基于边缘链码信息的番茄苗重叠叶面分割算法

Segmentation algorithm of overlapping tomato seedling leaves based on edge chaincode information

针对番茄穴盘苗自动移栽机在检测幼苗特征参数时,由于叶面重叠导致提取特征参数不准确问题,提出一种基于重叠叶面边缘链码信息逐层分割算法。首先将穴盘苗原图像预处理得到的边缘二值图进行重叠叶面边缘拐点检测和链码统计,然后通过拐点配对进行叶面逐层分割,对真实分割拐点对进行线性插值,最终实现重叠叶面分割。试验结果表明,采用逐层重叠叶面分割算法具有旋转不变性,针对72孔和128孔2种穴盘规格,重叠叶面分割成功率分别为100%和96%,分割平均耗时分别为0.835和0.99s。此分割算法分割速度快、精确度高,可满足实际工厂化全自动移栽作业要求。

According to overlapping leaves while robotic transplanters extracting seedling leaves＇ feature parameters with machine vision,a layer-by-layer segmentation algorithm was proposed based on overlapping leaves chaincode information. Firstly the chaincode was counted and corners and edges of overlapping leaves were detected based on edge binary image from original image preprocessing. Then the overlapping leaves were separated layer-by-layer through authentic corners and edges pairing,and the overlapping leaves segmentation was achieved by linear interpolation. The result of experiments showed that the segmented leaves images would not be changed by rotation,and the success rate of the algorithm was 100% and 96% respectively and the separating time was 0.835 and 0.99 s respectively for two types of plug size （72 cells and 128 cells）. This segmentation algorithm has the advantage of high speed and high accuracy with rotation invariance,and can meet the requirements of automatic transplanting.