Examination of Plant/Weed Image Dataset Using a Hybrid Image Processing Tool

Examination of Plant/Weed Image Dataset Using a Hybrid Image Processing Tool

V. Rajinikanth (St. Joseph's College of Engineering, India), S. Arunmozhi (Manakula Vinayagar Institute of Technology, India), N. Sri Madhava Raja (St. Joseph's College of Engineering, India), B. Parvatha Varthini (St. Joseph's College of Engineering, India) and K. Palani Thanaraj (St. Joseph's College of Engineering, India)
DOI: 10.4018/978-1-5225-8027-0.ch007
OnDemand PDF Download:
List Price: $37.50
10% Discount:-$3.75


Image evaluation procedures are widely employed in various domains to extract the useful information to make the necessary decision. This chapter implements a soft-computing tool to examine the benchmark plant/weed (BPW) pictures of computer vision problems in plant phenotyping (CVPPP2014) challenge database. The proposed work implements a hybrid image evaluation procedure based on social-group optimization algorithm (SGOA) and Shannon's entropy (SE)-based multi-thresholding and the Chan-Vese segmentation (CVS)-based extraction procedure. After extracting the crop/weed regions of BPW pictures, the superiority of the proposed tool is then assessed by implementing a comparative study between the extracted plant/weed region and its equivalent ground-truth. The results of this study substantiate that the proposed system is proficient in examining the BPW pictures and in future. This procedure can be considered to inspect the crop/weed pictures obtained with field supervising drones.
Chapter Preview


In the modern era, most of the country’s Gross Domestic Product (GDP) rate depends mainly on the agriculture sector. In order to get the better yield in farming, a substantial amount of modern cultivation procedures are already implemented in agricultural industries of economically grown countries (Bakhshipour et al., 2017). Modern techniques can be used to monitor the entire plant stages from the seed germination process to the harvesting process (Aitkenhead et al., 2003; Razmjooy et al., 2011).

Smart and Supervised Cultivation (SSC) procedures are widely implemented in most of the countries when there is a demand for the labors and the demand for agricultural products. In SSC, the entire agriculture process, such as seeding, watering, plant monitoring, and harvesting stage monitoring will be carried out with the help of dedicated monitoring devices. When an appropriate network with all the possible sensor and actuators are implemented in this system, the cultivator can monitor the entire process using a dedicated monitoring system such as the personal computer or a mobile phone and he can also initiate or modify the scheduled task based on the climatic condition, crop’s stage, etc (Haug et al., 2014; Herrera et al., 2014).

The main objective of this chapter is to develop an image processing procedure to extract the plant/weed section from the RGB scale pictures recorded using a dedicated digital camera. In most of the plant monitoring cases, a remote digital camera based on a programmed drone or an autonomous robot can be employed to collect the essential plant images from the field based on a chosen resolution. The captured image is simultaneously sent to the monitoring station, which is responsible to perform an image analysis to identify the growth rate of the plant/weed. After the analysis, the outcome is then sent to the operator, who can take the decision regarding the next action to be implemented on the field to protect the plant from the weed and other intimidation (Herrera et al., 2014).

Due to its significance, a considerable number of SSC procedures are proposed and realized to improve the outcome of the agriculture industry. The image-based analysis of plant/weed section is the heart of the SSC process and the ultimate outcome relies mainly on the decision made during the image examination. Hence, in recent years, a number of procedures are proposed by the researchers to examine the plant images recorded with various techniques (Razmjooy et al., 2012; Moallem et al., 2013; Moallem et al., 2014).

The work of this chapter proposes a hybrid image processing technique to extract and examine the plant/weed section from a chosen RGB scale image. The work adopts the Benchmark Crop-Weed (BCW) database supplied by Haug and Ostermann (CVPPP2014) for the investigation (2014). The detailed methodology implemented during the image collection and the associated earlier research works on the considered images can be found in (Haug, S. & Ostermann, 2016). The work of the proposed chapter implements a technique by integrating the pre-processing and post-processing technique to examine the plant/weed region of the CVPPP2014 database. Initially, a multi-level thresholding with Social-Group-Optimization-Algorithm (SGOA) and Shannon’s Entropy (SE) is employed to separate the plant/weed from the background and later the segmentation based on the Chan-Vese is implemented to extract the plant/weed section. Finally, the advantage of the proposed image processing technique is confirmed with relative examination between the extracted region and the ground-truth available in the database.

Complete Chapter List

Search this Book: