Abstract
Abstract. Decision tree (DT) induction has been widely used in different pattern classification. However, most traditional DTs have the disadvantage that they consider only non-spatial attributes (ie, spectral information) as a result of classifying pixels, which can result in objects being misclassified. Therefore, some researchers have proposed a co-location decision tree (Cl-DT) method, which combines co-location and decision tree to solve the above the above-mentioned traditional decision tree problems. Cl-DT overcomes the shortcomings of the existing DT algorithms, which create a node for each value of a given attribute, which has a higher accuracy than the existing decision tree approach. However, for non-linearly distributed data instances, the euclidean distance between instances does not reflect the true positional relationship between them. In order to overcome these shortcomings, this paper proposes an isometric mapping method based on Cl-DT (called, (Isomap-based Cl-DT), which is a method that combines heterogeneous and Cl-DT together. Because isometric mapping methods use geodetic distances instead of Euclidean distances between non-linearly distributed instances, the true distance between instances can be reflected. The experimental results and several comparative analyzes show that: (1) The extraction method of exposed carbonate rocks is of high accuracy. (2) The proposed method has many advantages, because the total number of nodes, the number of leaf nodes and the number of nodes are greatly reduced compared to Cl-DT. Therefore, the Isomap -based Cl-DT algorithm can construct a more accurate and faster decision tree.
Highlights
In the current researches on remote sensing image classification and information extraction, it is a key issue to meet certain classification accuracy based on multiple categories
The same remote sensing image is classified by using the Isometric Mapping (Isomap)-based Cl-DT algorithm and the Cl-DT algorithm proposed in this paper. (a) is the Cl-DT algorithm,(b) is the Isomap-based Cl-DT algorithm
The algorithm, which overcomes the deficiency of the traditional ClDT method that Euclidean distances of instances that are nonlinear distributions in higher space cannot accurately represent the real distances of instances through merging the maximum variance unfolding algorithm with the co-location decision tree
Summary
In the current researches on remote sensing image classification and information extraction, it is a key issue to meet certain classification accuracy based on multiple categories. Decision tree is a basic classification and regression method It deduces the decision rule set from the disorganized training sample set, and the decision rule set continues to classify the new data. His advantage is the high accuracy of processing high-dimensional data without the need for knowledge or setting of parameters in other areas. There are some weaknesses in itself: sometimes intelligence does not satisfy people's expectations and there are more tree nodes and hierarchies Another defect is the problem to be solved in this paper. The traditional decision tree does not consider the spatial relationship between attributes.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
