An Application-Oriented Virtual Microsensor Integration Platform

The availabilities of low-cost microsensors and wireless communication networks have extended the applications of sensor network into a broad spectrum. Currently available microsensor nodes such as Mote from Crossbow and sGate from Sensoria integrate many 1-D sensing devices and processing on one platform. However, 1-D signal in most time can not provide sufficient information for monitoring or detection purpose. Visual sensor network (VSN) that is a network including visual sensors provide an approach for 2-D signal demand. But the requirement of huge communication bandwidth, together with high power consumption, brings new challenges to real implementation. The ideal solution to these problems is system integration for on-site image processing. In this paper, we propose an image processing oriented virtual microsensor platform and target at system-on-chip (SoC) with embedded FPGA. The virtual microsensor platform does not provide a final microsensor but a design environment where a set of image processing intellectual property (IP) blocks can be selected with pre-defined metrics. The user-specified design is then integrated with pre-defined architectures, therefore meeting performance requirements of low power, high density, and sufficient processing capabilities at the same time. The IP library consisting of image pre-processing and processing blocks is the focal point of the virtual microsensor platform. Each IP block addresses earlier validation of various applications in both regular and faulty conditions. This paper presents the IP designs of polynomial approximation based geometric correction and independent component analysis (ICA) as pre-processing and processing examples. Several testbeds are also compared for prototyping and implementing the designs generated from the virtual microsensor platform