The requirements of the earth data for remote sensing products are generally very demanding in terms of data quality and coverage/revisit time. In this research, we proposed the image classification method aimed at multispectral earth observation using Commercial off-The-Shelf (COTS) instruments onboard a 1U CubeSat for automatically selecting images for downlink on a 1U CubeSat. The hardware systems will be designed and developed by us without changing the severe limitations of size, power, volume, and mass imposed by 1U CubeSat. If implemented, our research will significantly contribute to image classification and more acceptable image-finding methods.
Machine learning has been widely used in multispectral remote sensing image analysis and classification for many years. The applications for multispectral or hyperspectral satellite image classification tasks often used random forests, support vector machines, or decision trees in earlier years. However, these approaches are not optimized for smaller hardware.
Through this research, we will try to design the mission payload of our second satellite which will be developed inside Bangladesh by the students. Here are the objectives of our Research
Demonstrate the multispectral camera feasibility on 1U CubeSat with On-Board Image Classification Using Deep Learning.
Space Radiation Effect on Micro-controller.
Using software-defined radio for monitoring Radio Frequency Interference.