Research on Recognition of Working Area and Road Garbage for Road Sweeper Based on Mask R-CNN Neural Network

Abstract

In order to reduce the energy waste of the road sweeper and simplify the operation process of the driver, a set of intelligent cleaning device for road sweeper is designed, which can automatically identify road garbage and adjust the cleaning mechanism to the required power. This device is composed by a monocular camera, an industrial computer, a vehicle DC power, and a cleaning mechanism. In terms of algorithms, two Mask R-CNN neural network models are used to detect road garbage. First, the road surface information is obtained by the first model to obtain the workable area of the road sweeper, which can reduce the influence of factors such as vehicles and pedestrians. Secondly, the road surface information in the workable area is divided into two types, road-specific information and garbage, the garbage detection and marking are completed after the second model is tested. Finally, the garbage coverage rate is used as a feature to adjust the power of the cleaning device. The result of testing and analysis of this algorithms shows that the real-time performance and recognition accuracy can achieve the expected results.