Data Mining Practical Machine Learning Tools and Techniques

Highway Capacity Manual改訂の動向--テイラ-教授の講演より

Pattern recognition

The growing explosion in the use of surveillance cameras in public security highlights the importance of vehicle search from a large-scale image or video database. However, compared with person re-identification or face recognition, vehicle search problem has long been neglected by researchers in vision community. This paper focuses on an interesting but challenging problem, vehicle re-identification (a.k.a precise vehicle search). We propose a Deep Relative Distance Learning (DRDL) method which exploits a two-branch deep convolutional network to project raw vehicle images into an Euclidean space where distance can be directly used to measure the similarity of arbitrary two vehicles. To further facilitate the future research on this problem, we also present a carefully-organized largescale image database "VehicleID", which includes multiple images of the same vehicle captured by different realworld cameras in a city. We evaluate our DRDL method on our VehicleID dataset and another recently-released vehicle model classification dataset "CompCars" in three sets of experiments: vehicle re-identification, vehicle model verification and vehicle retrieval. Experimental results show that our method can achieve promising results and outperforms several state-of-the-art approaches.

/pdf/deep-relative-distance-learning-tell-the-difference-between-4ut5ydyh1o.pdf

Deep Relative Distance Learning: Tell the Difference between Similar Vehicles

https://www.tandfonline.com/doi/pdf/10.1080/15472450.2017.1291351

Policy and society related implications of automated driving: A review of literature and directions for future research

A lane changing assistance system that advises drivers of safe gaps for making mandatory lane changes at lane drops is developed. Bayes classifier and decision-tree methods were applied to model lane changes. Detailed vehicle trajectory data from the Next Generation Simulation (NGSIM) data set were used for model development (U.S. Highway 101) and testing (Interstate 80). The model predicts driver decisions on whether to merge or not as a function of certain input variables. The best results were obtained when both Bayes and decision-tree classifiers were combined into a single classifier using a majority voting principle. The prediction accuracy was 94.3% for nonmerge events and 79.3% for merge events. In a lane change assistance system, the accuracy of nonmerge events is more critical than merge events. Misclassifying a nonmerge event as a merge event could result in a traffic crash, whereas misclassifying a merge event as a nonmerge event would only result in a lost opportunity to merge. Sensitivity analysis performed by assigning higher misclassification cost for nonmerge events resulted in even higher accuracy for nonmerge events but lower accuracy for merge events.

Modeling Mandatory Lane Changing Using Bayes Classifier and Decision Trees

The vehicle reidentification problem is the task of matching a vehicle detected at one location with the same vehicle detected at another location from a feasible set of candidate vehicles detected at the other location. This paper formulates and solves the vehicle reidentification problem as a lexicographic optimization problem. Lexicographic optimization is a preemptive multi-objective formulation, and this lexicographic optimization formulation combines lexicographic goal programming, classification, and Bayesian analysis techniques. The solution of the vehicle reidentification problem has the potential to yield reliable section measures such as travel times and densities, and enables the measurement of partial dynamic origin/destination demands. Implementation of this approach using conventional surveillance infrastructure permits the development of new algorithms for ATMIS (Advanced Transportation Management and Information Systems). Freeway inductive loop data from SR-24 in Lafayette, California, demonstrates that robust results can be obtained under different traffic flow conditions.

Use of vehicle signature analysis and lexicographic optimization for vehicle reidentification on freeways

None of numerous existing traffic flow forecasting models focus on work zones. Work zone events create conditions that are different from both normal operating conditions and incident conditions. In this paper, four models were developed for forecasting traffic flow for planned work zone events. The four models are random forest, regression tree, multilayer feedforward neural network, and nonparametric regression. Both long-term and short-term traffic flow forecasting applications were investigated. Long-term forecast involves forecasting 24 h in advance using historical traffic data, and short-term forecasts involves forecasting 1 h and 45, 30, and 15 min in advance using real-time temporal and spatial traffic data. Models were evaluated using data from work zone events on two types of roadways, a freeway, i.e., I-270, and a signalized arterial, i.e., MO-141, in St. Louis, MO, USA. The results showed that the random forest model yielded the most accurate long-term and short-term work zone traffic flow forecasts. For freeway data, the most influential variables were the latest interval's look-back traffic flows at the upstream, downstream, and current locations. For arterial data, the most influential variables were the traffic flows from the three look-back intervals at the current location only.

Traffic Flow Forecasting for Urban Work Zones

Lane change assistance system was developed using ensemble learning methods.Proposed system has potential to prevent lane change crashes and thus reducing injuries and fatalities.Random forest and AdaBoost outperformed Bayes/Decision tree classifier.Higher classification accuracy and lower false positive rates achieved.Accuracies of 99.1% and 98.7% were achieved for lane keeping. Lane change maneuvers contribute to a significant number of road traffic accidents. Advanced driver assistance systems (ADAS) that can assess a traffic situation and warn drivers of unsafe lane changes can offer additional safety and convenience. In addition, ADAS can be extended for use in automatic lane changing in driverless vehicles. This paper investigated two ensemble learning methods, random forest, and AdaBoost, for developing a lane change assistance system. The focus on increasing the accuracy of safety critical lane change events has a significant impact on lowering the occurrence of crashes. This is the first study to explore ensemble learning methods for modeling lane changes using a comprehensive set of variables. Detailed vehicle trajectory data from the Next Generation Simulation (NGSIM) dataset in the US were used for model development and testing. The results showed that both ensemble learning methods produced higher classification accuracy and lower false positive rates than the Bayes/Decision tree classifier used in the literature. The impact of misclassification of lane changing events was also studied. A sensitivity analysis performed by varying the accuracy of lane changing showed that the lane keeping accuracy can be increased to as high as 99.1% for the AdaBoost system and 98.7% for the random forest system. The corresponding true positive rates were 96.3% and 94.6%. High accuracy of lane keeping and high true positive rates are desirable due to their safety implications.

Situation assessment and decision making for lane change assistance using ensemble learning methods

Travel time is the reciprocal of speed and is a useful measure of road congestion and traffic system performance. Travel time is also a basic traffic variable that is used in many intelligent transportation system strategies such as route guidance, incident detection, and traveler information systems. Previously, speeds were mainly acquired from double inductive loops configured as speed traps, because single loop speed estimates based on assumptions of a constant vehicle length were inaccurate. However, more accurate measurements of speed can now be accomplished with single loops by utilizing inductive waveforms of vehicles that are output from newer detector cards. An algorithm using signal processing and statistical methods was developed to extract speeds from inductive waveforms. The results show that the proposed algorithm performs better than conventional single loop estimation methods. The results also show that the algorithm is robust under different traffic conditions and is transferrable across surveillance sites without the need for recalibration. The use of the extensive single loop surveillance infrastructure is a cost-effective way of obtaining more accurate networkwide travel time information.

/pdf/individual-vehicle-speed-estimation-using-single-loop-28qgldvov0.pdf

Carlos Sun

Papers

Modeling Mandatory Lane Changing Using Bayes Classifier and Decision Trees

Use of vehicle signature analysis and lexicographic optimization for vehicle reidentification on freeways

Traffic Flow Forecasting for Urban Work Zones

Situation assessment and decision making for lane change assistance using ensemble learning methods

Individual Vehicle Speed Estimation Using Single Loop Inductive Waveforms