TY - GEN
T1 - A practical approach for counting and classifying vehicles using rising/falling edge thresholding in a virtual detection line
AU - Rosas-Arias, L.
AU - Portillo-Portillo, J.
AU - Sanchez-Perez, G.
AU - Toscano-Medina, K.
AU - Perez-Meana, H. M.
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/7/2
Y1 - 2018/7/2
N2 - In this paper, we present a methodology for finding a set of useful parameters to identify traffic patterns. First, we perform motion detection in the scene by applying the background subtraction method. We compute the Euclidean distance of 3-dimensional pixels-vectors in the YCbCr color space and then apply thresholding to filter out small Euclidean distances that does not represent motion pixels. After that, we count the number of vehicles by implementing a detection line in one of the road lanes. We make sure that our system does not get false positive detections and counts the vehicles only when they completely pass through the detection line. Later, we determine the size of the vehicle by counting up all the pixels that passed through the detection line. According to the number of pixels, the vehicle is classified into two of the following categories: Big vehicle or Small vehicle. Finally, we determine the vehicle color by clustering the vehicle image into 7 different color categories. The color with the maximum number of occurrences in the image histogram is determined to be the vehicle color. Results show that our system reaches a 100% of accuracy when counting vehicles and determining their size. However, the color determination process tends to present problems when images of vehicles of two or more colors are presented, but obtaining good results overall.
AB - In this paper, we present a methodology for finding a set of useful parameters to identify traffic patterns. First, we perform motion detection in the scene by applying the background subtraction method. We compute the Euclidean distance of 3-dimensional pixels-vectors in the YCbCr color space and then apply thresholding to filter out small Euclidean distances that does not represent motion pixels. After that, we count the number of vehicles by implementing a detection line in one of the road lanes. We make sure that our system does not get false positive detections and counts the vehicles only when they completely pass through the detection line. Later, we determine the size of the vehicle by counting up all the pixels that passed through the detection line. According to the number of pixels, the vehicle is classified into two of the following categories: Big vehicle or Small vehicle. Finally, we determine the vehicle color by clustering the vehicle image into 7 different color categories. The color with the maximum number of occurrences in the image histogram is determined to be the vehicle color. Results show that our system reaches a 100% of accuracy when counting vehicles and determining their size. However, the color determination process tends to present problems when images of vehicles of two or more colors are presented, but obtaining good results overall.
UR - http://www.scopus.com/inward/record.url?scp=85063866298&partnerID=8YFLogxK
U2 - 10.1109/ROPEC.2018.8661458
DO - 10.1109/ROPEC.2018.8661458
M3 - Contribución a la conferencia
AN - SCOPUS:85063866298
T3 - 2018 IEEE International Autumn Meeting on Power, Electronics and Computing, ROPEC 2018
BT - 2018 IEEE International Autumn Meeting on Power, Electronics and Computing, ROPEC 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 IEEE International Autumn Meeting on Power, Electronics and Computing, ROPEC 2018
Y2 - 14 November 2018 through 16 November 2018
ER -