In this manuscript, the real-time detection of power lines and lanes in agricultural fields is discussed. This proposed method will make a significant improvement in smart agriculture. Real-time scene parsing through object identification in embedded systems is extremely challenging due to processing resource constraints the accuracy is less. The detection is performed using the YOLO v5 (You Only Look Once) algorithm based on deep learning. The YOLO v5 algorithm and the deep neural network algorithm are trained by feeding image datasets of power lines and different types of lanes in agricultural fields. Transmission lines and lanes are detected in real-time using a system consisting of a Raspberry Pi 3, a Pi camera, and a single-channel relay that connects to a DC motor. The proposed deep learning approach for power line and lane detection can increase the accuracy to 99.3% and improve the robustness of the system compared to traditional methods.

YOLO v5, Deep neural network, Object identification, Smart agriculture, Raspberry Pi.