鸡蛋果大小测量的多种图像边缘检测算子对比  

Comparative Study of Multiple Image Edge Detection Operators Applied to Size Measurement of Passiflora edulia Sims

作　　者：刘现[1] 林营志[1] LIU Xian;LIN Yingzhi(Digital Agriculture Research Institute,Fujian Academy of Agricultural Sciences,Fuzhou,Fujian 350003,China)

机构地区：[1]福建省农业科学院数字农业研究所,福建福州350003

出　　处：《热带作物学报》2022年第12期2554-2563,共10页Chinese Journal of Tropical Crops

基　　金：福建省自然科学基金项目(No.2020J011377);智慧农林福建省高校重点实验室开放基金项目(No.2019LSAF01);福建省智慧农业科技创新团队项目(No.CXTD2021013-1)。

摘　　要：大小测量是鸡蛋果采后分选阶段一个十分重要且不可或缺的环节。阐述多种边缘检测算子以及基于边缘检测算法的图像识别步骤,模拟鸡蛋果采后大小测量分选现场,构建一个水果图像获取平台。为增加难度,选取40个表面具有一定皱皮现象的鸡蛋果,使用水果图像获取平台获取从顶端拍摄的40张鸡蛋果图像作为试验材料。基于Visual Studio2017集成开发环境,调用OpenCV3.2.0机器视觉库,采用C++编程语言,针对40张鸡蛋果图像,使用4种算子Canny、Sobel、Laplacian、Scharr进行边缘检测;使用Photoshop7.0.1软件基于图像测算鸡蛋果大小作为真实果实大小数据的近似值,参照该数值衡量算法检测的准确度;对4种算子的检测参数进行自动调优,依次使用SPSS软件对每一种算子的检测参数、4种算子的检测准确度及单果检测时间进行统计分析。试验结果表明:使用Canny算子在(70,105, 3)参数下进行鸡蛋果大小测量时,与其他几种算子比较,测量准确度最高,单果测量平均用时为1.98 ms,也是4种算子中最快的,可满足于鸡蛋果采后大小智能分选的实际需求。Size measurement was a very important and indispensable link in the postpartum sorting stage of passion fruit.This paper described a variety of edge detection operators and image recognition steps based on edge detection algorithm. A fruit image acquisition platform was constructed by simulating fruit sorting site. In order to increase the difficulty, forty passion fruits with certain wrinkle on the surface were obtained by using fruit image acquisition platform and forty images shoot from the top were selected as the test material. Based on the Visual Studio 2017 integrated development environment, called the OpenCV 3.2.0 machine vision library, C++ programming language was used to detect the edge of forty passion fruit images by Canny, Sobel, Laplacian and Scharr operators. The size of passion fruit was measured by photoshop 7.0.1 software based on images as the approximation of real fruit size and according to the value to measure the detection accuracy of algorithm. The detection parameters of the four operators were automatically optimized, and the detection parameters of each operator, the detection accuracy of the four operators and the single fruit detection time were statistically analyzed by SPSS software. The test results showed that the measurement accuracy of passion fruit size using Canny operator under(70, 105, 3) parameters was the highest, the average time consumed of the single fruit measurement was 1.98 ms, it was the fastest of four operators, which can satisfy the actual demand for the intelligent sorting of passion fruit.

关 键 词：鸡蛋果 边缘检测 算子 

分 类 号：S667.9[农业科学—果树学] TP391.4[农业科学—园艺学]