CAJ | 학술논문

大田成熟棉花的识别和定位是采棉机器人作业的基本前提. 为此,提出一种采棉机器人视觉系统解决方案. 首先根据棉纤维颜色特点,将彩色棉田图像分离为RGB 3 个通道,分别进行预处理和阈值分割,最后合并为彩色图像,继而检测大面积区域的轮廓并计算其质心点;在轮廓附近区域对预处理过的棉田图像进行 FAST 角点检测,使用SUFT特征描述并进行四层约束立体匹配,结合隐式定标参数可计算出特征点的空间坐标;最后,建立了一种考虑镜头缩放影响的轮廓面积/最小外接圆面积比值的成熟度估计模型. 实验结果表面:本方法可在准确识别和定位大田棉花的同时判断棉花的成熟度.
대전성숙면화적식별화정위시채면궤기인작업적기본전제. 위차,제출일충채면궤기인시각계통해결방안. 수선근거면섬유안색특점,장채색면전도상분리위RGB 3 개통도,분별진행예처리화역치분할,최후합병위채색도상,계이검측대면적구역적륜곽병계산기질심점;재륜곽부근구역대예처리과적면전도상진행 FAST 각점검측,사용SUFT특정묘술병진행사층약속입체필배,결합은식정표삼수가계산출특정점적공간좌표;최후,건립료일충고필경두축방영향적륜곽면적/최소외접원면적비치적성숙도고계모형. 실험결과표면:본방법가재준학식별화정위대전면화적동시판단면화적성숙도.
The mature cotton identification and positioning in the field are the basic premise for cotton -harvesting robot. A robot vision system solution was suggested by this manuscript.Firstly, cotton field image was separated into RGB chan-nels according to the characteristics of cotton fiber.The divided image was pre-processed and threshold-operated respec-tively and three channels were merged into a color image.The contours in the image were determined and the centroid was calculated.In the neighboring area of the contour, the feature points were detected using FAST corner detection algo-rithm and described with SURF feature description, then four constraints feature matching was adopted.Combining with the implicit calibration parameters, the 3 D coordinates value was calculated.A cotton maturity estimation model was es-tablished using the impact of the contour area /minimum area ratio of the circumscribed circle considering the influence of a zoom lens.Experimental results showed the proposed method may be identified and located the cotton in the field ac-curately and the maturity of cotton was confirmed at the same time.