Roadside LiDAR Research Articles

Automatic Background Construction and Object Detection Based on Roadside LiDAR

High-resolution micro traffic data are important to traffic safety and efficiency analysis. In this study, a roadside LiDAR sensor is used to collect 3D point clouds of surrounding objects. An automatic background construction and object detection method is proposed on the basis of the operation principle of the LiDAR sensor. In the algorithm, the discrete horizontal and vertical angular values can be considered as coordinates of pixels in digital images, and the farthest and mean distance of each azimuth are used to construct the background dataset. Then, vehicle and pedestrian points are extracted on the basis of the distance difference of each point with the same angular value between the target frame dataset and the background dataset. A density-based spatial clustering method is employed to group points into clusters and identify vehicles and pedestrians automatically. Finally, the performance of the algorithm is tested for roadside LiDAR data preprocessing under different traffic conditions. Our algorithm can perform well with high accuracy. Experimental results demonstrate that our algorithm can achieve a larger detection range and exhibit lower time complexity in comparison with a previously proposed algorithm.

An automatic skateboarder detection method with roadside LiDAR data

The increasing conflicts between skateboarders and pedestrians on the campuses have caused safety concerns. Although traffic planners on campus usually did not consider skateboarding into planning due to the lack of historical skateboarding data. The traditional data collection method such as manual counting or video detection took a lot of effort and could only provide macrolevel skateboarding data. Therefore, this research presented a new approach for skateboarder detection using the roadside LiDAR sensor. A two-stage classification method was developed to distinguish skateboarders and other road users. The first stage was to classify motor vehicles and nonvehicles (pedestrians, bicycles, and skateboarders). Then the second step was to distinguish skateboarders from pedestrians and bicycles. The proposed procedure was evaluated using real-world data on campus. The results showed that the proposed procedure can detect the skateboarder with the overall accuracy of 89.5%. The data collected in the real world showed that the speed of the skateboarder was usually higher than the pedestrian and lower than the bicycle. The skateboarding information extracted from the proposed detection method could be applied for skateboarder behavior analysis, volume counting, and safety analysis. A level-based procedure to reinforce the safety between skateboarders and pedestrians was recommended.

Revolution and rotation-based method for roadside LiDAR data integration

Recently, roadside LiDAR has been applied to different transportation areas. Data integration (point registration) for multiple LiDAR sensors is an important step to extend the data range and improve the density of scanned points. A challenge for the three-dimensional LiDAR is to find the corresponding features in different sensors. It is even unrealistic to identify the correspondences due to the different installation heights of LiDAR sensors. This paper proposes a generalized approach to integrate point clouds automatically based on a revolution and rotation-based method. The proposed method matches the ground points with an optimizing algorithm. The new approach was validated by the data collected in the real world. The testing results showed that the length error and width error of the proposed method were 13.26% and 13.15%, respectively. Compared to the state-of-the-art method, the proposed method has a higher level of automation and improved accuracy.

Points Registration for Roadside LiDAR Sensors

Roadside LiDAR deployment provides a solution to obtain the real-time high-resolution micro traffic data of unconnected road users for the connected-vehicle road network. Single roadside LiDAR sensor has a lot of limitations considering the scant coverage and the difficulty of handling object occlusion issue. Multiple roadside LiDAR sensors can provide a larger coverage and eliminate the object occlusion issue. To combine different LiDAR sensors, it is necessary to integrate the point clouds into the same coordinate system. The existing points registration methods serving mapping scans or autonomous sensing systems could not be directly used for roadside LiDAR sensors considering the different feature of point clouds and the spare points in the cost-effective roadside LiDAR sensors. This paper developed an approach for roadside LiDAR points registration. The developed points-aggregation-based partial iterative closest point algorithm (PA-PICP) is a semi-automatic points registration method, which contains two major parts: XY data registration and Z adjustment. A semi-automatic key point selection method was introduced. The partial iterative closest point was applied to minimize the difference between different LiDARs in the XY plane. The intersection of ground surface between different LiDARs was used for Z-axis adjustment. The performance of the developed procedure was evaluated with field-collected LiDAR data. The results showed the effectiveness and accuracy of data integration using PA-PICP was greatly improved compared with points registration using the traditional iterative closest point. The case studies also showed that the occlusion issue can be fixed after PA-PICP points registration.

