Perbandingan Metode Haar Cascade, YoloV3, dan TinyYoloV3 Dalam Mendeteksi Kendaraan Bermotor Berbasis Video

Marzuarman (Politeknik Negeri Bengkalis)
Stephan (Politeknik Negeri Bengkalis)
Muharnis (Politeknik Negeri Bengkalis)
Azizul (Politeknik Negeri Bengkalis)
Doni Mirza Rinaldi (Politeknik Negeri Bengkalis)
Bagas Prasetyo (Politeknik Negeri Bengkalis)

Proceeding Applied Business and Engineering Conference
Vol. 10 (2022): 10th Applied Business and Engineering Conference

Motorized vehicles are one of the most important needs in everyday life. Every year in Indonesia there is always an increase in the number of motorized vehicles along with the increase in population. Many researchers in the field of information technology use image processing systems to investigate and develop systems that can be used on public roads, one of which is to detect motor vehicles. In general, the methods that are often used to detect objects are Haar cascade, YOLOV3, and TinyYOLOV3. In this study, a comparison was made, to determine the best accuracy of the three methods in detecting motorized vehicles. The test was carried out using Python 3.10 software that has been installed with OpenCV, where the test was carried out using a video with a duration of 1 minute 23 seconds which was downloaded from the youtube.com site. Based on the test results, the YOLOV3 method gets the best level of accuracy, which is 74%. For the Haar cascade method, the accuracy value is 41%, and TinyYOLOV3 produces an accuracy rate of 25%.