-
在高光谱成像仪电子学设计技术流程中,筛选短波红外探测器阶段在组装之前,对短波红外高光谱成像仪的坏像元进行识别后,然后按其特征进行分类,这样为有效的修正提供了便利。短波红外高光谱成像仪采用定制大面阵制冷型探测器(外观如图8所示)。
图 8 制冷型1000×256大面阵高速短波红外探测器
Figure 8. Outline drawing of 1 000×256 refrigerated large area array shortwave infrared detector
按照短波红外高光谱成像仪的设计工作模式,测试增益模板中其不同增益(G表示Gain,分别取值 G1、G2、G3、G4、G5)下的图像,统计面阵图像中的坏像元数量;再按照帧频模板中工作模式,测试帧频变化范围内图像的响应值,特别保留选取模板帧频为180、220、260、300、340、380 fps的图像,复核增益模板。对两次模板统计出的坏像元,按类分别筛查其面阵坐标位置,生成坏像元坐标文件。图9是设计模板检测坏像元的流程图。
暗背景下采集图像,识别亮点像元如图10所示。
通过增益模板的检测,可检测出短波红外高光谱成像仪图像中亮点像元、暗点像元、弱响应像元、非线性响应像元[12]。
通过对星载干涉成像短波红外高光谱成像仪进行亮点及暗点像元检测,在12 bit量化下,在增益4下,图像响应灰度值在2500附近时,检测结果最为准确。
图 9 基于增益模板和帧频模板的坏像元识别流程图
Figure 9. Flow chart of bad pixel identification based on gain template and frame rate template
一般用来表示坏像元检测准确率的参数有两个,分别是:漏检测系数(Detection Missed Factor):定义为漏检坏像元数占比实际坏像元数的百分比。虚假检测系数(Detection False Factor):定义为虚假检测坏像元数占比实际坏像元数的百分比。
$$ {{{N}}}_{\text {miss }}=\left(1-\frac{b}{d+h+w}\right) {\text{%}} $$ (1) 式中:b代表所采用方法检测到的坏像元;d表示实际没有任何响应的暗点像元;h表示响应值过高的亮点像元;w表示有响应但响应辐照不随正常像元随曝光时间增加呈现线性响应的像元。
$$ {N}_{\mathrm{err}}=\left(\frac{a}{M \times N-d-h-w}\right) {\text{%}} $$ (2) 式中:a代表被所用方法检测出的坏像元;d表示实际没有任何响应的暗点像元;h表示响应值过高的亮点像元;w表示对光照有响应但响应辐照不随正常帧长像元呈现线性响应的像元;分母中M表示图像面阵行数, N表示图像列数[13],在文中项目中M=256,N=1000。
图 10 高速高光谱成像仪不同增益下亮元及暗元检测
Figure 10. Bright pixels and dark pixels recognition for high speed hyperspectral imager at different gains
为了比对本方法的便捷性,通过对比同一图像数据处理工程师编写实现的三个方法在对相同的5组图像数据进行坏像元识别时所附加消耗的时长(见表1,运算用计算处理平台相同),考核处理防范的便捷性。方法1:记作M1,代表文中提出的模板法;方法2:记作M2,同一软件设计师参考标准流程编写的“滑动窗口法”;方法3:记作M3,同一设计师参考标准流程编写“图像差分域识别法”。附加时间指的是,软件不识别短波红外图像用时与读入并遍历显示一遍对应图像数据进行坏像元识别所额外占用的附加时间,所增加的额外附加时间。
表 1 采用短波红外图像三种方法检测坏像元的附加用时对比
Table 1. Comparing additional times of three bad pixels detection methods using in SWIR images
Image data# Time of M1/s Time of M2/s Time of M3/s Data1 2.283 9.573 12.796 Data2 2.265 9.591 12.889 Data3 2.236 9.599 12.989 Data4 2.267 9.577 12.799 Data5 2.273 9.594 12.896
Identification of bad pixels in shortwave infrared high-speed hyperspectral imager for spaceborne remote sensing
-
摘要: 为改善干涉成像短波红外高速高光谱成像仪的坏像元对复原光谱的影响,利用高光谱成像仪测试流程建立了坏像元识别模板,以提高坏像元识别效率。首先,按照高光谱成像仪测试流程设置增益模板和帧频模板并采集图像数据,依据正常像元增益响应设定合理判定阈值Th1,识别不同增益下异常像元并记录对应坐标值;再依据正常像元帧频响应灰度值设定合理判定阈值Th2,识别不同帧频下异常像元并记录坐标值。最后,对比增益模板和帧频模板判定的异常像元,融合确定坏像元。实验结果表明基于增益模板和帧频模板的识别方法在不增加设备研制测试成本的同时有效识别出短波红外高光谱成像仪探测器的坏像元,为可靠识别短波红外高光谱成像仪坏像元提供了一种低成本、高效可靠的新方法,提高了干涉成像高光谱成像仪光谱反演准确性。Abstract: To reduce the influence of bad pixels in large-aperture interferometric imaging shortwave infrared high-speed hyperspectral imagers on recovering spectra, a bad pixel identification template was established using a hyperspectral imager test process to effectively raise the efficiency of identification of bad pixels. First, image data were collected based on the gain template and frame-rate template for the hyperspectral imager test. Then, the judgment threshold Th1 was set reasonably according to the gain response of normal pixels to identify abnormal pixels under different gains and record the corresponding coordinate values, and the judgment threshold Th2 was set reasonably according to the frame-rate response gray values of normal pixels to identify abnormal pixels under different frame rates and record the coordinate values. Finally, after the abnormal pixels identified by the gain template and those identified by the frame-rate template were compared, they were used together to identify bad pixels. The experimental results show that the identification method based on the gain and frame-rate templates can effectively identify bad pixels in a shortwave infrared hyperspectral imager detector without increasing the cost of equipment development and testing, while providing an economical, practical, efficient, and reliable technical means for correcting bad pixels in a shortwave infrared hyperspectral imager. The method also offers a useful reference to improve the accuracy of inversion of interferometric imaging hyperspectral imager spectral data.
