Cassava brown steak diseases (CBSD) has caused serious reduction to the cassava productivity in Africa and is hard to detect it at the early stage of infection. Exist approaches requires large number of samples or complex experiment process, which cannot be implement in practice. This dissertation proposed a model that can detect infected plants at an early stage based on dual information fusion. Two main steps applied in this model is feature extraction and spatial optimization. For feature extraction, kernel principal component analysis (KPCA) is used to extract crucial information and transform original data to linear separable data, the data is then classified by support vector machines (SVM) to obtain a probability map. For spatial optimization, extended random walker is applied to generate another probability map. Then, the probability map created through feature extraction and spatial optimization are combined following a decision fusion algorithm. The model was applied on three groups of plants from three different trials, the result shows that this model has the potential to classify infected plants, but more trials are required to evaluate the reliability of this model in practice environment.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。