{"id":{"repo_id":"auckland-ms","oai_identifier":"oai:researchspace.auckland.ac.nz:2292/72880"},"canonical_url":"https://search.dev.ndltd.org/etd/auckland-ms/oai:researchspace.auckland.ac.nz:2292/72880","repository":{"repo_id":"auckland-ms","name":"University of Auckland","base_url":"https://researchspace.auckland.ac.nz/server/oai/request"},"display":{"title":"Artificial Intelligence-Based Decision Automation for Apple Orchard Thinning","abstract":"Aotearoa New Zealand is a leading apple producer and is projected to export NZ$ 2 billion worth of high-quality pip fruit by 2030. Apple thinning is essential to quality but is also seasonal, leading to a high demand for skilled labour. However, labour shortages have been experienced globally due to the labour-intensive nature of hand-thinning and low wages. Apple thinning involves removing a proportion of fruitlets from clusters to ensure the remaining apples get adequate sunlight, nutrients, and water. Automating this task requires a system capable of accurately detecting clusters and deciding which fruitlets to thin based on physical metrics. This thesis aimed to develop a fruitlet mapping system that extracts key metrics and guides a robot in making precise decisions for accurately removing fruitlets from tree branches. This thesis presents a novel vision system capable of mapping the load and clustering information of apple fruitlets as accurately as a human expert. Accurate mapping of fruitlets along each branch is a complex task which involves quality stereo data collection, point cloud data generation, instance segmentation, fruitlet 3D geometrical representation, and fruitlet tracking. Fruitlet maps have been validated against data collected from a real-world commercial apple orchard. These maps provide the spatial information of the fruitlet clusters and the decisions required to thin them. The best results are achieved using an over-the-row robot platform called “Archie Senior”, which scans both sides of the fruitlet branches, yielding an 87.3% precision and 93.3% recall for load estimates, a 2.06 mm mean absolute error (MAE) for size estimates, an 80.8% orientation success rate, 66.14% thinning accuracy, and 66.7% correct decision-making. These results were obtained after validation using actual fruitlet data in a comprehensive set of experiments conducted during field trials in orchards.","abstract_html":"Aotearoa New Zealand is a leading apple producer and is projected to export NZ$ 2 billion worth of high-quality pip fruit by 2030. Apple thinning is essential to quality but is also seasonal, leading to a high demand for skilled labour. However, labour shortages have been experienced globally due to the labour-intensive nature of hand-thinning and low wages. Apple thinning involves removing a proportion of fruitlets from clusters to ensure the remaining apples get adequate sunlight, nutrients, and water. Automating this task requires a system capable of accurately detecting clusters and deciding which fruitlets to thin based on physical metrics. This thesis aimed to develop a fruitlet mapping system that extracts key metrics and guides a robot in making precise decisions for accurately removing fruitlets from tree branches. This thesis presents a novel vision system capable of mapping the load and clustering information of apple fruitlets as accurately as a human expert. Accurate mapping of fruitlets along each branch is a complex task which involves quality stereo data collection, point cloud data generation, instance segmentation, fruitlet 3D geometrical representation, and fruitlet tracking. Fruitlet maps have been validated against data collected from a real-world commercial apple orchard. These maps provide the spatial information of the fruitlet clusters and the decisions required to thin them. The best results are achieved using an over-the-row robot platform called “Archie Senior”, which scans both sides of the fruitlet branches, yielding an 87.3% precision and 93.3% recall for load estimates, a 2.06 mm mean absolute error (MAE) for size estimates, an 80.8% orientation success rate, 66.14% thinning accuracy, and 66.7% correct decision-making. These results were obtained after validation using actual fruitlet data in a comprehensive set of experiments conducted during field trials in orchards.","abstract_has_math":false,"creators":["Qureshi, Ans Hussain"],"institution":"ResearchSpace@Auckland","degree_name":"PhD","degree_level":"Doctoral","degree_discipline":"Computer Systems Engineering","degree_department":null,"school":null,"contributors":[],"advisors":["Ahn, Ho Seok","MacDonald, Bruce"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-06-17","date_published":"2025-06-17","updated_at":"2026-07-24T01:05:49Z","subjects":["Agricultural Robotics","Orchard Automation","Robot Vision"],"languages":[],"rights":["Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated."],"rights_urls":["https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm"],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/2292/72880","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Ahn, Ho Seok","MacDonald, Bruce"]},{"key":"dc:creator","label":"Author","values":["Qureshi, Ans Hussain"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-07-13T21:35:19Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-07-13T21:35:19Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-06-17"]},{"key":"dc:publisher","label":"Institution","values":["ResearchSpace@Auckland"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Systems Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Doctoral"]},{"key":"thesis:degree_name","label":"Degree Name","values":["PhD"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["The University of Auckland"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Agricultural Robotics","Orchard Automation","Robot Vision"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated."]