光学 精密工程, 2016, 24 (3): 651, 网络出版: 2016-04-13
基于形态学图像检测的机械手移栽穴苗识别技术
Recognition of seedlings in mechanical transplanting processing by Morphological image detection
摘要
针对机械移栽穴苗过程中如何有效识别出根系受损的幼苗, 进而及时进行分类、补救这一实际问题, 提出一种基于机器视觉的移栽穴苗识别方法。该方法根据机械手移栽穴苗特点, 对比检测图像土壤基质面积与基准面积; 进而通过测定土壤基质完整率, 判断样本合格与否。文中从理论和实际的角度说明了形态学图像处理方法处理穴苗移栽图像特点, 设计了机械手移栽穴苗识别在线测试系统。最后, 对基于形态学图像检测方法的机械手移栽穴苗识别技术和普通图像检测方法进行对比实验。结果表明, 本文提出的形态学检测方法对一个72孔穴盘合格检出率提高了6.6%, 不合格检出率提高了54.5%。同时, 一个穴苗的平均处理时间约为1.82 s。结果表明提出的方法可靠, 耗时较短, 能够匹配机械手移栽流水线作业时间上的要求, 并满足实时处理要求。
Abstract
How to discover and remedy the damaged seedlings quickly in batch processing seedlings by automatic transplanting technology was researched. A recognition method of seedlings in mechanical transplanting processing based on machine vision was proposed. Based on the characteristics of mechanical transplanting seedlings, the method was used to measure soil matrix intact rate of seedlings by comparison of the image areas between soil matrix and benchmark. By which the samples were qualified or not could be judged. Then, the image characteristics processed by morphological image processing method in mechanical transplanting seedlings were described from the theory and practice and an online testing system for recognition of mechanical transplanting seedlings was designed. Finally, the manipulator transplanting seedling method based on morphological image detection and the common image detection method were compared. Experimental results indicate that the morphological detection method proposed in this paper for a seedling tray containing 72 holes increases the detection rate by 6.6%, and unqualified detection rate by 54.5%. At the same time, the average time that processes one seeding is about 1.82 s. It is shown that the proposed method is reliable, short time-consuming, satisfied with the requirement of real-time processing.
王跃勇, 于海业, 刘媛媛. 基于形态学图像检测的机械手移栽穴苗识别技术[J]. 光学 精密工程, 2016, 24(3): 651. WANG Yue-yong, YU Hai-ye, LIU Yuan-yuan. Recognition of seedlings in mechanical transplanting processing by Morphological image detection[J]. Optics and Precision Engineering, 2016, 24(3): 651.