-
Key words:
- bad pixel recognition /
- shortwave infrared /
- interference imaging /
- hyperspectral imager
-
表 1 采用短波红外图像三种方法检测坏像元的附加用时对比
Table 1. Comparing additional times of three bad pixels detection methods using in SWIR images
Image data# Time of M1/s Time of M2/s Time of M3/s Data1 2.283 9.573 12.796 Data2 2.265 9.591 12.889 Data3 2.236 9.599 12.989 Data4 2.267 9.577 12.799 Data5 2.273 9.594 12.896 -
[1] Chen Jianjun, Cui Jicheng, Liu Jianan, et al. Correction technology of HgCdTe short-wave infrared focal plane arrays [J]. Acta Optica Sinica, 2019, 39(2): 0204001. (in Chinese) doi: 10.3788/AOS201939.0204001 [2] Li Yingying, Zhao Yongchao, Wu Hao, et al. Detection and correction of bad pixels in large aperture static imaging spectrometer [J]. Spacecraft Recovery & Remote Sensing, 2015, 36(5): 76-82. (in Chinese) [3] Yu Yi, Chang Songtao, Wang Min, et al. Fast non-uniformity correction for high dynamic infrared radiometric system [J]. Optics and Precision Engineering, 2015, 23(7): 2315-2319. (in Chinese) [4] Huang Fuyu, He Yongqiang, Zhou Bing, et al. A new method of bad pixels detection based on SWIR FPA [J]. Infrared Technology, 2010, 32(11): 611-614. (in Chinese) [5] Li Qian, Yuan Yan, Cao Jun. Bad pixel correction in temporally and spatially modulated fourier transform imaging spectrometer [J]. Spacecraft Recovery & Remote Sensing, 2016, 37(6): 76-84. (in Chinese) [6] Wang Rongbin, Zeng Chao, Jiang Wan, et al. Terra MODIS band 5th stripe noise detection and correction using MAP-based algorithm [J]. Infrared and Laser Engineering, 2013, 42(1): 273-277. (in Chinese) [7] Huang Fuyu, He Yongqiang, Zhou Bing, et al. A new method of bad pixels detection based on SWIR FPA [J]. Infrared Technology, 2010, 32(10): 611-614. (in Chinese) [8] Hu Mingpeng, Ma Dongmei, Liu Zhixiang, et al. Study of IRFPA non-effective pixel discrimination [J]. Semiconductor Optoelectronics, 2008, 29(2): 184-187. (in Chinese) [9] Zhang Jinyue, Lv Junpeng, Ni Zhenhua. Highly sensitive infrared detector based on a two-dimensional heterojunction [J]. Chinese Optics, 2021, 14(1): 87-96. (in Chinese) [10] Li Yanjun, Cui Ruiqing, Zhao Guilan, et al. A new method of bad pixel detection and compensation of IRFPA [J]. Laser & Infrared, 2007, 37(1): 51-52. (in Chinese) [11] Lai Rui, Liu Shangqian, Zhou Huixin, et al. A new method of bad pixel detection and compensation of IRFPA [J]. Semiconductor Optoelectronics, 2005, 26(3): 51-52. (in Chinese) [12] Liang Chao, Ma Tianxiang. Design of infrared imaging nonuniformity correction system based on black body calibration [J]. Chinese Optics, 2016, 9(3): 385-392. (in Chinese) [13] Zhang Changxing, Liu Chengyu, Qi Hongxing, et al. Blind pixel detection algorithm using spectral matching for thermal infrared hyperspectral imager [J]. Infrared and Laser Engineering, 2020, 49(1): 0104002. (in Chinese) [14] Zhu Shuangshuang, Wu Yang, Zou Peng, et al. Environmental adaptability analysis and performance optimization of space-borne infrared detector assembly [J]. Infrared and Laser Engineering, 2020, 49(2): 0204001. (in Chinese) [15] Huang Liang, Li Mingxuan, Lv Hengyi, et al. Large-range and high-resolution temperature measurement system for satellite-borne infrared detector [J]. Optics and Precision Engineering, 2019, 27(11): 2315-2319. (in Chinese)