A full-hardware edge detection system based on a highly robust dual random number generator

Conventional edge computing hardware at autonomous driving terminals is limited by chip area and power efficiency for complex image data. Recently, researchers have used memristors to construct dedicated hardware systems, achieving advancements in multiple image processing fields such as pattern recognition and edge detection. However, a full-hardware image processing system with anti-interference, low power, and high real-time performance remains challenging. Here, we fabricated a volatile Ag/AlO x /n-Si memristor with exceptional stability and low power. High-resolution electron microscopy visualized Ag conductive filament formation. Integrating two memristors with a logic circuit, we designed a low-power edge detection unit using dual random number hardware computation. Its energy consumption is 92.3% lower than traditional logic circuits, with strong anti-interference. To validate practical use, we constructed a 32-channel parallel full-hardware system with 64 devices, achieving 31.25 frames per second and excellent edge detection. The system exhibits excellent scalability, stability, and anti-interference, offering great potential for edge vision hardware in autonomous driving, safety monitoring, and surgical assistance.