Classification and Counting of Moving Vehicle at Night with Similarity of Rear Lamp

Abstract

Abstract Congestion is caused by a large number of vehicles exceeding road capacity. If we know the amount of the average density of vehicles passing by, it can be a consideration of infrastructure development. Therefore, the authors made a study to classify and calculate vehicles at night using the similarity of vehicles rear lamp. The technique used by authors is to pair each vehicle’s rear lamps that have been detected which have the same characteristics, in this case, We used the similarity of pixels for the pairing process. After that, the pair of rear lamps will be calculated and classified as the type of motorbike or car. This study resulted in a calculation that in Video 1 with not-so-busy traffic conditions able to detect 79 of 88 motorbikes and 32 of 35 cars with accuracy 90,24%. Video 2 with fairly quiet conditions was able to detect 52 of 56 motorcycles and 9 of 11 cars with accuracy 91,04%. Video 3 with crowded traffic conditions can detect 63 of 71 motorcycles and 23 of 29 cars in actual conditions with accuracy 86,00%.