An intelligent navigation system for oil palm plantations
In order to introduce automation to oil palm plantations as a solution to a reduction in workforce, a smart navigation system capable of operating autonomously was required. This thesis focuses on the creation of an intelligent navigation system for oil palm plantations. For the successful creation...
| Main Author: | |
|---|---|
| Format: | Thesis (University of Nottingham only) |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | https://eprints.nottingham.ac.uk/55822/ |
| _version_ | 1848799222497804288 |
|---|---|
| author | Juman, Mohammed Ayoub |
| author_facet | Juman, Mohammed Ayoub |
| author_sort | Juman, Mohammed Ayoub |
| building | Nottingham Research Data Repository |
| collection | Online Access |
| description | In order to introduce automation to oil palm plantations as a solution to a reduction in workforce, a smart navigation system capable of operating autonomously was required. This thesis focuses on the creation of an intelligent navigation system for oil palm plantations. For the successful creation of the system, an object detection system to detect tree trunks, a path planner to calculate the optimum path to the tree and a trajectory controller to accurately move along the path was necessary. A robot rover was designed and built to test the created system, and novelties were presented in the object detection as well as the trajectory control systems to tackle the difficulties faced due to the plantation environment.
A novel tree trunk detection algorithm was proposed for object detection, focusing on oil palm trees in the plantation, as they were the key objects of interest. The algorithm used the Viola and Jones detector with a proposed pre-processing method along with depth information to tackle the issue of high false positive rates when the Viola and Jones detector was used on its own. The pre-processing method used colour space combination and segmentation to eliminate the ground not covered by trees and fed the resulting image into a cascade detector. The method had better performance when compared to both Neural Network and Support Vector Machine based detectors with a detection rate of 91.7% and had the lowest false acceptance rate overall, as well as a 97.8% detection rate during real time testing with the use of the robot, proving its high accuracy. Depth information obtained via the Microsoft KINECT sensor, resulted in a 100% detection rate of tree trunks within the sensor’s range during low light conditions.
The nearest detected tree as well as the distance to it was obtained and marked as the goal point, while the surrounding obstacles were detected via ultrasonic sensors. This information was then passed onto the path planning system, created by a modified D*lite algorithm, to plan an optimal path to the tree trunk from the robot’s current position.
The trajectory control system was based on the Enhanced Self Organizing Incremental Neural Network (ESOINN), which was able to produce exceptional trajectory control without the use of a kinematic / dynamic model of the robot while consistently reducing errors during incremental learning. However, the ESOINN was only capable of discrete outputs, which limited the motions capable by the robot. A novel incremental unsupervised learning algorithm that is capable of producing continuous outputs, called IDW-ESOINN, was proposed by incorporating an interpolation method known as Inverse Distance Weighting (IDW) to extract outputs from three of the nearest nodes to the data point and combining them based on their distance to the data. The method also led to the creation of virtual nodes that enabled more nodes to be produced in empty areas during the next stage of incremental learning. The IDW-ESOINN had a higher performance accuracy than the original ESOINN as well as the Kalman filter method in simulated path following tests, with a consistently low RMSE, while having the lowest RMSE 60% of the time when compared with the other two methods. In real-time testing, the IDW-ESOINN outperformed the original ESOINN again, proving its functionality.
Based on the results obtained via the various tests done, it can be concluded that the created navigation system had a high level of performance, which could lead to fully autonomous robots being used in oil palm plantations one step closer to realisation. |
| first_indexed | 2025-11-14T20:32:14Z |
| format | Thesis (University of Nottingham only) |
| id | nottingham-55822 |
| institution | University of Nottingham Malaysia Campus |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-14T20:32:14Z |
| publishDate | 2019 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | nottingham-558222025-02-28T14:21:01Z https://eprints.nottingham.ac.uk/55822/ An intelligent navigation system for oil palm plantations Juman, Mohammed Ayoub In order to introduce automation to oil palm plantations as a solution to a reduction in workforce, a smart navigation system capable of operating autonomously was required. This thesis focuses on the creation of an intelligent navigation system for oil palm plantations. For the successful creation of the system, an object detection system to detect tree trunks, a path planner to calculate the optimum path to the tree and a trajectory controller to accurately move along the path was necessary. A robot rover was designed and built to test the created system, and novelties were presented in the object detection as well as the trajectory control systems to tackle the difficulties faced due to the plantation environment. A novel tree trunk detection algorithm was proposed for object detection, focusing on oil palm trees in the plantation, as they were the key objects of interest. The algorithm used the Viola and Jones detector with a proposed pre-processing method along with depth information to tackle the issue of high false positive rates when the Viola and Jones detector was used on its own. The pre-processing method used colour space combination and segmentation to eliminate the ground not covered by trees and fed the resulting image into a cascade detector. The method had better performance when compared to both Neural Network and Support Vector Machine based detectors with a detection rate of 91.7% and had the lowest false acceptance rate overall, as well as a 97.8% detection rate during real time testing with the use of the robot, proving its high accuracy. Depth information obtained via the Microsoft KINECT sensor, resulted in a 100% detection rate of tree trunks within the sensor’s range during low light conditions. The nearest detected tree as well as the distance to it was obtained and marked as the goal point, while the surrounding obstacles were detected via ultrasonic sensors. This information was then passed onto the path planning system, created by a modified D*lite algorithm, to plan an optimal path to the tree trunk from the robot’s current position. The trajectory control system was based on the Enhanced Self Organizing Incremental Neural Network (ESOINN), which was able to produce exceptional trajectory control without the use of a kinematic / dynamic model of the robot while consistently reducing errors during incremental learning. However, the ESOINN was only capable of discrete outputs, which limited the motions capable by the robot. A novel incremental unsupervised learning algorithm that is capable of producing continuous outputs, called IDW-ESOINN, was proposed by incorporating an interpolation method known as Inverse Distance Weighting (IDW) to extract outputs from three of the nearest nodes to the data point and combining them based on their distance to the data. The method also led to the creation of virtual nodes that enabled more nodes to be produced in empty areas during the next stage of incremental learning. The IDW-ESOINN had a higher performance accuracy than the original ESOINN as well as the Kalman filter method in simulated path following tests, with a consistently low RMSE, while having the lowest RMSE 60% of the time when compared with the other two methods. In real-time testing, the IDW-ESOINN outperformed the original ESOINN again, proving its functionality. Based on the results obtained via the various tests done, it can be concluded that the created navigation system had a high level of performance, which could lead to fully autonomous robots being used in oil palm plantations one step closer to realisation. 2019-02-23 Thesis (University of Nottingham only) NonPeerReviewed application/pdf en arr https://eprints.nottingham.ac.uk/55822/1/Mohammed%20Ayoub%20Juman%20-%20PhD%20Thesis.pdf Juman, Mohammed Ayoub (2019) An intelligent navigation system for oil palm plantations. PhD thesis, University of Nottingham. navigation system |
| spellingShingle | navigation system Juman, Mohammed Ayoub An intelligent navigation system for oil palm plantations |
| title | An intelligent navigation system for oil palm plantations |
| title_full | An intelligent navigation system for oil palm plantations |
| title_fullStr | An intelligent navigation system for oil palm plantations |
| title_full_unstemmed | An intelligent navigation system for oil palm plantations |
| title_short | An intelligent navigation system for oil palm plantations |
| title_sort | intelligent navigation system for oil palm plantations |
| topic | navigation system |
| url | https://eprints.nottingham.ac.uk/55822/ |