Rawat, Ishank Kumar (2016) Plant Identification from Leaves using Pattern Recognition Techniques. MTech thesis.
Medicinal plants have been used throughout the human history. Ayurveda is one of the oldest medicine system, which is even recognized in the modern medical society, uses plants for the preparation of medicines. There are thousands of species of plants used in the preparation of medicines. The difficulty lies in the identification of plant species. An individual with deep knowledge of plants can only differentiate between these species. This makes leaf identification very difficult. A reference guide to plants identification may ease up the problems. This is where nature needs engineering. In this work, a system is being developed which helps in the identification of the plants based on the leaf. This system takes input as a leaf image and outputs the name of the species and other relevant details which are stored in the database. The system is designed using the technique of image identification using pattern recognition. The approach of shape and texture identification, both are combined for designing such a system. The segmentation of the images was done using the techniques of graph-cuts. The descriptor used for shape identification was Shape Context and textures were described using Local Binary Patterns. The classification was done using feed forward Multi-Layered Perceptron (MLP) neural network with backpropagation training algorithm. The system was tested of certain class of leaves and the performance of the system is compared with an existing system.
|Item Type:||Thesis (MTech)|
|Uncontrolled Keywords:||Graph Cuts, Shape Context, Local Binary Patterns, Multi-Layered Perceptron (MLP).|
|Subjects:||Engineering and Technology > Electronics and Communication Engineering > Signal Processing|
|Divisions:||Engineering and Technology > Department of Electronics and Communication Engineering|
|Deposited By:||Mr. Sanat Kumar Behera|
|Deposited On:||18 Sep 2017 11:40|
|Last Modified:||05 Dec 2019 18:09|
|Supervisor(s):||Sahoo, Ajit Kumar|
Repository Staff Only: item control page