基于图像处理技术的棚室番茄果实识别  

Tomato Fruit Recognition in Greenhouse Based on Image Processing Technology

作　　者：伍蓥芮 张志勇[1] WU Yingrui;ZHANG Zhiyong(College of Agricultural Engineering,Shanxi Agricultural University,Taigu 030801,China)

机构地区：[1]山西农业大学农业工程学院,山西太谷030801

出　　处：《山西农业科学》2021年第7期898-902,共5页Journal of Shanxi Agricultural Sciences

基　　金：山西省重点研发计划项目(201803D221027-4)。

摘　　要：番茄采摘机器人进行自动采摘前需要准确识别和分割果实目标,为准确识别棚室番茄果实,研究采集了棚室环境内拍摄的60幅番茄果实图像,分别采用色差法、K-means算法和DBSCAN算法这3种图像分割算法对采集的番茄图像进行分割识别,分析比较了3种算法对棚室环境下拍摄的番茄果实图像分割结果。结果表明,色差法、K-means算法和DBSCAN算法的平均耗时分别为503、591、292 ms;DBSCAN算法耗时最短,且其平均信噪比最高,为95.2%,优于色差法(88.4%)和K-means算法(73.8%),对噪声抑制能力最强,可以初步满足棚室环境下番茄自动采摘机器人对果实目标的识别需求。Tomato picking robot needs to accurately identify and segment the fruit before automatic picking.In this paper,60 tomato fruit images were collected in the greenhouse environment.Three image segmentation algorithms,namely color difference method,K-means algorithm and DBSCAN algorithm,were used to segment and recognize the collected tomato images,and the segmentation results of the three algorithms were analyzed and compared.The results showed that the average time-consuming of color difference method,K-means algorithm and DBSCAN algorithm were 503,591,292 ms,respectively.DBSCAN algorithm had the shortest time-consuming,and had the highest average signal-to-noise ratio,which was 95.2%,better than 88.4%of color difference method and 73.8%of K-means algorithm.It had the strongest noise suppression ability,and could initially meet the requirements of tomato automatic picking robot for fruit recognition in greenhouse environment.

关 键 词：番茄 图像分割 色差法 K-MEANS算法 DBSCAN算法 图像识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]