Research and Implement of Head Milled Rice Detection High-Speed Algorithm Based on FPGA

Head milled rice online detection implements dynamical image sampling and analysis, and then generates decision signals in the process of rice falling. Thereby, an efficient detection algorithm is very important for the fast falling rice selection. Since the traditional diameter classical search algorithm causes much loss of time and resources, we propose a center-edge rotation search algorithm whose complexity is only O(n). The main principle of the algorithm is to search the max diameter by traversing from the image center point to edge pixe points. The algorithm is simulated and tested on the ISE8.2i software platform and head milled rice testing machine. Finite-state machine is written with VHDL, and operations of multiplication and nuclear-rooting are realized by using IP core. The results show that processing an image consumes 2.7 ms, the error is below 10%. The algorithm improves efficiency, and also meets the demands in the accuracy of diameter results. There will be great prospects on the dynamical and real-time selection of head milled rice.