基于机器视觉的果树树冠体积测量方法研究

针对人工测量精度低、费时费力,而基于三维激光扫描技术、超声波技术等自动测量方法成本高、操作复杂的不足,提出了基于机器视觉的果树树冠体积测量方法,搭建了可移植性果树树冠体积自动测量平台。基于机器视觉实现待测树冠图像获取,通过图像处理算法获得树冠图像面积特征,并采用最小二乘法和五点参数标定法获得普适性树冠面积与体积相关关系模型,从而得到树冠体积,通过对梨树以及桂花树样本的试验,可以发现预测树冠体积平均误差分别为13.73%和10.18%。对于不具备系列样本无法构建模型的树冠,采用单点测量法,根据树冠轮廓拟合椭球结构体,然后根据体积求算补偿公式,完成体积测量,测量误差在10%左右。表明树冠形态特征的图像提取算法可行有效,通过面积以及轮廓特征量均能很好地表达树冠体积特征。

Measurement Methods of Fruit Tree Canopy Volume Based on Machine Vision

There were some problems of artificial and sensor measurement for tree canopy volume, such as inefficiency, low precision, high cost, complex operation. In order to solve those problems, a new measurement method based on machine vision was proposed. The previous research indicated that there was significant correlation between tree canopy area and canopy. Based on this, the new method was proposed. Firstly, tree canopy image was obtained by machine vision according to the set standards. Secondly, tree canopy area was extracted by using a series of image processing operations. Meanwhile, the least square method and the 5-point calibration method were used to obtain the model of tree canopy volume. Finally, the corresponding volume was got. Experimental result showed that the average prediction error of the model of pear tree and Osmanthus fragrans were 13.73% and 10.18%, respectively. In view of the conditions of tree canopy, the structure estimation method was used to fit ellipsoid structure according to the contour of tree canopy that without a series of samples. Then, the volume of tree canopy was got by the compensation formula. Experimental result showed that the average prediction error of the model of peach trees and Osmanthus fragrans was about 10%. Consequently, characteristics extraction method of fruit tree canopy images was effective and feasible. The tree canopy volume characteristics can be perfectly expressed by tree canopy area and contour.