Relative radiometric calibration of yaw data without calibration field of hyperspectral images based on harmonic analysis

Abstract

ABSTRACT Radiometric calibration of satellite sensor could be divided into absolute radiometric calibration and relative radiometric calibration. Relative radiometric calibration can greatly reduce the impact of the sensor’s instability due to many factors like failure to perform calibration experiments or satellite post-operative attenuation. One of the methods is to use the 90° yaw data for relative radiation calibration, where the camera or satellite rotates by 90° so that the linear charge-coupled device (CCD) detector is parallel to the flight direction and sequentially passes through the ground uniform field to obtain the same radiance to achieve the satellite in order to achieve relative radiometric calibration of satellite sensor and eliminate stripe phenomenon in image. The orbita hyperspectral satellites (OHS) are commercial hyperspectral remote sensing satellites with the largest image width and the highest spatial resolution in China. It has realized the operation of multiple hyperspectral satellites network for the first time. However, there are still problems such as inaccurate calculation of radiation calibration parameters, difficulty in realizing uniform field ground, and difficulty in achieving full dynamic calibration of the sensor. In this study, the hyperspectral images acquired by OHS satellites were taken as the research object to propose a method to achieve field-free relative radiometric calibration based on harmonic analysis of yaw data. This method combines the advantages of harmonic analysis technology and the least squares theory and is applied to the relative radiation calibration of hyperspectral yaw data. It is called the harmonic analysis radiometric calibration (HARC) algorithm. First, HARC algorithm performed the first-order forward and reverses harmonic analysis on the 86° yaw data of OHS satellites which were pre-processed into the data of the conventional image mode to obtain the theoretical value of the pre-processed yaw data. Then, the actual radiance value of yaw data in column direction would be fitted to the theoretical value using the least squares theory to calculate the offset and gain of radiometric calibration parameter after eliminating the high and low radiance values in yaw data. Finally, the offset and gain parameters were used to carry out relative radiometric calibration on conventional push-broom satellite image. The comparative analysis of the test showed that relative radiometric calibration of all band images of six pieces of CCD arrays in the hyperspectral sensor of OHS satellites was realized based on yaw data without uniform ground calibration field. The HARC algorithm proposed in this paper can overcome the inaccurate calculation of calibration parameters caused by the yaw angle less than 90°. It not only retains the image detail information but also effectively removes various types of thick and thin stripe noise in the sweep direction, and is not limited by the different bands. The algorithm has good universality in applications on various bands. Qualitative and quantitative comparison analyses were conducted between HARC algorithm and classical relative radiometric calibration methods, and the corrected image has a great improvement in image quality and sharpness compared with the original image. The effect of stripe removal and the accuracy of radiation correction of the HARC algorithm are superior to the classical relative radiation calibration method.