| Summary: | The number of vehicles increases every year in big cities around Malaysia, causing a higher flow of traffic on the road, prompting authorities to increase traffic monitoring to ensure smooth traffic conditions. Traffic surveillance normally conducted using cameras and observed manually by the authorities before they can make decisions on controlling the traffic flow. Continuous monitoring is tedious and prone to error especially during rush hour where traffic volume drastically increased. Intelligent traffic monitoring is possible via trajectory analysis and prediction where the artificial intelligent algorithms learns and cluster the trajectories of vehicles movement. Similarity measure based on distance-based such as Euclidean-distance, Dynamic Time Warping (DTW) and Longest Common Subsequence (LCSS) are less accurate and computationally costly. This paper proposed a combined modified DTW method which merged the piecewise and segmentation measurement in DTW, and the proposed method will be known as the Piecewise-Segment Dynamic Time Warping Distance (PSDTW). PSDTW algorithm is tested on CROSS dataset and compared the time taken and accuracy of the algorithm with previous studied algorithm. The proposed method improves the execution time by average of factor of 4 compared to the SDTW, DTW and LCSS with good results accuracy which have less than 0.01 error rate.
|
|---|