On solving leaf classification using linear regression
- PDF / 2,054,021 Bytes
- 19 Pages / 439.642 x 666.49 pts Page_size
- 40 Downloads / 207 Views
On solving leaf classification using linear regression Neha Goyal1
· Nitin Kumar1 · Kapil1
Received: 18 March 2020 / Revised: 28 July 2020 / Accepted: 16 September 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Plant’s conservation is getting close attention nowadays. It requires awareness about ecology among masses. Plant species identification has been proved as a primary step in literature for biodiversity conservation. It is a sequential process from leaf images as input followed by image enhancement algorithms, and feature extraction phase to classification. The complete process of identifying a leaf image requires substantial time. The article focuses on introducing a simpler and computationally inexpensive framework with a performance at par or better as compared to the existing framework. The article covers several findings and results while transforming the proposed framework for plant identification to a parameter specific optimized framework. The findings include optimizing the leaf image dimension, the impact of RGB to grayscale conversion method, and comparative analysis of the proposed framework for classification from images with other frameworks that first extract specific features and then classify. It also represents the whole framework as a regression problem. Further, improvement is incorporated by integrating the benefits of kernel trick in linear regression. Our finding confirms that the framework not only recognizing the leaf images with comparable accuracy but also reduces the computational time significantly to identify leaf images as compared to other frameworks. Keywords Linear regression · Kernel function · Color to gray-scale conversion · Image down-sampling · Image projection
1 Introduction Plants are globally identified as an essential factor in biological diversity and an essential resource for the planet and also underpins the functioning of the entire ecosystem. There are an estimated 500,000 species of plants [3]. One-third of plant species are threatened Neha Goyal
[email protected] Nitin Kumar [email protected] Kapil [email protected] 1
NIT, Kurukshetra, Uttarakhand, India
Multimedia Tools and Applications
and several plant families that are still un-subscribed or un-documented. Only a few plant species are specific to human use. Several plants play crucial roles in natural ecosystems due to the rare services provided by such plant species. Some plant species are more likely to have important characteristics that may prove to be of utmost importance in future. The significant threats to plant diversity include habitat loss, fragmentation, forest degradation, over exploitation of resources, invasive species, increasing pollution, and climate change [3, 11, 26] etc. Conservation of plant diversity is a challenging task worldwide. Plant species recognition is a necessary step towards diversity conservation due to several reasons e.g. it helps to aware people of plant traits and resources we are getting from the ecosystem and a
Data Loading...
