METHOD OF PRIMARY PROCESSING OF MULTISSPECTRAL IMAGES OF THE EMBEDDED SYSTEM OF A NANOSATELLITE

Abstract

Currently, CubeSat satellites enable universities, public and private companies to conduct scientific research in zero gravity at the altitude of low Earth orbit. The work considers a nanosatellite carrying a camera as a payload that occupies one module (1U). The main purpose of the camera is monitoring the use of Earth’s resources.

The goal of this work is a creation of the noise model of the Sony IMX304LQR image sensor and development of a method for primary processing of multispectral images before their sending to the ground information complex.

The sensor’s noise measurement is fulfilled by recording the difference between the maximum and minimum shade values of the 40x30 sample taken from each of the spectral channels having the camera aimed at a sheet of paper with a whiteness percentage of at least 92% illuminated by the light of the color of a cloudless sky. The results of measuring the white balance showed that each of three channels has its own sensitivity level: 141, 236, 136 out of 255 shades. It is shown that a transition from analog to digital amplification is accompanied with a jump in the noise level of a panchromatic image. The relation of the number of noise shades of the Sony IMX304LQR sensor to the gain has been obtained. The yield made possible to assess the prospects for compressing a multispectral image by reducing the Shannon entropy. The noise level of the channels at the gain level of 0db is no more than 5 shades.

A method for processing the data obtained from the Bayer matrix has been developed. The method provides a progressive improvement in the image as its blocks are sent to Earth. The way of blocks organization makes it possible to obtain a usable image even in the event of an abrupt data receiving interruption.

Key words: multispectral images, noise model, YUV color space, satellite images.