The use of electric cars as a means of transportation for pediatric patients has the main purpose of having a positive effect on the psychology of pediatric patients before surgery. Therefore, it is expected to accelerate the healing process. An electric car navigation system that can recognize the environment is needed. This article aims to develop a camera-based semi-autonomous navigation system using the faster R-CNN method to detect markers as electric car direction. This method optimizes the range of interest (RoI) layer to produce optimal features. Faster R-CNN is faster in generating accurate region proposals compared to R-CNN and Fast R-CNN. Various Faster R-CNN models were tested in image data processing for marker detection as the electric car steering system. Test results on FPS variations show that the best results were obtained when using the Faster R-CNN MobileNet V3 Large 320 FPN model with a value of 11.3f ps for the forward marker, 18.9 fps for the stop marker, 22.6 fps for the left turn marker and 11.1 fps for the right turn marker. With this model, the results obtained are quite good in testing the performance of the car navigation system. The results obtained in the success of the test are 70% for the forward marker test, 100% for the stop marker test, 90% for the left turn marker and 100% for the right turn marker.
Read full abstract