{"id":{"repo_id":"stellenbosch","oai_identifier":"oai:scholar.sun.ac.za:10019.1/135731"},"canonical_url":"https://search.dev.ndltd.org/etd/stellenbosch/oai:scholar.sun.ac.za:10019.1/135731","repository":{"repo_id":"stellenbosch","name":"Stellenbosch University","base_url":"https://scholar.sun.ac.za/server/oai/request"},"display":{"title":"Development of a Low-Cost System for Monitoring Root Growth Dynamics","abstract":"Understanding root system dynamics is crucial for improving plant resilience and optimising resource allocation in agriculture and forestry. Despite the importance of monitoring root dynamics, roots remain difficult to study because of their below-ground structure. Minirhizotrons – root imaging tools that use transparent tubes installed in the soil to allow for the non-destructive, in situ observation of plant roots – offer a non-invasive solution for capturing root images over time. However, existing products are typically expensive or manually operated. These limitations restrict access to dynamic root data, particularly for small-scale research. The solution presented in this study bridges this gap through the development of a low-cost and automated minirhizotron to collect dynamic root data non-invasively. This automated data collection system was integrated with a deep-learning root segmentation model and feature extraction algorithms for quantifying key root traits over time. The minirhizotron hardware was constructed from a glass cylinder for high optical clarity, controlled by a RPi Zero 2 W and ZeroCam module, and powered by 5 000 mAh LiPo batteries and a solar panel for long-term operation in field-like conditions. Custom scripts on the RPi Zero 2W were developed to automate image capture, illumination and wireless image transmission. The system was installed in an outdoor nursery and left unattended for a period of 52 days, capturing images at frequent intervals. The total cost of this system was R2 774 – substantially lower than existing commercial alternatives, improving the accessibility of dynamic root data to small-scale research and making large-scale deployment financially feasible. To segment the roots from the minirhizotron data, a deep-learning model was trained on an external dataset of 62 000 annotated minirhizotron images, spanning five root species and diverse soil conditions. The final segmentation model achieved an accuracy of 98,7% and a loss of 0,0317%. Although quantitative performance on the external dataset was limited by coarsely annotated ground truth masks – indicated by a Dice coefficient of 0,63 – qualitative assessment showed that the model produced realistic organic root shapes. When applied to unseen Eucalyptus time-series data, the model achieved an accuracy of 99,6% and a Dice coefficient of 0,67, demonstrating its ability to generalise to new root species. A set of feature extraction algorithms was developed and implemented to quantify temporal root traits – including total root length, convex hull area, surface area and primary root length – from the predicted masks. Despite systematic underestimation of absolute trait magnitudes due to the segmentation model’s limitation in detecting fine lateral roots, the predicted growth trends closely aligned with the ground truth growth trends and developmental phase transitions, with high correlation coefficients obtained when comparing predicted and ground truth dynamic behaviour. This research contributes to precision forestry and agriculture by quantifying root growth trends and developmental phases to support optimal resource allocation. Overall, this study demonstrates a low-cost and scalable system for dynamic root data collection and root growth monitoring.","abstract_html":"Understanding root system dynamics is crucial for improving plant resilience and optimising resource allocation in agriculture and forestry. Despite the importance of monitoring root dynamics, roots remain difficult to study because of their below-ground structure. Minirhizotrons – root imaging tools that use transparent tubes installed in the soil to allow for the non-destructive, in situ observation of plant roots – offer a non-invasive solution for capturing root images over time. However, existing products are typically expensive or manually operated. These limitations restrict access to dynamic root data, particularly for small-scale research. The solution presented in this study bridges this gap through the development of a low-cost and automated minirhizotron to collect dynamic root data non-invasively. This automated data collection system was integrated with a deep-learning root segmentation model and feature extraction algorithms for quantifying key root traits over time. The minirhizotron hardware was constructed from a glass cylinder for high optical clarity, controlled by a RPi Zero 2 W and ZeroCam module, and powered by 5 000 mAh LiPo batteries and a solar panel for long-term operation in field-like conditions. Custom scripts on the RPi Zero 2W were developed to automate image capture, illumination and wireless image transmission. The system was installed in an outdoor nursery and left unattended for a period of 52 days, capturing images at frequent intervals. The total cost of this system was R2 774 – substantially lower than existing commercial alternatives, improving the accessibility of dynamic