Articles
Construction of a high-resolution digital map to support citrus breeding using an autonomous multicopter
Article number
1135_9
Pages
73 – 84
Language
English
Abstract
This study aimed to develop an end-to-end high-resolution digital mapping system to support evaluation of citrus breeding at an orchard.
Phenotyping of many offspring of citrus breeding by conventional methods is laborious, time-consuming, and often unreliable.
Here, we proposed the implementation of a customized robotic multicopter with autonomous flight control using the global positioning system.
The multicopter can carry a 1.5-kg digital camera with tele-lens for 20 min.
We obtained images and videos from a low altitude at the NARO Institute of Fruit Tree Science, Okitsu Citrus Research Station, Shizuoka, Japan.
We then converted the videos into frames and performed a feature extraction to distinguish structures in the images, such as points, edges, or objects using algorithms.
Furthermore, image alignment based on these feature points was performed to generate a panorama image.
To overcome the requirement of a high-resolution image, a super-resolution algorithm was used to enhance the resolution.
Finally, a 4D map construction was performed.
Both software and hardware were optimized to assist breeders in observing and evaluating candidate offspring.
The system is useful for citrus breeding and can be implemented to other field-based monitoring, such as in rice fields or red clover cultivation.
In general, we overcame the needs of an aerial monitoring system by developing a complete system using an autonomous multicopter and optimizing the image-processing algorithms.
Phenotyping of many offspring of citrus breeding by conventional methods is laborious, time-consuming, and often unreliable.
Here, we proposed the implementation of a customized robotic multicopter with autonomous flight control using the global positioning system.
The multicopter can carry a 1.5-kg digital camera with tele-lens for 20 min.
We obtained images and videos from a low altitude at the NARO Institute of Fruit Tree Science, Okitsu Citrus Research Station, Shizuoka, Japan.
We then converted the videos into frames and performed a feature extraction to distinguish structures in the images, such as points, edges, or objects using algorithms.
Furthermore, image alignment based on these feature points was performed to generate a panorama image.
To overcome the requirement of a high-resolution image, a super-resolution algorithm was used to enhance the resolution.
Finally, a 4D map construction was performed.
Both software and hardware were optimized to assist breeders in observing and evaluating candidate offspring.
The system is useful for citrus breeding and can be implemented to other field-based monitoring, such as in rice fields or red clover cultivation.
In general, we overcame the needs of an aerial monitoring system by developing a complete system using an autonomous multicopter and optimizing the image-processing algorithms.
Authors
M. Haris, K. Ishii, L. Ziyang, T. Sugiura, M. Qi, T. Watanabe, S. Sukisaki, T. Tanabata, S. Isobe, T. Shimizu, T. Yoshioka, H. Nobuhara
Keywords
image processing, feature extraction, image alignment, map viewer, phenotyping, super resolution
Online Articles (18)