基于矩阵广义逆和奇异值分解的运动水果模糊图像恢复

Blurred Image Resuming of Moving Fruit Based on GMI and SVD

针对利用机器视觉对进行水果分级时,由于水果运动所造成的模糊问题,提出了基于矩阵广义逆和奇异值分解的恢复方法,实验表明恢复的图像比较清晰,并且在保证实时的条件下可将水果大小检测的相对误差从4．17%减小为0．671%,相比于传统的恢复方法而言,提高了速度,消除了误差积累,为后续的边缘检测、形状分析、缺陷分类等打下了基础．

In order to resume the blurred image of moving fruit in the process of fruit quality inspecting and sorting using machine vision, an algorithm which based on General Matrix Inverse and Singular Value Decomposition is put forwarded. The experiment results indicated that, compared to the traditional algorithm, this algorithm can eliminate error accumulation, the resume the blurred image well, and improve the accuracy of the fruit size detection, the relative error of which is decreased from 4.173% to 0.671%. This lay a solid foundation for elevate the speed and accuracy of on-line edge detection, shape analyses, defect classification of fruit.