An adaptive plant leaf mobile informatics using RSSC

• Published in: Multimedia tools and applications Vol. 76; no. 20; pp. 21339 - 21363
• Main Authors: Prasad, Shitala, Peddoju, Sateesh Kumar, Ghosh, Debashis
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.10.2017; Springer Nature B.V
• Subjects: Adaptive systems; Biodiversity; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Datasets; Electronic devices; Entropy (Information theory); Informatics; Machine learning; Mathematical analysis; Matrix methods; Multimedia Information Systems; Special Purpose and Application-Based Systems; Human mobile interaction (HMI); NN; Mobile vision (MV); Shape descriptor; Relative sub-image sparse coefficients (RSSC); Leaf image informatics
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-016-4040-8

An automated plant biometric system is now an important step in preserving nature’s biodiversity. This paper presents a novel Relative Sub-image Sparse Coefficient (RSSC) algorithm for mobile devices (MDs) representing plant leaves into a mathematically compact vector for its classification. The RSSC feature vector includes local Statistical Entropy Texture (SET) information inter-related to all the sub-images within a leaf. RSSC space is merged with Gray Level Co-occurrence Matrix (GLCM) feature to refine the outputs using best -Nearest Neighbor ( best -NN), designed for MDs. The experiments were performed on three different types of leaf datasets: (i) Flavia, (ii) ICL and (iii) Diseased leaf datasets. The results proves our method more accurate and better compared to other existing plant identification systems. The proposed approach is also tolerant under shape distortion caused while capturing. The mobile machine learning system for leaf image informatics is deployed on Android devices which helps botanists, agriculturists and medical biologists to recognize ubiquitously the herbs and plant species anywhere-anytime .