自然光照下基于粒子群算法的农业机械导航路径识别

发布时间：2018-03-16 18:37

本文选题：农业机械　切入点：机器视觉　出处：《农业机械学报》2016年06期 　论文类型：期刊论文

【摘要】：针对农业机械视觉导航线提取易受光照变化影响及常规导航线识别算法实时性低、抗干扰能力差等问题,对自然光照条件下基于机器视觉的农业机械导航路径识别技术进行了研究。首先,在YCr Cb颜色模型的基础上构建与光照无关的Cg分量,选择2Cg-Cr-Cb特征因子对图像进行灰度化处理,以降低光照变化对图像分割的影响;然后,采用改进K-means聚类方法进行图像分割,将绿色作物信息从土壤背景中分离出来,并通过形态学滤波方法滤除二值图像中存在的杂草干扰信息;最后,根据图像中作物行的特点建立作物行直线方程约束模型,利用粒子群算法对作物行直线进行寻优求解,进而得到导航线。实验结果表明,不同光照条件下对2Cg-Cr-Cb灰度图像进行图像分割,可以清晰完整地将作物从土壤背景中分离出来,分割图像受光照变化影响较小并且不会引入背景噪声;基于粒子群算法的导航线检测方法可以快速准确地提取出导航路径,对于不同农田作物和作物不同生长阶段具有较高的适应性,相比于常规导航线识别算法具有实时性高、鲁棒性好等优点。
[Abstract]:According to the agricultural machinery navigation line extraction is easily affected by illumination changes and conventional navigation line recognition algorithm of low real-time problems, poor anti-interference ability, the natural light condition of machine vision recognition technology based on agricultural machinery navigation path is studied. Firstly, the construction and light independent Cg component based on YCr Cb the color model, 2Cg-Cr-Cb feature selection factor on gray image processing, in order to reduce the illumination effect on image segmentation; then, using the improved K-means clustering method for image segmentation, green crop information from the soil background, and through filtering the morphological filtering method of two value of weed interference information in an image finally, the establishment of crop rows; linear equation constraint model according to the characteristics of image in crop rows, optimization of crop line using particle swarm algorithm, and obtain the Route. The experimental results show that under different illumination conditions on 2Cg-Cr-Cb image segmentation, can be clearly separated from the crop soil background, image segmentation illumination changes is small and does not introduce background noise; navigation line detection method based on particle swarm algorithm can extract the navigation path quickly and accurately. Has high adaptability for different crops and different crop growth stages, compared to the conventional navigation line recognition algorithm has high real-time performance and good robustness.

【作者单位】：天津职业技术师范大学自动化与电气工程学院;天津市信息传感与智能控制重点实验室;中国农业大学现代精细农业系统集成研究教育部重点实验室;
【基金】：国家自然科学基金项目(31571570) 国家国际科技合作专项(2015DFG12280)
【分类号】：TP391.41;TP18

【相似文献】