Traffic Regulator Detection Using GPS Trajectories

Abstract

This paper explores the idea of enriching maps with features predicted from GPS trajectories. More specifically, it proposes a method of classifying street intersections according to traffic regulators (traffic light, yield/priority-sign and right-of-way rule). Intersections are regulated locations and the observable movement of vehicles is affected by the underlying traffic rules. Movement patterns such as stop events or start-and-stop sequences are commonly observed at those locations due to traffic regulations. In this work, we test the idea of detecting traffic regulators by learning them in a supervised way from features derived from GPS trajectories. We explore and assess different settings of the feature vector being used to train a classifier that categorizes the intersections based on traffic regulators; also, we test several experimental setups. The results show that a Random Forest classifier with oversampling and Bagging booster enabled can predict the intersection regulators with 90.4% accuracy. We discuss future research directions and recommend next steps for improving the results of this research.

Impact Assessing of Traffic Lights via GPS Vehicle Trajectories

“…In addition, they calculated the percentage of cross-sectional measurements for each road segment containing at least one stop. Recently, Golze et al [10] used different sampling settings of the same nature: random oversampling [11], SMOTE (Synthetic Minority Oversampling TEchnique) [12] and ADASYN (ADAptive SYNthetic sampling approach) [13]. They investigated the effect of learning traffic rules (traffic lights, yield rules) on physical feature vectors obtained from trajectory data and compared the classification accuracy of decision trees, random forests, support vector machines, and neural network algorithms to initially explore the difference between machine learning and deep learning.…”

Traffic Control Recognition with Speed-Profiles: A Deep Learning Approach

“…In this paper, we propose a sequence-to-sequence deep learning model using GPS tracks as a time sequence for realizing the goal mentioned above. The majority of previous works suggests using the statistical features extracted from GPS tracks, such as stop times and stop duration [5,25,26], slowdown and standstill events [27][28][29] or speed-profiles [30][31][32]. These types of features summarize the dynamics of vehicles' motion in the relevant junctions over time.…”

“…They recognize locations where pedestrians stay over a time threshold (dwelling time) and categorize the regulators to the two categories accordingly. Similar to the study [5], the most recent work by Golze et al [26] uses the speed-related statistics (e.g., mean and maximum crossing speed) extracted from GPS tracks for traffic regulator classification.…”

“…The proposed CVAE model was compared with the performance of a random forest model [26] using the same dataset (1328 in total regulators) as in [26], enlarged by additional 1609 regulators from the same city/road-network (in total 2937 regulators). Different from the sequential features mentioned above, the random forest model uses two types of features summarized from the GPS tracks: physical and statistical features.…”

“…The statistical features are the statistics, such as minimum, maximum, mean and variance of all the aforementioned physical features. Different strategies were leveraged to boost the performance of the random forest model, such as, random oversampling and bagging or AdaBoost [26].…”