[Multi-DSP parallel processing technique of hyperspectral RX anomaly detection]

Area of Science:

• Computer Engineering
• Signal Processing
• Remote Sensing

Context:

• Hyperspectral imaging generates massive datasets requiring high-speed processing.
• Real-time RX anomaly detection is crucial for applications like target recognition and environmental monitoring.
• Existing systems face limitations due to internal storage capacity and processing speed.

Purpose:

• To propose a multi-Digital Signal Processor (DSP) parallel processing system for hyperspectral image RX anomaly detection.
• To design a hardware architecture using CPCI Express bus and tightly coupled DSPs.
• To develop a 4-DSP parallel processing technique for efficient computation of mean and covariance matrices.

Summary:

• A multi-DSP parallel processing system based on CPCI Express bus architecture is presented.
• The system utilizes four tightly coupled DSPs sharing data bus and memory, interconnected via Link ports.
• A novel 4-DSP parallel processing technique spatially partitions the hyperspectral image for efficient matrix computation.

Impact:

• Achieves 4x higher time efficiency compared to single DSP processing for RX anomaly detection.
• Overcomes the internal storage capacity constraints of DSPs for processing large hyperspectral images.
• Meets the real-time processing demands for hyperspectral spectral data analysis.