Vineyard trunk detection using deep learning - An experimental device benchmark

被引:45
作者
Pinto de Aguiar, Andre Silva [1 ,2 ]
Neves dos Santos, Filipe Baptista [1 ]
Feliz dos Santos, Luis Carlos [1 ,2 ]
de Jesus Filipe, Vitor Manuel [1 ,2 ]
Miranda de Sousa, Armando Jorge [1 ,3 ]
机构
[1] INESC TEC INESC Technol & Sci, P-4200465 Porto, Portugal
[2] Univ Tras Os Montes & Alto Douro, Sch Sci & Technol, P-5000801 Vila Real, Portugal
[3] Univ Porto, Fac Engn, P-4200465 Porto, Portugal
关键词
Deep learning; Object detection; Agricultural robots; NEURAL-NETWORKS; AGRICULTURE;
D O I
10.1016/j.compag.2020.105535
中图分类号
S [农业科学];
学科分类号
09 ;
摘要
Research and development in mobile robotics are continuously growing. The ability of a human-made machine to navigate safely in a given environment is a challenging task. In agricultural environments, robot navigation can achieve high levels of complexity due to the harsh conditions that they present. Thus, the presence of a reliable map where the robot can localize itself is crucial, and feature extraction becomes a vital step of the navigation process. In this work, the feature extraction issue in the vineyard context is solved using Deep Learning to detect high-level features - the vine trunks. An experimental performance benchmark between two devices is performed: NVIDIA's Jetson Nano and Google's USB Accelerator. Several models were retrained and deployed on both devices, using a Transfer Learning approach. Specifically, MobileNets, Inception, and lite version of You Only Look Once are used to detect vine trunks in real-time. The models were retrained in a built in-house dataset, that is publicly available. The training dataset contains approximately 1600 annotated vine trunks in 336 different images. Results show that NVIDIA's Jetson Nano provides compatibility with a wider variety of Deep Learning architectures, while Google's USB Accelerator is limited to a unique family of architectures to perform object detection. On the other hand, the Google device showed an overall Average precision higher than Jetson Nano, with a better runtime performance. The best result obtained in this work was an average precision of 52.98% with a runtime performance of 23.14 ms per image, for MobileNet-V2. Recent experiments showed that the detectors are suitable for the use in the Localization and Mapping context.
引用
收藏
页数:9
相关论文
共 52 条
[1]   Monocular Visual Odometry Using Fisheye Lens Cameras [J].
Aguiar, Andre ;
Santos, Filipe ;
Santos, Luis ;
Sousa, Armando .
PROGRESS IN ARTIFICIAL INTELLIGENCE, PT II, 2019, 11805 :319-330
[2]   The Alto Douro Wine Region greenway [J].
Andresen, T ;
de Aguiar, FB ;
Curado, MJ .
LANDSCAPE AND URBAN PLANNING, 2004, 68 (2-3) :289-303
[3]  
[Anonymous], ABS151202325 CORR
[4]  
[Anonymous], 2018, ARXIV180703284
[5]   Optimized EIF-SLAM algorithm for precision agriculture mapping based on stems detection [J].
Auat Cheein, F. ;
Steiner, G. ;
Perez Paina, G. ;
Carelli, R. .
COMPUTERS AND ELECTRONICS IN AGRICULTURE, 2011, 78 (02) :195-207
[6]   Compound Autoregressive Network for Prediction of Multivariate Time Series [J].
Bai, Yuting ;
Jin, Xuebo ;
Wang, Xiaoyi ;
Su, Tingli ;
Kong, Jianlei ;
Lu, Yutian .
COMPLEXITY, 2019, 2019
[7]  
Bargoti Suchet, 2017, 2017 IEEE International Conference on Robotics and Automation (ICRA), P3626, DOI 10.1109/ICRA.2017.7989417
[8]  
Basanez Munoz A, 2018, Serv. Robots, DOI [10.5772/intechopen.69874, DOI 10.5772/INTECHOPEN.69874, 10.5772/intechopen.76975]
[9]  
Dai JF, 2016, ADV NEUR IN, V29
[10]  
Deng J, 2009, PROC CVPR IEEE, P248, DOI 10.1109/CVPRW.2009.5206848