基于FPGA和分水岭算法的甜瓜图像分割

Hardware Accelerated Watershed Based Musk Melon Image Segmentation

准确从背景中分割出瓜体以及从瓜体表面分割出缺陷是甜瓜外观品质分级计算机视觉系统的重要任务。采用分水岭算法可以一次完成这2项工作。针对软件实现的分水岭算法在图像分割过程中实时性较差的缺点,采用了并行分水岭算法,并在FPGA平台实现该算法。试验表明,基于硬件的算法实现具有较好的实时性,提高了甜瓜图像的分割速度。

Accurately segmenting melon from the background and defects from the melon surface are the important tasks of computer-based melon quality grading system.The watershed algorithm could carry out these both tasks simultaneously.To overcome software-implemented watershed＇s disadvantages of poor real-time,the parallel watershed algorithm was performed on the FPGA platform.Experimental results showed that this hardware-accelerated implementation had a good effect of real-time,and could improve the speed of melon image segmentation.