root data to small-scale research and making large-scale deployment financially feasible. To segment the roots from the minirhizotron data, a deep-learning model was trained on an external dataset of 62 000 annotated minirhizotron images, spanning five root species and diverse soil conditions. The final segmentation model achieved an accuracy of 98,7% and a loss of 0,0317%. Although quantitative performance on the external dataset was limited by coarsely annotated ground truth masks – indicated by a Dice coefficient of 0,63 – qualitative assessment showed that the model produced realistic organic root shapes. When applied to unseen Eucalyptus time-series data, the model achieved an accuracy of 99,6% and a Dice coefficient of 0,67, demonstrating its ability to generalise to new root species. A set of feature extraction algorithms was developed and implemented to quantify temporal root traits – including total root length, convex hull area, surface area and primary root length – from the predicted masks. Despite systematic underestimation of absolute trait magnitudes due to the segmentation model’s limitation in detecting fine lateral roots, the predicted growth trends closely aligned with the ground truth growth trends and developmental phase transitions, with high correlation coefficients obtained when comparing predicted and ground truth dynamic behaviour. This research contributes to precision forestry and agriculture by quantifying root growth trends and developmental phases to support optimal resource allocation. Overall, this study demonstrates a low-cost and scalable system for dynamic root data collection and root growth monitoring.","abstract_has_math":false,"creators":["De Raay, Yasmin"],"institution":"Stellenbosch : Stellenbosch University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Van der Merwe, Andre","Drew, David"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-03","date_published":"2026-03","updated_at":"2026-07-24T04:40:09Z","subjects":[],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://scholar.sun.ac.za/handle/10019.1/135731","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Van der Merwe, Andre","Drew, David"]},{"key":"dc:contributor.other","label":"Dc Contributor Other","values":["Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering."]},{"key":"dc:creator","label":"Author","values":["De Raay, Yasmin"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-04-09T06:04:44Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-04-09T06:04:44Z"]},{"key":"dc:date.issued","label":"Date","values":["2026-03"]},{"key":"dc:publisher","label":"Institution","values":["Stellenbosch : Stellenbosch University"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://scholar.sun.ac.za/handle/10019.1/135731"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (MEng)--Stellenbosch University, 2026.","De Raay, Y. 2026. Development of a Low-Cost System for Monitoring Root Growth Dynamics. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/711000ac-a528-4f4f-bff2-a0a80749c5fe"]},{"key":"dc:description.abstract","label":"Abstract","values":["Understanding root system dynamics is crucial for improving plant resilience and optimising resource allocation in agriculture and forestry. Despite the importance of monitoring root dynamics, roots remain difficult to study because of their below-ground structure. Minirhizotrons – root imaging tools that use transparent tubes installed in the soil to allow for the non-destructive, in situ observation of plant roots – offer a non-invasive solution for capturing root images over time. However, existing products are typically expensive or manually operated. These limitations restrict access to dynamic root data, particularly for small-scale research. The solution presented in this study bridges this gap through the development of a low-cost and automated minirhizotron to collect dynamic root data non-invasively. This automated data collection system was integrated with a deep-learning root segmentation model and feature extraction algorithms for quantifying key root traits over time. The minirhizotron hardware was constructed from a glass cylinder for high optical clarity, controlled by a RPi Zero 2 W and ZeroCam module, and powered by 5 000 mAh LiPo batteries and a solar panel for long-term operation in field-like conditions. Custom scripts on the RPi Zero 2W were developed to automate image capture, illumination and wireless image transmission. The system was installed in an outdoor nursery and left unattended for a period of 52 days, capturing images at frequent intervals. The total cost of this system was R2 774 – substantially lower than existing commercial alternatives, improving the accessibility of dynamic root data to small-scale research and making large-scale deployment financially feasible. To segment the roots from the minirhizotron data, a deep-learning model was trained on an external dataset of 62 000 annotated minirhizotron images, spanning five root species and diverse soil conditions. The final segmentation model achieved an accuracy of 98,7% and a loss of 0,0317%. Although quantitative performance on the external dataset was limited by coarsely annotated ground truth masks – indicated by a Dice coefficient of 0,63 – qualitative assessment showed that the model produced realistic organic root shapes. When applied to