The paper presents the results of the study of a plant detection program on agriculture and forestry surveillance quadcopter companion computers. The plant detection program uses an optimized convolution neural network to process the drone camera input video frame by frame and can process up to 38 FPS on the companion computer. The inference speeds up efficiently compared to the original SSD Mobilenet Lite V2 reach approximately 304 times. This performance is satisfied by most real-time applications for agriculture and forestry monitoring flight missions. The network was integrated on a NDIVIA Jetson Nano embedded computer and succeeded in detecting “coconut tree” in different simulation scenarios of a drone flight in real-time. The results demonstrate that the proposed approach could be used for further development of a fully plant detection system using only cameras. They also showed that a good outcome is achievable needing only cheap devices and can be implemented easily on forestry monitoring drones or agricultural drones which are familiar nowadays in Vietnam.