Abstract
Exact yield estimation of fruits on plants guaranteed fine and timely decisions on harvesting and marketing practices. Automatic yield estimation based on unmanned agriculture offers a viable solution for large orchards. Recent years have witnessed notable progress in computer vision with deep learning for yield estimation. Yet, the current practice of vision-based yield estimation with successive frames may engender fairly great error because of the double counting of repeat fruits in different images. The goal of this study is to provide a wise framework for fruit yield estimation in sequence images. Specifically, the anchor-free detection architecture (CenterNet) is utilized to detect fruits in sequence images from videos collected in the apple orchard and orange orchard. In order to avoid double counts of a single fruit between different images in an image sequence, the patch matching model is designed with the Kuhn–Munkres algorithm to optimize the paring process of repeat fruits in a one-to-one assignment manner for the sound performance of fruit yield estimation. Experimental results show that the CenterNet model can successfully detect fruits, including apples and oranges, in sequence images and achieved a mean Average Precision (mAP) of 0.939 under an IoU of 0.5. The designed patch matching model obtained an F1-Score of 0.816 and 0.864 for both apples and oranges with good accuracy, precision, and recall, which outperforms the performance of the reference method. The proposed pipeline for the fruit yield estimation in the test image sequences agreed well with the ground truth, resulting in a squared correlation coefficient of R2apple = 0.9737 and R2orange = 0.9562, with a low Root Mean Square Error (RMSE) for these two varieties of fruit.
Highlights
Smart farming is becoming increasingly pervasive in modern agriculture, from crop planting with commonly automatic equipment in the fields to the current trend of intelligent monitoring for fruit tree growing in the orchard, providing effective management tools to support precise cultivating [1]
To train models and evaluate the performance of the proposed method for fruit yield estimation, all experiments are implemented on a workstation platform containing an NVIDIA (R) TITAN Xp GPU with 16 GB of graphics memory, an Intel(R) Core i7 7700 CPU
Compute unified device architecture (CUDA) toolkit 10.0 and CUDA deep neural network v7.5 are both applied to faster graphic calculation and less memory access latency
Summary
Smart farming is becoming increasingly pervasive in modern agriculture, from crop planting with commonly automatic equipment in the fields to the current trend of intelligent monitoring for fruit tree growing in the orchard, providing effective management tools to support precise cultivating [1]. The conventional approach for estimating yield primarily relies on humans, which is sampling a fixed percentage (e.g., 5% or 10%) of trees randomly and fruit counting before extrapolating the total yield of the entire orchard [5]. This sampling and extrapolation practice for long hours is labor intensive and time consuming, and prone to the error caused by brain fatigue or other interference. Few critical reviews concerning fruit tasks in orchards are reported by Gongal et al [16] and Koirala et al [17], pointing out that machine learning yields better results than traditional image processing techniques
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
