Development of a Real-Time Plant Species Recognizing Rover

Abstract

Real-time plant species recognition in unconstrained backgrounds such as scale changes, illumination variations, orientation or viewpoint variations, cluttered backgrounds and complex structure of leaf (simple or compound), is a challenging and time-consuming process. In this paper, a land rover incorporated with plant species recognition system (deployed in Raspberry Pi board) is developed to perform real-time recognition of plant species. Plant species recognition is carried out using a custom-created Convolutional Neural Network (CNN) Model (Two Convolutional Blocks and Two Fully connected blocks). The leaf-6 dataset is custom created with six plant species namely, Aloe vera, Murraya koenigii, Lawsonia inermis, Ocimum tenuiflorum, Moringa oleifera and Sesbania grandiflora. The proposed CNN model achieved an accuracy of about 93 % (Training) and 91.87 % (Testing), on the Leaf-6 dataset. The rover is remotely controlled using an RF Transceiver. The rover's onboard data such as temperature, humidity and plant species prediction results are transferred to an authenticated user's mobile phone.