In-Field Pine Seedling Counting Using End-to-End Deep Learning for Inventory Management

Abstract

Highlights A deep learning-based machine vision system was developed for pine seedling counting. Automated seedling counting achieved less error than the manual sampling-based practice. Regression-based counting from optical flow inputs achieved the best performance. Machine counting produces seedling density maps for management practice improvement. Abstract. The southern U.S. produces over 1 billion pine seedlings for market sales per year, with prices varying from 50 to 435 dollars per thousand seedlings. An accurate inventory of seedlings provides nursery management with insights into how many seedlings can be sold and/or if there is any loss due to washout, mechanical damage, or pest/diseases that can still be mitigated. In this study, we developed a system to count pine seedlings at production sites and map the seedling density in the field. A system with three cameras was developed to collect video from different drills in the seedling bed. The videos were preprocessed to restrict the region of interest to the center portion of the image in each camera and separated each drill into individual videos. Two different modalities, i.e., video and optical flow, were evaluated as inputs to a convolutional neural network followed by a long short-term recurrent network to model the sequence of frames and regress to the seedling count for each plot. The mean absolute percentage error (MAPE) of our best performing model was 7.5%, which is an improvement over the baseline manual sampling-based approach with a MAPE of 11%. The results showed that the proposed approach was able to count seedlings in a crowded scene under complex field conditions with higher accuracy than the standard manual practice. Therefore, the proposed system and results demonstrated the potential to replace manual counting and even provide further information such as a seedling density map over the field for precision forest nursery management and seedling harvesting. Keywords: CNN, LSTM, Nursery Inventory, Optical Flow, Pine Seedling, Regression.