60°上空。仿真参数设置见表1,仿真效果如图4、图5所示。
Parameters (a) (b) (c) (d) (e) (f) (g) (h) (i) (j) (k) (l) Focal length/mm 20 20 20 20 40 40 40 40 60 60 60 80 Obliquity of satellite orbit/rad 0 0 0 0 20 20 20 20 0 20 0 20 Altitude of satellite orbit/km 2 000 2 000 2 000 2 000 10 000 10 000 10 000 10 000 20 000 20 000 20 000 20 000 Longitude of virtual target/rad 5π/12 5π/12 7π/12 7π/12 5π/12 5π/12 7π/12 7π/12 5π/12 5π/12 7π/12 7π/12 Latitude of virtual target/rad π/18 π/6 π/18 π/6 π/18 π/6 π/18 π/6 π/18 π/6 π/18 π/6 Altitude of virtual target/km 100 200 200 200 500 500 3 000 3 000 200 200 5 000 5 000 Longitude of direct sunlight/rad 0 2π/3 4π/3 0 2π/3 4π/3 0 2π/3 4π/3 0 2π/3 4π/3 Latitude of direct sunlight/rad −π/9 −π/9 −π/9 0 0 0 π/9 π/9 π/9 −π/9 0 π/9 Table 1. Simulation parameters
图4所示为不考虑噪声的仿真图像,图5所示为考虑传感器传输噪声与镜头振动的仿真图像。图中可以明显区分海洋、云和深空因红外辐射特性的不同而导致的图像灰度差异。云的图像存在单方向的条纹,离地球边缘越近,条纹现象越明显。在倾斜角度拍摄真实的云图时,存在单方向的条纹特性,而在地球临边观测中,同样采用倾斜角度来观测云,同样出现了条纹,符合观测经验。
使用拉普拉斯算子和(Laplacian Sum, LS)、灰度梯度(Grayscale Mean Gradient, GMG)评价仿真图像的清晰度[14]。拉普拉斯算子和比较了3×3 pixel邻域内的像素偏差程度。首先使用拉普拉斯算子对图像求卷积,该算子定义如下:
基于拉普拉斯算子和的图像清晰度定义为:
式中:$ \left|G\left(i,j\right)\right| $为像素点$ \left(i,j\right) $处的拉普拉斯算子卷积;M、N为图像的行数和列数;LS值越大,图像信息越丰富,边缘越锐利,图像质量越好。
灰度梯度函数反映了图像的细节和纹理变化特征,定义为:
式中:$ f\left(i,j\right) $为像素点$ \left(i,j\right) $处的图像灰度;GMG值越大,高频信息越丰富,图像越清晰。
对仿真实验结果进行图像清晰度评价,分别计算图像的拉普拉斯算子和、灰度梯度,其结果见表2。不含噪声的仿真结果图像的平均拉普拉斯算子和(LS值)为0.15,平均灰度梯度(GMG值)为0.70。含噪声图像的LS值为0.14,GMG值为0.68。仿真图像边缘易检测,高频信息丰富,画质清晰。传感器噪声对图像质量的影响较小。图像仿真时间见表2第6列,平均单帧图像仿真时间为123.45 s,运行速度较快。
Evaluation index
image in figuresLS GMG Runtime/s Fig.4 Fig.5 Fig.4 Fig.5 (a) 0.085 0.078 0.34 0.33 98.57 (b) 0.090 0.080 0.44 0.40 98.71 (c) 0.11 0.10 0.50 0.50 164.22 (d) 0.13 0.11 0.71 0.61 87.30 (e) 0.13 0.12 0.58 0.53 105.50 (f) 0.14 0.12 0.63 0.57 111.58 (g) 0.17 0.15 0.81 0.76 154.30 (h) 0.16 0.15 0.85 0.77 108.92 (i) 0.22 0.23 0.98 1.16 156.75 (j) 0.19 0.17 0.94 0.85 139.99 (k) 0.19 0.19 0.93 0.99 144.22 (l) 0.14 0.13 0.68 0.68 111.33 Average 0.15 0.14 0.70 0.68 123.45 Table 2. Image evaluation metrics
Infrared remote sensing imaging simulation method for earth’s limb scene
doi: 10.3788/IRLA20210896
- Received Date: 2021-11-20
- Rev Recd Date: 2021-12-25
- Available Online: 2022-03-04
- Publish Date: 2022-02-28
-
Key words:
- infrared physics /
- image of limb scene /
- remote sensing imaging simulation /
- atmospheric infrared radiation /
- cloud scene
Abstract: Simulation of earth’s limb scene plays a key role in satellite infrared detection field. It is an important basis for long-range detection of high-speed airborne targets. In limb detection, the traditional infrared ocean simulation method based on three-dimensional ocean appearance and the calculation of radiation characteristics is not applicable, because the earth surface approximates a sphere. Also, the thickness and height of clouds have important influence on the calculation of infrared radiative transmission characteristics, where the method of considering the cloud as particle cluster would greatly reduce the speed of simulation. Therefore, the infrared remote sensing imaging simulation method for earth’s limb scene was established by conducting the infrared radiation model of ocean and cloud, the transformation relationship between earth-space coordinate system and infrared camera coordinate system, and the atmospheric transmission model. According to the components of scene, the ocean distribution model and multi-layer clouds distribution model were established respectively, and the infrared radiation model of the earth’s limb scene was established according to the infrared radiation and reflection characteristics of ocean and clouds. The infrared remote sensing simulation images of the earth’s limb scene under various observation angles were calculated by the conversion relationship between earth-space coordinate system and camera coordinate system, the theory of atmospheric transmission and the sensor effect. The simulation results show that the infrared image accord with the infrared radiation characteristics of earth’s limb scene. The average Laplacian sum of simulation images is 0.15, and the grayscale gradient average value of the images is 0.70.