基于机器视觉的番茄实时分级系统设计

Tomato Real-time Classification Based on Machine Vision Detection System Design

针对人工检测番茄品质效率低、主观性强等问题,设计了基于机器视觉技术的适应于番茄外部品质检测的多方法融合的图像采集及图像处理系统.首先,采用颜色直方图获取番茄的颜色特征;其次,通过边界跟踪算法得到番茄的半径序列点,经过傅里叶变换与反变换处理,计算出番茄的不规则度,从而对其形状特征进行描述,同时,将格林公式变形求得番茄形心坐标,并利用圆形度计算其最大横径,以获取番茄的大小特征;最后,用基于线性判别函数和决策树的模式分类器对番茄的大小、颜色和形状特征综合进行分级处理.实验结果表明,系统分级结果基本稳定,分级精度达到92%.

A multiple method fusion image acquisition and processing system based on machine vision technology which applies to detecting tomatoes external quality is designed to solve the problem of low efficiency and strong subjectivity of artificial detection. Firstly, the color histogram is used to obtain the color features of tomatoes. Then, the characteristics of tomatoes＇ shape is described after calculating the irregular degree by using the boundary tracking algorithm which could get the radius of tomato sequence points and the Fourier transform and inverse transform. In the meantime, a deformation of Green＇s formula is introduced into obtaining tomato centroid coordinates and the maximum transverse diameter to identify the size. Finally, according to previous image information such as size, color and shape feature, tomatoes are graded effectively by using pattern classifier based on linear discriminant function and decision tree. The classification accuracy is improved to 92% and the stability of classification results of the system are proved by experiments.