Object-oriented greenhouse cantaloupe identification by remote sensing technology

Abstract

Cantaloupe (Cucumis melo var. saccharinus) is one of the favorite fruits for consumers and widely cultivated in China. Rapidly and accurately obtaining the distribution information of cantaloupe greenhouses is very helpful to local government. The traditional field survey method by workers is laborious and time-consuming. Remote sensing technology has distinct advantages to obtain distribution information of cantaloupe than the traditional method. The plastic greenhouses were usually adopted in order to improve the yield and quality of cantaloupe. The effect of plastic film on cantaloupe spectral reflectance leads to difficulties estimating the growth information of cantaloupe using remote sensing technology. Therefore, the purpose of this study is to provide a remote sensing method to accurately identify cantaloupe in greenhouses. In this study, with GF-2 image as the data source, we proposed an object-oriented remote sensing monitoring method for cantaloupe growth information based on random forest algorithm by using the specific spatial distribution details (texture features and shape features) and spectral features of plastic greenhouse. In the meantime, the method only using spectral features was also conducted as a contrast. In addition, because of the difference of the spectral features between different periods of cantaloupes, in order to improve the identify accuracy, the cantaloupes at early period and middle-late period were separately monitored in this study. The results show that the method with the combination of spectral features, texture features and shape features can effectively improve the classification accuracy than the method with only spectral features. The overall classification accuracy and the Kappa coefficient of the proposed method were 92.36% and 90.17%, respectively. This study successfully implemented the greenhouse cantaloupe identification meeting the requirements of actual production accuracy, and also is a preliminary study for monitoring the growth period of cantaloupe in greenhouses and can provide a basis for monitoring the growth period and predicting the maturity period of cantaloupe in the future.