The next generation of iSIM binocular cameras that the SATLANTIS company develops and markets for installation in artificial microsatellites will have multispectral sensors up to 5 bands (GRB + NIR + Panchromatic) with high image resolution. This technology will dramatically expand the applicability of iSIM cameras in the field of satellite imagery terrestrial observation. The high spatial resolution offered by iSIM cameras, combined with their compact design and low weight, represent a technological achievement that significantly reduces the costs of missions for Earth observation. However, it is precisely these characteristics that make it more necessary to maximize and optimize image processing in the camera itself (onboard processing). On the one hand, the time and bandwidth limitations of communications between satellites and ground stations make it advisable to transmit only the relevant information obtained by their sensors. On the other hand, the power consumption in microsatellites is a determining factor for the operability of its instruments and the success of its missions. In multispectral imaging applications in particular, these requirements become more important due to the vast amount of data that is generated during observation. Thus, instead of storing and transmitting raw images (multispectral cubes), it is necessary to process the information generated to discriminate the relevant information before being stored and transmitted. In other words, the camera must incorporate a certain “intelligence” to “understand” what it is seeing and discriminate the information, thus relieving the satellite computer from carrying out these tasks and providing great added value to SATLANTIS products.

The purpose of the project is twofold. First, we want to develop a functional real-time cloud detection system adapted to the specific characteristics of SATLANTIS iSIM cameras and their embedded computer, SmartCIA. SmartCIAs feature two Zynq UltraScale ZU3EG MPSoCs and 2GB of external DDR. The starting hypothesis is that a detection system of this type can be designed so that it can be executed on this processor with a latency low enough to process the images acquired by the sensor in real time. To achieve this, both the design of the detection algorithm to be used (simplicity vs. accuracy) and the design of the processor architecture necessary to speed up the execution of the algorithm to the maximum (heterogeneous processor with accelerator cores) must be investigated. The final technical solution will depend on the conclusions drawn from each phase of the investigation, the final objective being to maximize the precision and robustness of the detector by squeezing the available computational power in the SmartCIA. In any case, the design of the processing system will be carried out following a hardware/software co-design methodology supported by design, simulation, co-simulation and verification techniques and tools. The second purpose of the project is to establish a consistent methodology that allows, in the future, to expand the capabilities of this detection system to enable the automatic detection of other types of objects in the images acquired by iSIM cameras, such as ships in the ocean, lines of coast etc.