unseen Eucalyptus time-series data, the model achieved an accuracy of 99,6% and a Dice coefficient of 0,67, demonstrating its ability to generalise to new root species. A set of feature extraction algorithms was developed and implemented to quantify temporal root traits – including total root length, convex hull area, surface area and primary root length – from the predicted masks. Despite systematic underestimation of absolute trait magnitudes due to the segmentation model’s limitation in detecting fine lateral roots, the predicted growth trends closely aligned with the ground truth growth trends and developmental phase transitions, with high correlation coefficients obtained when comparing predicted and ground truth dynamic behaviour. This research contributes to precision forestry and agriculture by quantifying root growth trends and developmental phases to support optimal resource allocation. Overall, this study demonstrates a low-cost and scalable system for dynamic root data collection and root growth monitoring."]},{"key":"dc:title","label":"Title","values":["Development of a Low-Cost System for Monitoring Root Growth Dynamics"]}]}],"canonical_facts":{"dc:contributor.advisor":["Van der Merwe, Andre","Drew, David"],"dc:contributor.other":["Stellenbosch University. Faculty of Engineering. Dept. of Industrial Engineering."],"dc:creator":["De Raay, Yasmin"],"dc:date.accessioned":["2026-04-09T06:04:44Z"],"dc:date.available":["2026-04-09T06:04:44Z"],"dc:date.issued":["2026-03"],"dc:description":["Thesis (MEng)--Stellenbosch University, 2026.","De Raay, Y. 2026. Development of a Low-Cost System for Monitoring Root Growth Dynamics. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/711000ac-a528-4f4f-bff2-a0a80749c5fe"],"dc:description.abstract":["Understanding root system dynamics is crucial for improving plant resilience and optimising resource allocation in agriculture and forestry. Despite the importance of monitoring root dynamics, roots remain difficult to study because of their below-ground structure. Minirhizotrons – root imaging tools that use transparent tubes installed in the soil to allow for the non-destructive, in situ observation of plant roots – offer a non-invasive solution for capturing root images over time. However, existing products are typically expensive or manually operated. These limitations restrict access to dynamic root data, particularly for small-scale research. The solution presented in this study bridges this gap through the development of a low-cost and automated minirhizotron to collect dynamic root data non-invasively. This automated data collection system was integrated with a deep-learning root segmentation model and feature extraction algorithms for quantifying key root traits over time. The minirhizotron hardware was constructed from a glass cylinder for high optical clarity, controlled by a RPi Zero 2 W and ZeroCam module, and powered by 5 000 mAh LiPo batteries and a solar panel for long-term operation in field-like conditions. Custom scripts on the RPi Zero 2W were developed to automate image capture, illumination and wireless image transmission. The system was installed in an outdoor nursery and left unattended for a period of 52 days, capturing images at frequent intervals. The total cost of this system was R2 774 – substantially lower than existing commercial alternatives, improving the accessibility of dynamic root data to small-scale research and making large-scale deployment financially feasible. To segment the roots from the minirhizotron data, a deep-learning model was trained on an external dataset of 62 000 annotated minirhizotron images, spanning five root species and diverse soil conditions. The final segmentation model achieved an accuracy of 98,7% and a loss of 0,0317%. Although quantitative performance on the external dataset was limited by coarsely annotated ground truth masks – indicated by a Dice coefficient of 0,63 – qualitative assessment showed that the model produced realistic organic root shapes. When applied to unseen Eucalyptus time-series data, the model achieved an accuracy of 99,6% and a Dice coefficient of 0,67, demonstrating its ability to generalise to new root species. A set of feature extraction algorithms was developed and implemented to quantify temporal root traits – including total root length, convex hull area, surface area and primary root length – from the predicted masks. Despite systematic underestimation of absolute trait magnitudes due to the segmentation model’s limitation in detecting fine lateral roots, the predicted growth trends closely aligned with the ground truth growth trends and developmental phase transitions, with high correlation coefficients obtained when comparing predicted and ground truth dynamic behaviour. This research contributes to precision forestry and agriculture by quantifying root growth trends and developmental phases to support optimal resource allocation. Overall, this study demonstrates a low-cost and scalable system for dynamic root data collection and root growth monitoring."],"dc:identifier.uri":["https://scholar.sun.ac.za/handle/10019.1/135731"],"dc:language.iso":["en"],"dc:publisher":["Stellenbosch : Stellenbosch University"],"dc:title":["Development of a Low-Cost System for Monitoring Root Growth Dynamics"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T04:40:09Z"}