文献类型：期刊文献

中文题名：幼龄降香黄檀冠层叶片全磷含量的无损估计

英文题名：Nondestructive Estimation of Total Phosphorus Content in Canopy Leaves of Young Dalbergia odorifera

作者：陈珠琳[1] 王雪峰[1] 管青军[2]

第一作者：陈珠琳

通信作者：Wang, Xuefeng

机构：[1]中国林业科学研究院资源信息研究所,北京100091;[2]内蒙古莫尔道嘎林业局,额尔古纳022250

年份：2019

卷号：55

期号：11

起止页码：145-152

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;EI(收录号：20200808194901);Scopus;北大核心:【北大核心2017】；CSCD:【CSCD2019_2020】；

基金：国家自然科学基金项目（31670642）;林业科学技术推广项目（[2016]11号）

语种：中文

中文关键词：降香黄檀;图像理解;图像提取;混合效应模型;全磷含量估计

外文关键词：Dalbergia odorifera;image understanding;image extraction;mixed effects model;phosphorus content estimation

分类号：S758.6

摘要：【目的】以树木图像为数据源,采用图像理解方法估算幼龄降香黄檀单位质量叶片的全磷含量,为林农在施肥时间与数量方面经营该树种提供参考。【方法】首先提出从图像中分割降香黄檀树木冠层的算法;然后构建用于估计叶片全磷含量的统计模型形式和有效图像参数;最后采用混合模型方法,引入随机效应,建立以图像参数为自变量的植物叶片全磷含量预测模型,实现基于图像的叶片全磷含量预测。【结果】由于森林图像的自然属性,林木分类提取成为图像处理中的难点问题,利用林木前景与背景存在颜色差异这一特性,提出简洁的绿率树冠图像提取方法,通过大量图像测试,获知当绿率取0.35~0.42时,能够有效屏蔽背景保留树冠;进一步组合分析图像参数构建养分含量计算模型,确定将标准化灰度值作为指数并用暖距进行调整的二元叶片全磷含量预测模型,该模型能够实现对树冠叶片单位质量全磷含量的较高精度估算;同时,在模型参数估计时引入随机效应,对于各地区土壤条件等存在差异的降香黄檀全磷含量预测表现出较好适应性。【结论】对于与背景存在一定差异的林木图像,绿率是一种很好的树冠图像分割提取方法;在全磷含量预测模型中,双图像参数模型能够有效提高估计精度;对于各地区土壤或环境等存在差异的降香黄檀冠层叶片全磷含量预测,混合效应模型有效融合差异到一个模型中,表现出强大的适应性。
【Objective】 In this paper, the total phosphorus content of young Dalbergia odorifera leaves in per unit-mass was estimated by using the image understanding method with the tree image as the data source.【Method】 Firstly, the algorithm which was used to extract the canopy image of Dalbergia odorifera from the image is provided. Then the statistical model form and the effective image parameters used to estimate the total phosphorus content of leaves are constructed. Finally, the plant leaf total phosphorus content prediction model with image parameters as independent variables is established by using nonlinear mixed-effects model.【Result】 Based on the color difference between foreground and background, this paper proposes a simple method to extract canopy image with green rate. Through a large number of image tests, we know that it can effectively erase the background when the green rate is set between 0.35 and 0.42. Furthermore, we combined and analyzed the image parameters, and built the nutrient content estimation model. The leaf total phosphorus content prediction model is established, which takes the standardized gray value as indicators and adjusts them with the the warm data. The model can achieve a high precision estimation of the phosphorus content per unit weight of canopy leaves. At the same time, the random effects are introduced to the model parameter estimation, and the results show a good adaptability to the prediction of total phosphorus content of Dalbergia odorifera with different soil conditions in different regions.【Conclusion】 The result indicate that the green rate is a good method for tree crown image segmentation and extraction when there exist a certain difference between the background and foreground. The two-image parameter model could effectively improve the estimation accuracy of total phosphorus content prediction. For the prediction of total phosphorus content in canopy leaves of Dalbergia odorata with different soils or environments in different regions, the mixed effect model integrates these differences into one model, and shows a strong adaptability.