A Novel Target Recognition System for the Amphibious Robot based on Edge Computing and Neural Network

Abstract

In the past, the accuracy of target recognition was low and the transmission efficiency was low, in this paper, a target recognition system based on Edge calculation is proposed for the platform of Amphibious robot, which is mainly used for sea rescue and marine garbage search. Because of the flexible application of Amphibious robot to sea, land and air, the platform of Amphibious robot is equipped with camera to collect image information. We train neural networks on computers, We store the weight file on the edge calculation chip. The method of Edge calculation is used to run the Convolutional neural network to rescue the victims at sea, We have carried out many experiments to verify the accuracy of target recognition. Because the judgment of Edge nodes can greatly reduce the time of communication with terminals, thus improving the efficiency of rescue.