Intuition-1

Intuition-1 is a 6U hyperspectral satellite developed by KP Labs, launched in November 2023 as part of the SpaceX Transporter-9 mission. The mission aims to demonstrate AI-based on-board data processing to reduce the amount of raw data transmitted to Earth and enable environmental monitoring. By processing hyperspectral imagery directly in orbit, Intuition-1 enhances the efficiency of remote sensing applications, particularly in vegetation analysis, soil composition mapping, and environmental change detection. The mission also serves as a testbed for evaluating the applicability of nanosatellites for medium-resolution hyperspectral imaging with onboard deep-learning-based processing.

The satellite integrates two primary payloads developed by KP Labs:

• A 192-band hyperspectral sensor (HSI) that captures spectral data in the visible to near-infrared range (465–940 nm) with 3–6 nm spectral resolution and a ground sampling distance (GSD) of 25 m at a reference orbit of 600 km altitude. The instrument captures HSI in a push-broom manner using a CMOS image sensor with linear variable filters (LVFs), enabling high-speed imaging at 340 fps. A keypoint-based co-registration process is applied to raw frames to generate a sub-pixel accurate hyperspectral cube.
• The Leopard Data Processing Unit (DPU), an AI-powered computing platform capable of executing up to 3 trillion operations per second (TOPS) for complex image analysis and filtering directly on-board. It features two cold-redundant processing nodes (Zynq UltraScale+ ZU9EG, 4 GiB SLC NAND Flash, 16 GiB DDR4, 2 × 240 GiB pSLC SSD) and a Supervisor subsystem with radiation-hardened MCU, triple NOR Flash memories, and an FPGA device performing TMR voting.

On-Board Data ProcessingOne of the most critical aspects of Intuition-1 is its ability to analyze hyperspectral data in orbit rather than relying solely on ground-based processing. This significantly reduces the need for high-bandwidth data transmission, allowing only relevant and processed insights to be sent to Earth.The Leopard DPU serves as the satellite’s processing core, running AI and machine learning algorithms designed to optimize Earth observation tasks. These include:

• Cloud detection and filtering to identify cloud-covered regions, ensuring that only clear images are transmitted.
• Image segmentation to differentiate between land cover types, water bodies, and vegetation for more efficient data utilization.
• Hyperspectral data classification using AI-driven models to extract valuable environmental information from multi-band imagery.

By leveraging FPGA-accelerated deep learning inference, the system processes large hyperspectral datasets, reducing latency and transmission costs while increasing data accuracy and usability. The Leopard DPU also handles data preprocessing tasks, including raw image acquisition, sensor correction (bad pixels, dark frames, flat-field correction), and geometric distortions. Additionally, it performs data compression using CCSDS-123 algorithms to optimize storage and transmission efficiency. The unit manages S-band radio uplink (256 kbps) and X-band downlink (3–50 Mbps) for communication.Hyperspectral Imaging CapabilitiesThe hyperspectral sensor onboard captures 192 spectral bands, offering a significantly more detailed view of Earth’s surface than traditional RGB or multispectral sensors. This expanded spectral range enables:

• Vegetation and soil analysis, including plant species differentiation, crop health monitoring, and soil composition assessment (e.g., moisture and nutrient levels).
• Environmental monitoring by detecting pollution levels, analyzing water quality, and tracking deforestation patterns.
• Disaster response support, providing near-real-time insights into wildfire damage, flooding, and other natural disasters.

Since launch, Intuition-1 has captured hyperspectral images from multiple locations worldwide, including Chad, Turkey, Alaska, Saudi Arabia, Algeria, Poland, Brazil, and Spain. The data collected has been processed directly on-board, validating the Leopard DPU’s AI models for analysis.System Resilience & Performance in OrbitOperating in space presents numerous challenges, including exposure to radiation and extreme temperature fluctuations. The Leopard DPU has been designed with radiation-tolerant hardware and fault-tolerant computing to ensure mission continuity.Key reliability features include:

• Redundant processing nodes that maintain operation even if one processing core fails.
• Error-correction mechanisms to prevent single-event functional interrupts (SEFIs) caused by radiation exposure.
• A conduction-based thermal management system that dissipates heat into space, ensuring stable operation under high processing loads.

The DPU runs a customized Linux OS and utilizes the Xilinx Vitis AI framework to accelerate deep convolutional neural networks (CNNs) on FPGA hardware. This architecture enables energy-efficient inference (0.3 TOPS/W) and supports in-flight reconfigurable deep models, allowing the system to update processing applications dynamically during the mission.Over the past year, the DPU has demonstrated uninterrupted operation, successfully executing deep-learning models in orbit and validating AI-based decision-making in space environments.Key Achievements to Date

• AI-driven hyperspectral image processing successfully demonstrated in orbit.
• Cloud detection and intelligent image filtering implemented.
• Stable image segmentation and object detection at 3 TOPS.
• Terabytes of hyperspectral data processed from multiple global locations.
• Radiation-hardened AI computing validated for extended space operations.

Intuition-1 continues its technology demonstration phase, refining AI models and enhancing its on-board data processing capabilities. The mission serves as a critical testbed for next-generation hyperspectral imaging and autonomous satellite decision-making, setting a new standard for in-orbit AI applications. With its capability to receive in-orbit updates (uplinking new processing applications or pipelines), Intuition-1 is effectively a flying laboratory, ready to adapt to new scientific and commercial use cases as the mission progresses.