A Smart Leaf Blow Robot Based on Deep Learning Model

Abstract

Although leaves are everywhere in the world, and they also play a vital role in our daily life, they tend to fall all over the ground in due course, thereby making it difficult for pedestrians and vehicles to move. In this paper, an automatic leaf blower was designed and based on the concept of convolutional neural network(CNN). This system can automatically collect leaves into a garbage bag. A four-wheel driving robot was implemented to drive a blow machine. The control sensors of this robot mainly include a camera, ultrasound, the electronic compass and acceleration. Besides, an ultra-wide band located module was used to obtain the position of the current robot during the working process. Also, the computer vision was employed to recognize whether the leaves are on the ground. For this, ResNet50 deep CNN was used as the training model to recognize the fallen leaves. Since there are many kinds of trees, their leaves are different shape. We collected the images of these leaves as dataset for training, and the recognition rate achieved 92.5%. The obtained result was sent to the controller to control the moving direction of the robot. For the real-time operation, the embedded system was used to sense the leaf data to decide the movement made by the machine based on a control algorithm. The CNN model was implemented with an accelerator on the embedded system for the real-time purpose, which the recognition speed can achieve 20 frames per second form the camera. The automatic leaf blow machine can be possibly used in an effective way instead of human power.

Keywords

• Leaf recognition
• robot
• blow machine
• recognition
• convolutional neural networks (CNNs)