Traffic data analysis using image processing technique on Delhi-Gurgaon expressway

摘要

With the advancements in video image processing system (VIPS), detection mechanism has made a significant improvement over traditional methods for traffic data analysis. Traffic on Delhi-Gurgaon expressway is heterogeneous in nature with non-lane based behaviour. Moreover, automation and instrumentation are also not implemented. In view of this, TRaffic AnalyZer and EnumeratoR (TRAZER), a VIPS was used to process video-captured data on Delhi-Gurgaon expressway to check accuracy based on traffic count, speed and lateral placement. The motivation behind using TRAZER is to evaluate its efficiency and robustness for extracting micro and macro-level traffic parameters under heterogeneous traffic conditions. To achieve this, data were extracted manually on above parameters and compared with those obtained from TRAZER. The volume count data from TRAZER generated a lesser accuracy of 60% detection under heavy traffic conditions, using default parameters. Thus, refinements were carried out in the software as part of calibration: (i) redefining maximum and minimum detection widths for each vehicle category, and (ii) selecting the optimum trap length for reducing the occlusion effect, which increased the detection percentage as well as reduced the error. After implementing these refinements, 80% of the vehicles were detected. Further, relationships between vehicle speed and its lateral placement from median across road width, at a given point were also developed. The models were developed for both aggregate (considering all vehicles) and disaggregate (vehicle category-wise) levels. The polynomial relationship was found to be best fitted function to estimate vehicle speed based on its lateral placement.