from the collar to the base of the highest fully extended leaf .
An open-source photogrammetric software , ColMap , was used to generate the 3D point cloud using the captured images ( Figure 1b ) and exported into another open-source 3D modelling software to filter the environment of any unwanted objects ( Figure 1c ).
To determine the virtual height of the plant using the 3D point cloud , a scaling factor that relates the real world to the virtual measurement is required . For this purpose , the planting medium , with a 2cm height ( Figure 1c ), was used to determine the scaling factor . The virtual plant height was measured from the highest point to the base of the plant ’ s point cloud . This value is then multiplied by the scaling factor and carries the same unit as the real-world dimension .
To compare the real world to the virtual measurements , linear regression was performed , and the root mean square error ( RMSE ) was calculated . Particularly good agreements between the two measurements were found ( see Figure 2 ) highlighting that the 3D point cloud reconstructed with the photogrammetry technique using images captured from mobile phone cameras can produce comparable real-world plant height measurements for crop growth monitoring .
At the Centre , we also utilised 3D modelling to assist in the post-harvest handling of papaya fruit . A 3D model of the fruit was used with finite element analysis to determine the mechanical injury extent ( see Figure 3 ). The results informed the best orientation for storing or transporting papayas to minimise mechanical damage . This is crucial due to the papayas ’ thin skin , irregular shape and high water content , particularly as they ripen .
Pros and Cons of using 3D Models for the Agrifood Industry
Using 3D point clouds for crop growth monitoring enables farmers and agronomists to gain a deeper understanding of plant health , growth patterns , and environmental interactions . This allows for precise monitoring and management strategies tailored to specific areas within a field , optimising resource use such as water , fertiliser , and electricity . Furthermore , 3D models enable enhanced data analysis through tools like spatial analytics and machine learning algorithms , facilitating predictive modelling for yield estimation and disease detection . The ability to visualise and analyse data in three dimensions provides a holistic view of crop conditions , empowering decision-making processes aimed at maximising productivity and sustainability in agriculture .
Despite its advantages , 3D modelling in agricultural settings faces several challenges . Most importantly , ensuring data accuracy and reliability remains a critical concern , as errors in data collection and / or processing can lead to inaccurate models and subsequent decisionmaking . It is well established that environmental factors such as varying light conditions , weather changes , and seasonal variations can impact the quality and consistency of data collection , affecting the accuracy of 3D models . Moreover , the complexity of agricultural landscapes , including uneven terrain and dense vegetation , poses challenges for data acquisition using traditional methods like UAVs or ground-based sensors . Additionally , the scalability and cost-effectiveness of deploying 3D modelling technologies across large agricultural operations require careful consideration , balancing the potential benefits with the challenges of practical implementation . Addressing these challenges involves on-going advancements in technology , methods , and data processing techniques tailored to the unique needs of agricultural applications .
Future Directions
For crop monitoring , emerging trends in 3D modelling technology include advancements in remote sensing techniques such as hyperspectral imaging and multi-sensor fusion . These technologies allow for more comprehensive data collection , providing insights into crop health beyond visual and geometric data . Integration of artificial intelligence and machine learning algorithms is another trend , enabling automated analysis of 3D models to detect subtle changes in crop conditions , predict yield outcomes , and diagnose diseases at early stages . Furthermore , advancements in cloud computing and data analytics offer scalability and real-time processing
39