發(fā)布時間：2019-05-23 23:27

【摘要】：華南地區(qū)種植園地廣泛分布,類型混雜多樣,導(dǎo)致園地分布信息難以正確獲取,為農(nóng)業(yè)管理造成了較大困難。本研究基于Landsat8 OLI數(shù)據(jù),通過數(shù)據(jù)融合、特征優(yōu)化,應(yīng)用隨機森林算法構(gòu)建面向?qū)ο蟮姆N植園地分類規(guī)則集,對華南地區(qū)典型經(jīng)濟作物香蕉、柑橘、葡萄、蒲葵、海棗、番木瓜和火龍果等進行類別識別,同時對比貝葉斯分類法、K最鄰近分類法、支持向量機法、決策樹分類法的分類效果。結(jié)果表明:數(shù)據(jù)融合會在一定程度上影響分類結(jié)果精度;植株形態(tài)、光譜特征接近,種植期交錯是影響華南地區(qū)典型園地分類精度的重要原因;以中分辨率影像為數(shù)據(jù)源,面向?qū)ο蟮碾S機森林算法應(yīng)用于種植園地分類研究總體精度可達88.05%,Kappa系數(shù)0.87,可以有效區(qū)分華南地區(qū)典型種植園地類別;相比于其他算法,隨機森林算法在分類精度、可靠性和穩(wěn)定性上具有一定優(yōu)勢,可為園地作物生長監(jiān)測和種植管理提供科學(xué)依據(jù)。
[Abstract]:The planting gardens in South China are widely distributed and the types are mixed and diverse, which makes it difficult to obtain the information of garden distribution correctly, which makes it difficult for agricultural management. Based on Landsat8 OLI data, through data fusion and feature optimization, an object-oriented classification rule set of planting gardens was constructed by using random forest algorithm. Banana, citrus, grape, sunflower and sea jujube, a typical cash crop in South China, were constructed. Papaya and dragon fruit were identified by category recognition, and the classification effects of Bayesian classification, K nearest neighbor classification, support vector machine method and decision tree classification were compared. The results showed that data fusion would affect the accuracy of classification results to a certain extent, and the plant morphology and spectral characteristics were close, and the interlaced planting period was an important reason for the classification accuracy of typical gardens in South China. Taking the medium-resolution image as the data source, the object-oriented stochastic forest algorithm can be applied to the classification of planting gardens with the overall accuracy of 88.05%, and the Kappa coefficient is 0.87, which can effectively distinguish the typical planting garden types in South China. Compared with other algorithms, stochastic forest algorithm has some advantages in classification accuracy, reliability and stability, and can provide scientific basis for crop growth monitoring and planting management in garden land.
【作者單位】： 西南大學(xué)地理科學(xué)學(xué)院三峽庫區(qū)生態(tài)環(huán)境教育部重點實驗室;中國科學(xué)院深圳先進技術(shù)研究院;
【基金】：深圳市科技計劃項目(JCYJ20150831194835299) 國家重點研發(fā)計劃子課題(2016YFC0500201-07)
【分類號】：S127

【相似文獻】

相關(guān)期刊論文 前2條

1 張雷;王琳琳;張旭東;劉世榮;孫鵬森;王同立;;隨機森林算法基本思想及其在生態(tài)學(xué)中的應(yīng)用——以云南松分布模擬為例[J];生態(tài)學(xué)報;2014年03期

2 錢永蘭,楊邦杰,雷廷武;數(shù)據(jù)融合及其在農(nóng)情遙感監(jiān)測中的應(yīng)用與展望[J];農(nóng)業(yè)工程學(xué)報;2004年04期

，

本文編號：2484332