Label propagation based supervised locality projection analysis for plant leaf classification

• Published in: Pattern recognition Vol. 46; no. 7; pp. 1891 - 1897
• Main Authors: Zhang, Shanwen, Lei, Yingke, Dong, Tianbao, Zhang, Xiao-Ping
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.07.2013; Elsevier
• Subjects: Agronomy. Soil science and plant productions; Algorithms; Applied sciences; Artificial intelligence; Biological and medical sciences; Biometrics, statistics, experimental designs, modeling, agricultural computer applications; Classification; Computer science; control theory; systems; Data points; Data processing. List processing. Character string processing; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; Generalities. Biometrics, experimentation. Remote sensing; Information retrieval. Graph; Label propagation; Labels; Memory organisation. Data processing; Optimization; Pattern recognition; Pattern recognition. Digital image processing. Computational geometry; Plant leaf classification; Projection; Running; Software; Supervised locality projection analysis (SLPA); Theoretical computing; Warshall algorithm; Warshall algorithm; Label propagation; Supervised locality projection analysis (SLPA); Plant leaf classification; Transitive closure; Taxonomy; Labelling; Linear time; Plant leaf; Graph theory; Projection method; Dimension reduction; Experimental result; Multidimensional analysis; Supervised learning; Semi-supervised learning; Classification; Database
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2013.01.015

The label propagation has the benefits of nearly-linear running time and easy implementation. In this paper, we make use of the label propagation to propose a new weight measure, and present a supervised locality projection analysis (SLPA) method for plant leaf classification. Firstly, we apply Warshall algorithm to label propagation and get the label matrix, then incorporate it into the weight, which has a clear physical meaning. Secondly, multi-class data points in high-dimensional space are to be pulled or pushed by discriminant neighbors to form an optimum projecting to low dimensionality. Finally, the experimental results on two plant leaf databases show that the proposed method is quite effective and feasible. ► Applying Warshall algorithm to label propagation and obtain the label matrix. ► Incorporate the label matrix into the weight, which has clear physical meaning. ► We integrate the class information of all points before computing the weight. ► Multi-class data points are to be pulled or pushed by discriminant neighbors. ► We apply the method to leaf classification.