Early corn stand count of different cropping systems using UAV-imagery and deep learning

Optimum plant stand density and uniformity is vital in order to maximize corn (Zea mays L.) yield potential. Assessment of can occur shortly after seedlings begin emerge, allowing for timely replant decisions. The conventional methods evaluating an early rely on manual measurement visual observation, which are time consuming, subjective because the small sampling areas used, unable capture field-scale spatial variability. This study aimed evaluate feasibility unmanned aerial vehicle (UAV)-based imaging system estimating count three cropping systems (CS) with different tillage crop rotation practices. A UAV equipped on-board RGB camera was used collect imagery (~14 days planting) CS, i.e., minimum-till corn-soybean (MTCS), no-till (NTCS), corn-corn cover implementation (NTCC). An image processing workflow based a deep learning (DL) model, U-Net, developed segmentation estimation. Results showed that DL model performed best segmenting MTCS, followed by NTCS NTCC. Similarly, accuracy estimation highest MTCS (R2 = 0.95), (0.94) NTCC (0.92). Differences CS were related amount distribution soil surface residue cover, increasing generally reducing performance proposed method Thus, using modeling qualified influenced management