CAJ | 학술논문

在玉米育种和品质研究中，经常需要对玉米的果穗长度、果穗宽度、穗行数、穗粒数等参数进行测量。该研究提出了一种基于机器视觉的玉米果穗参数图像测量方法。使用PC摄像头连续采集旋转台上的玉米果穗图像，经过图像处理，获得玉米穗的图像区域，进而得到玉米果穗的穗长和穗宽参数；通过对玉米果穗局部区域的x方向和y方向累计像素值曲线进行分析，提取出玉米穗行，获得每一穗行的穗粒数和穗行宽度；通过图像匹配，获得玉米果穗的穗行数。试验表明，使用该研究方法对玉米果穗的长度、宽度和穗行数的参数测量准确率可达98%以上，对穗行宽及总穗粒数测量准确率达95%以上，整穗的平均检测时间约102 s/穗。该研究实现了玉米果穗参数快速有效的自动检测，相对于目前采用的人工检测，大大提供检测效率，降低劳动强度，可应用于玉米千粒质量检测、产量预测、育种和品质分析等场合。
재옥미육충화품질연구중，경상수요대옥미적과수장도、과수관도、수행수、수립수등삼수진행측량。해연구제출료일충기우궤기시각적옥미과수삼수도상측량방법。사용PC섭상두련속채집선전태상적옥미과수도상，경과도상처리，획득옥미수적도상구역，진이득도옥미과수적수장화수관삼수；통과대옥미과수국부구역적x방향화y방향루계상소치곡선진행분석，제취출옥미수행，획득매일수행적수립수화수행관도；통과도상필배，획득옥미과수적수행수。시험표명，사용해연구방법대옥미과수적장도、관도화수행수적삼수측량준학솔가체98%이상，대수행관급총수립수측량준학솔체95%이상，정수적평균검측시간약102 s/수。해연구실현료옥미과수삼수쾌속유효적자동검측，상대우목전채용적인공검측，대대제공검측효솔，강저노동강도，가응용우옥미천립질량검측、산량예측、육충화품질분석등장합。
The parameters such as the length, the number of ear rows, and the quantity of kernels in an ear of corn were measured during corn breeding and quality studies. It is usually done mainly manually. This research proposes an efficient image processing algorithm to detect the parameters of an ear of corn based on a machine vision. An experimental device was designed to detect the parameters. It mainly included a computer, a module of data acquisition and control, a stepper motor, a stepper motor driver, a PC camera, and other mechanical components. The computer was used to control the stepper motor to rotate the ear of corn and trigger the PC camera to capture images. The image was segmented after the ear of corn was captured. Its contour was traced. The length and the width of it were obtained by measuring the contour. The horizontal and vertical accumulated pixel values histograms were used in this research. One point in the upper edge and one point in the lower edge of the central ear row were found by first searching for the concaves of the horizontal accumulated pixel values histogram in a specified region. All the points in the upper and the lower edges of the central row were obtained by searching for the concaves of the horizontal accumulated pixel values histograms in a specified moving region which moved following the edge of the central ear row direction. So the image of this central ear row was determined. Each gap between the adjacent kernels could be distinguished by searching for the concaves of the vertical accumulated pixel values histogram in the image area of the central ear row. Then the width of the central ear row and the quantity of kernels in this ear row were recorded. The image of the next adjacent ear row was taken while this ear row was rotated to the location in which the former ear row was imaged. The condition of stopping detection was judged by matching the image of the current ear row with the first. So the number of the ear rows was determined. The quantity of the kernels in this ear of corn could be obtained by accumulating the kernels of all ear rows. In this research, an experimental device was designed to detect the parameters of an ear of corn. And an algorithm was supplied on the base of a machine vision for the same purpose. The image of each ear row in the ear of corn was effectively taken with no repeat. The parameters were detected such as the length and the width of the ear of corn, the width of one ear row, the number of ear rows, and the quantity of kernels in the ear of corn. Experiments showed that the measurement accuracy of the length, the width, and the number of the ear rows of the ear of corn was up to 98%. The measurement accuracy of the width of each ear row and the quantity of kernels was up to 95%. The detection speed was about 102 seconds per ear of corn.