Machine learning driven non-invasive approach for the detection of anomalies in living plant leaves and water at cellular level using terahertz sensing

Zahid, Adnan (2020) Machine learning driven non-invasive approach for the detection of anomalies in living plant leaves and water at cellular level using terahertz sensing. PhD thesis, University of Glasgow.

Full text available as:
Download (4MB) | Preview


In recent times, an increasing global aridification due to climate transformations and unceasing expansion of population have posed enormous challenges on the environment and its agricultural provision. Researchers and scientists are faced with significant challenges to enhance yield while facing shortage of fertile land due to environmental changes. In this regard, many technologies have been employed to monitor and enhance the crops production. However, certain limitations such as low resolution, destructive nature, cost, sensitivity and reactive nature of technology have markedly reduce their application in modern agriculture. The mounting pressure of more yield with limited fertile land due to environmental changes demands for proactive, cost-effective, real-time, feasible and non-destructive technique in perpetual plants’ health monitoring in order to maintain a healthy physiological status of plants leaves, and to drive the crops productivity and achieve economic benefits. With this motivation in mind, we potentially highlight the evolving application of terahertz (THz) technology (due to its non-ionising and less pervasive radiation properties) with machine learning (ML) for the proactive vegetation monitoring.

In this thesis, we proposed a novel, non-invasive, and cost-effective technique to characterise and estimate the real-time information of water contents (WC) in plants leaves and fruits at cellular level in terms of electromagnetic parameters at THz frequency range from 0.75 to 1.1 THz. It is was noticed that loss observed in WC on day 1 was in the range of 5% to 22%, and increased from 83.12% to 99.33% on day 4. Furthermore, we observed an exponential decaying trend in the peaks of the real part of the permittivity from day 1 to 4, which was reminiscent of the trend observed in the weight of all leaves.

The study also highlights the proactive approach by integrating THz with ML for the accurate and precise estimation of WC in plants and fruits slices including apple and mango, respectively. The results obtained from the amalgamation of ML with THz for the estimation of WC in plants leaves demonstrated that support vector machine (SVM) outperformed other classifiers using tenfold and leave-one-observations out cross-validation for different days classification with an overall accuracy of 98.8%, 97.15%, and 96.82% for Coffee, pea shoot, and baby spinach leaves respectively. In addition, using sequential forward selection (SFS) technique, coffee leaf showed a significant improvement of 15%, 11.9%, 6.5% in computational time for SVM, K-nearest neighbour (KNN) and Decision-tree (D-Tree). For pea-shoot, 21.28%, 10.01%, and 8.53% of improvement was noticed in operating time for SVM, KNN and D-Tree classifiers, respectively. Lastly, baby spinach leaf exhibited a further improvement of 21.28% in SVM, 10.01% in KNN, and 8.53% in D-tree in overall operating time for classifiers.

The results illustrated that the performance of SVM exceeded other classifiers results using 10-fold validation and leave-one-observation-out-cross-validation techniques. Moreover, all three classifiers exhibited 100% accuracy for day 1 and 4 with 80% Moisture content (MC) value (freshness) and 2% MC value (staleness) of both fruits’ slices, respectively. Similarly, for day 2 and 3, an accuracy of 95% was achieved with intermediate MC values in both fruits’ slices.

In addition, in this work, the preservation of clean water without any harmful impurities is also addressed for the health, environmental protection, and economic development. For this purpose, a realistic technological solution method and application of Fourier transform Infrared Spectroscopy (FTIR) operates at THz waves enabled by ML is also discussed in detail. The suggested technique can provide the approximate prediction and detection of even the smallest of contaminants in distilled water due to high sensitivity and non-destructive nature and also produce high optical throughput. Moreover, it was found that random forest (RF) with 97.98%, outperformed other classifiers for estimation of salts concentration added in aqueous solutions. However, for sugar and glucose concentrations, SVM exhibited a higher accuracy of 93.11% and 96.88%, respectively, compared to other classifiers.

The proposed novel study using THz wave and incorporating ML are beneficial and provide prolific recommendations, and insights for cultivators, and horticulturists to take proactive actions in relations to both vegetation and water health monitoring, which in turn, can help in reducing the health and purification expenses by providing early alerts to protect the public health, increase yield with limited land, which will ultimately optimise economic benefits.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Subjects: T Technology > T Technology (General)
Colleges/Schools: College of Science and Engineering > School of Engineering > Electronics and Nanoscale Engineering
Supervisor's Name: Abbasi, Dr. Qammer Hussain and Imran, Professor Muhammad Ali
Date of Award: 2020
Depositing User: MR ADNAN ZAHID
Unique ID: glathesis:2020-81939
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 20 Jan 2021 17:40
Last Modified: 21 Jan 2021 10:07
Thesis DOI: 10.5525/gla.thesis.81939
Related URLs:

Actions (login required)

View Item View Item


Downloads per month over past year