Deep Interpretation of Parkland Environment for Autonomous Landscaping Robot for the Green Smart City

In the past decade, making the cities green and environmentally friendly is becoming a major issue for many countries around the world. The green city is an urban environment in which the parklands and natural habitats are integrated into the living and the working space of the communities. This significant increase in the scale of parklands in the green cities requires autonomous landscaping. In this paper, a computer vision system for an autonomous landscaping robot which is capable of seeding various types of grass and designing patterns in the lawns is developed. The proposed robotic platform uses deep convolutional neural networks for finding the required patches for the replanting of the grass and the obstacle avoidance for the robot. A dataset of real parkland environment is collected and the proposed vision system is evaluated for various scenarios. The experimental results show that the proposed vision system is capable of operating in complex parkland areas.