Comparison of different design methodologies of hardware-based image processing for automation in microrobotics

Object-detection and classification is a key task in micro- and nanohandling. The microscopy image is often the only available sensor to detect information about the positions and orientations of objects. FPGA-based image processing is superior to state of the art PC-based image processing in terms of achievable update rate, latency and jitter. Development of specialized FPGA solutions for object detection and tracking using a hardware-description language is time consuming and requires deep knowledge of the target system. Using SystemC, a C++ based class library, fast implementation with less knowledge of the system is possible. However, the gain in development speed is accompanied by higher resource usage of the FPGA and lower performance regarding computation speed. In this paper, several image processing algorithms are implemented in both manners and are compared regarding achievable update-rate, resource consumption and development time. The SystemC implementations could be implemented twice as fast and with less knowledge about the system. The increased resource usage of SystemC is negligible for small series and prototyping applications, while the decreased computation speed may be problematic for high-speed applications.