【作者】 孙仁山；

【导师】 李文彬；

【作者基本信息】 北京林业大学 ， 机械设计及理论， 2006， 博士

【摘要】 立木整枝是优质工业用材林定向抚育的重要环节。传统的整枝方法效率低、易引起工人的疲劳和事故、作业高度也受到限制,因此立木整枝机的自动化及智能化是实现高效、高质量和安全可靠整枝的有效途径。本文在研究传统遥控整枝机工作原理的基础上,综合运用数字图像处理、小波分析、模式识别、模板匹配及多元统计分析等技术,设计了一套用于立木自动整枝机全自动作业的机器视觉识别系统,从而来达到立木枝干自动识别的要求,为后续的自动识别研究提供了前期基础。本研究的主要内容和结论如下:1、确定了智能立木整枝机视觉系统的总体方案,并设计了一套行之有效的计算机视觉识别系统。2、通过对北京市西山林场的116组落叶松林的干径和枝径进行统计分析研究,通过拟合优度检验的方法,验证了落叶松林的干径和枝径符合正态分布;并运用两个总体的距离判别法,对判断枝丫是否为分岔枝进行了判别分析,建立了分类枝丫为分岔枝的判别函数,取得了良好的分类效果。3、通过对几种常见的图像压缩、图像增强和图像滤波技术进行比较分析,介绍了KL变换、DCT编码及小波变换等压缩编码,对灰度变换和直方图调整进行了阐述,并讨论了均值滤波和中值滤波技术。最后在分析和试验的基础上,选定运用二维离散小波变换对图像进行压缩,采用对比度增强的方法进行图像增强,最后采用中值滤波技术对图像进行消噪,通过以上技术对图像进行了有效的预处理,实际效果较好。4、首次将二次模板匹配的方法应用到立木图像标定尺检测中。通过自行设计的标定尺模板,利用二次模板匹配方法在图像在自动检测出了标定尺的位置,同时通过角点检测方式对标定尺线段长度像素数进行了测量,利用标定原理,得到了标定比例系数,为立木枝干直径测量做好了前期准备工作。5、通过数学形态学方法,综合运用图像分割、骨架化和卷积运算等方法对立木树枝生长点检测进行了研究。通过计算树木主干垂直直方图,提取到树木的主干信息并对主干图像进行骨架化运算,将得到的细化结构与设计的卷积模板做卷积运算,将卷积结果与给定的阈值相比较,得到立木枝干的交叉点处的具体空间位置。运用标定原理,通过对交叉点处树干和枝丫直径进行测量,得到树干和枝丫直径的实际生长尺更多还原

【Abstract】 Pruning is one of the most important and necessary step of cultivating plantation forest. The traditional pruning machine has low working efficiency, high casualty, limited prunning height and is difficult to meet the need of fostering the large area of commercial plantation in China. So the key method is to develop an automatic pruning technology, which should be safe, high quality and efficient.Based on studying of the theory of the traditional pruning machine, this paper synthetically applied digital image processing, wavelet, pattern recognization, template matching and multi-statistics technology, and designed a machine vision for automatically pruning to recognize the branch and trunk to be pruned. This paper applied the prophase base for the automatically recognization. The main contents and conclusions drawn in the dissertation:1. This thesis works out the mode of the machine vision system of automatic pruning machine, and designs an effective computer vision system.2. Through statistical analysis and research for the 116 group larch forests trunk and branch from Beijing Xishan forestry, and through superior fit for the test methods, we conclude the trunk and branch of larch forests obey the normal distribution; And apply the distance determine method of two person, a discrimination function of the larch fork branch and non-fork branch is established by statistics. Excellent discrimination results are obtained by the function and rule.3. Through comparing several common image compression, image enhancement and image filtering technologies, this paper introduced KL, DCT and DWT encoding, explained the gray and hist adjustment, and described and discussed in the mean filtering and medial filtering technology. In the final analysis, this paper selected for their use of更多还原