Smart parallel spectral imager based on heterogeneous FPGA system on chip
In the last years, industrial image processing has been shifting to areas and tasks that are increasing in complexity. This results in new challenges in order to contrast features to be detected or evaluated. Systems for the acquisition and interpretation of multispectral images are thus becoming more and more interesting. A major issue is, depending on the sensor principle, the time to acquire this spectral data. FPGA (Field Programmable Gate Array) and in particular heterogeneous FPGA SoC (System-on-Chip) offer the possibility to accelerate these acquisition methods decisively. In addition to the image acquisition, it is also possible to calculate decisive preprocessing steps in the hardware. A frequently used algorithm for analyzing but also compressing hyperspectral data is the PCA (principal component analysis). This paper presents a research setup that combines a heterogeneous FPGA SoC with a 12-channel filter wheel camera. With the help of the device a parallel working PCA is to be integrated, which works distributed in hardware and software. The paper presents the concept for this implementation and the current state of development in the project. In addition, restrictions on the development of algorithms with hardware systems and the current distribution in hardware and software are discussed.