Seed Kernel Counting using Domain Randomization and Object Tracking Neural Networks

TL;DR

This work tackles the cost barrier of seed kernel counting by combining domain-randomization–generated synthetic data with YOLOv8 object detection and two tracking frameworks (StrongSORT and ByteTrack) to count seeds in video. The hardware setup is deliberately low-cost (mechanical hopper, backlit surface, mobile camera) and DR is used to create synthetic datasets for soy and wheat seeds, with training on synthetic data and validation on real images. Detection performance on real data is high (precision and recall in the high 90s for soy and wheat, with AP50 around the low 90s), and counting accuracy improves with higher frame rates, achieving seed-counting accuracy in the mid-to-high 90s for favorable conditions but undercounting in clustered scenes due to occlusion and tracking ID fragmentation. The approach demonstrates the feasibility of DR-trained detectors plus tracking for affordable seed counting, with practical impact for the seed packaging industry and potential mobile app deployment for end users.

Abstract

High-throughput phenotyping (HTP) of seeds, also known as seed phenotyping, is the comprehensive assessment of complex seed traits such as growth, development, tolerance, resistance, ecology, yield, and the measurement of parameters that form more complex traits. One of the key aspects of seed phenotyping is cereal yield estimation that the seed production industry relies upon to conduct their business. While mechanized seed kernel counters are available in the market currently, they are often priced high and sometimes outside the range of small scale seed production firms' affordability. The development of object tracking neural network models such as You Only Look Once (YOLO) enables computer scientists to design algorithms that can estimate cereal yield inexpensively. The key bottleneck with neural network models is that they require a plethora of labelled training data before they can be put to task. We demonstrate that the use of synthetic imagery serves as a feasible substitute to train neural networks for object tracking that includes the tasks of object classification and detection. Furthermore, we propose a seed kernel counter that uses a low-cost mechanical hopper, trained YOLOv8 neural network model, and object tracking algorithms on StrongSORT and ByteTrack to estimate cereal yield from videos. The experiment yields a seed kernel count with an accuracy of 95.2\% and 93.2\% for Soy and Wheat respectively using the StrongSORT algorithm, and an accuray of 96.8\% and 92.4\% for Soy and Wheat respectively using the ByteTrack algorithm.

Figures (4)

• Figure 1: Mechanical hopper delivering seed kernels
• Figure 2: Wheat seed kernels flowing down the light box
• Figure 3: Image capture of soy seed kernels
• Figure 4: Sample images of (a) soy (b) wheat from the synthetic image dataset