},{"key":"dc:rights.uri","label":"Rights URI","values":["https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/2292/72880"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Aotearoa New Zealand is a leading apple producer and is projected to export NZ$ 2 billion worth of high-quality pip fruit by 2030. Apple thinning is essential to quality but is also seasonal, leading to a high demand for skilled labour. However, labour shortages have been experienced globally due to the labour-intensive nature of hand-thinning and low wages. Apple thinning involves removing a proportion of fruitlets from clusters to ensure the remaining apples get adequate sunlight, nutrients, and water. Automating this task requires a system capable of accurately detecting clusters and deciding which fruitlets to thin based on physical metrics. This thesis aimed to develop a fruitlet mapping system that extracts key metrics and guides a robot in making precise decisions for accurately removing fruitlets from tree branches. This thesis presents a novel vision system capable of mapping the load and clustering information of apple fruitlets as accurately as a human expert. Accurate mapping of fruitlets along each branch is a complex task which involves quality stereo data collection, point cloud data generation, instance segmentation, fruitlet 3D geometrical representation, and fruitlet tracking. Fruitlet maps have been validated against data collected from a real-world commercial apple orchard. These maps provide the spatial information of the fruitlet clusters and the decisions required to thin them. The best results are achieved using an over-the-row robot platform called “Archie Senior”, which scans both sides of the fruitlet branches, yielding an 87.3% precision and 93.3% recall for load estimates, a 2.06 mm mean absolute error (MAE) for size estimates, an 80.8% orientation success rate, 66.14% thinning accuracy, and 66.7% correct decision-making. These results were obtained after validation using actual fruitlet data in a comprehensive set of experiments conducted during field trials in orchards."]},{"key":"dc:title","label":"Title","values":["Artificial Intelligence-Based Decision Automation for Apple Orchard Thinning"]}]}],"canonical_facts":{"dc:contributor.advisor":["Ahn, Ho Seok","MacDonald, Bruce"],"dc:creator":["Qureshi, Ans Hussain"],"dc:date.accessioned":["2025-07-13T21:35:19Z"],"dc:date.available":["2025-07-13T21:35:19Z"],"dc:date.issued":["2025-06-17"],"dc:description.abstract":["Aotearoa New Zealand is a leading apple producer and is projected to export NZ$ 2 billion worth of high-quality pip fruit by 2030. Apple thinning is essential to quality but is also seasonal, leading to a high demand for skilled labour. However, labour shortages have been experienced globally due to the labour-intensive nature of hand-thinning and low wages. Apple thinning involves removing a proportion of fruitlets from clusters to ensure the remaining apples get adequate sunlight, nutrients, and water. Automating this task requires a system capable of accurately detecting clusters and deciding which fruitlets to thin based on physical metrics. This thesis aimed to develop a fruitlet mapping system that extracts key metrics and guides a robot in making precise decisions for accurately removing fruitlets from tree branches. This thesis presents a novel vision system capable of mapping the load and clustering information of apple fruitlets as accurately as a human expert. Accurate mapping of fruitlets along each branch is a complex task which involves quality stereo data collection, point cloud data generation, instance segmentation, fruitlet 3D geometrical representation, and fruitlet tracking. Fruitlet maps have been validated against data collected from a real-world commercial apple orchard. These maps provide the spatial information of the fruitlet clusters and the decisions required to thin them. The best results are achieved using an over-the-row robot platform called “Archie Senior”, which scans both sides of the fruitlet branches, yielding an 87.3% precision and 93.3% recall for load estimates, a 2.06 mm mean absolute error (MAE) for size estimates, an 80.8% orientation success rate, 66.14% thinning accuracy, and 66.7% correct decision-making. These results were obtained after validation using actual fruitlet data in a comprehensive set of experiments conducted during field trials in orchards."],"dc:identifier.uri":["https://hdl.handle.net/2292/72880"],"dc:publisher":["ResearchSpace@Auckland"],"dc:rights":["Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated."],"dc:rights.uri":["https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm"],"dc:subject":["Agricultural Robotics","Orchard Automation","Robot Vision"],"dc:title":["Artificial Intelligence-Based Decision Automation for Apple Orchard Thinning"],"dc:type":["Thesis"],"thesis:degree_discipline":["Computer Systems Engineering"],"thesis:degree_level":["Doctoral"],"thesis:degree_name":["PhD"],"thesis:institution_name":["The University of Auckland"]},"updated_at":"2026-07-24T01:05:49Z"}