THE COMPARISON OF VEHICLE SPEED ACCURACY USING VIDEO BASED MIXTURE OF GAUSSIAN 2 METHOD AND K-NEAREST NEIGHBOR METHOD

Abstract

Background subtraction has various of methods including mixture of gaussian 2 and k-nearest neighbor. Some vehicle speed detection system in previous research have different accuracy in detecting speed of vehicle. The different of accuracy arose the idea to compare the accuracy in detecting the speed of vehicles using the mixture of gaussian 2 with k-nearest neighbor supplied by OpenCV 3.0 library. During this research, four video has been recorded with scenario ofvehicle speed 20 km/h, 40 km h, 50 km/h and 60 km/h. Each video has different vehicle’s speed. The program was designed and built using mixture of 2 gaussian and k-nearest neighbor. Vehicle speed parameter retrieval conducted by 10 tests of four videos,which are 20km.avi, 40km.avi, 50km.avi and 60km.avi.The speed detected at each videois compared with the actual speed to obtain such further information as percentage of error accuracy of the two methods. From the test results, it is obtained that percentage error of mixture of gaussian 2 method is 0,36%-23,73%of the actual speed. The percentage error of k-nearest neighbor method is 64.2%-58,85% of the actual speed.