Discriminating invasive Solanum mauritianum using image texture and sparse PLS discriminant analysis

Solanum mauritianum causes severe ecological and economic damage to commercial forest plantations, which is due to its destructive and resourceful nature. Therefore, monitoring its spatial distribution using remote sensing is essential for effective management of these plantations. As a result, this study used co-occurrence image texture parameters derived from WorldView-2 (WV-2) imagery integrated with sparse partial least squares discriminant analysis (SPLS-DA) to map the spatial distribution of invasive S. mauritianum in a South African commercial forest estate. The findings validated the performance of SPLS-DA in conjunction with co-occurrence image texture parameters to effectively discriminate S. mauritianum from neighboring commercial forest species with a 77% overall classification accuracy. In addition, correlation was the most significant co-occurrence image texture parameter selected by SPLS-DA for model development, followed by homogeneity then second moment. Finally, the 5x5 moving window was the most selected window size used for model development over the 3x3 and 7x7 moving windows. In essence, this study confirmed the ability of co-occurrence image texture parameters and SPLS-DA in effectively discriminating S. mauritianum from neighboring commercial forest species.