Automatic Vehicle Classification using Roadside LiDAR Data

This research presented a new approach for vehicle classification using roadside LiDAR sensor. Six features (one feature, object height profile, contains 10 sub-features) extracted from the vehicle trajectories were applied to distinguish different classes of vehicles. The vehicle classification aims to assign the objects into ten different types defined by FHWA. A database containing 1,056 manually marked samples and their corresponding pictures was provided for analysis. Those samples were collected at different scenarios (roads and intersections, different speed limits, day and night, different distance to LiDAR, etc.). Naïve Bayes, K-nearest neighbor classification, random forest (RF), and support vector machine were applied for vehicle classification. The results showed that the performance of different methods varied by class. RF has the highest overall accuracy among those investigated methods. Some types were merged together to serve different types of users, which can also improve the accuracy of vehicle classification. The validation indicated that the distance between the object and the roadside LiDAR can influence the accuracy. This research also provided the distribution of the overall accuracy of RF along the distance to LiDAR. For the VLP-16 LiDAR, to achieve an accuracy of 91.98%, the distance between the object and LiDAR should be less than 30 ft. Users can set up the location of the roadside LiDAR based on their own requirements of the classification accuracy.

Vehicle Detection and Tracking in Complex Traffic Circumstances with Roadside LiDAR

The problem of traffic safety has become increasingly prominent owing to the increase in the number of cars. Traffic accidents often occur in an instant, which makes it necessary to obtain traffic data with high resolution. High-resolution micro traffic data (HRMTD) indicates that the spatial resolution reaches the centimeter level and that the temporal resolution reaches the millisecond level. The position, direction, speed, and acceleration of objects on the road can be extracted with HRMTD. In this paper, a LiDAR sensor was installed at the roadside for data collection. An adjacent-frame fusion method for vehicle detection and tracking in complex traffic circumstances is presented. Compared with the previous research, objects can be detected and tracked without object model extraction or a bounding box description. In addition, problems caused by occlusion can be improved using adjacent frames fusion in the vehicle detection and tracking algorithms in this paper. The data processing procedure are as follows: selection of area of interest, ground point removal, vehicle clustering, and vehicle tracking. The algorithm has been tested at different sites (in Reno and Suzhou), and the results demonstrate that the algorithm can perform well in both simple and complex application scenarios.

Automatic ground points filtering of roadside LiDAR data using a channel-based filtering algorithm

Ground points information is valuable for various transportation applications. To date, limited studies have been applied for ground points identification with the roadside LiDAR sensor. This paper presents an innovative approach to extract the ground points for the roadside LiDAR effectively. The proposed method can be divided into four major parts: region of intersection (ROI) Selection, background filtering, channel-based clustering, and slope-based filtering. The channel information is used for the ground points extraction. Data collected at different scenarios were used to evaluate the performance of the proposed method. It was shown that the algorithm can successfully extract ground points regardless of slope on the road and package loss issue. The developed method can also accommodate the different types of the LiDAR. Compared to the state-of-the-art methods, the overall performance of the proposed method is superior.

Trajectory tracking and prediction of pedestrian's crossing intention using roadside LiDAR

Trajectory tracking and crossing intention prediction of pedestrians at intersections are important to intersection safety. Recently, on-board video sensors have been developed for detection of pedestrians. However, both the detection range and operating environment of video-based systems seem to be constrained by the advancement of image-processing technologies. Additionally, on-board systems cannot alarm pedestrians to take evasive actions when at risk, a feature which is critical to saving lives. This paper summarises the authors' practice on using roadside LiDAR sensors to monitor and predict pedestrians' crossing intention, as part of an ongoing effort to develop a pioneering LiDAR-based system to systematically reduce pedestrian and vehicle collisions at intersections. The LiDAR sensors were installed at intersections to collect pedestrian data such as presence, location, velocity, and direction. A new method based on deep autoencoder - artificial neural network (DA-ANN) was used to process data and predict pedestrian crossing intention. The case study shows the proposed model is about 95% prediction accuracy and computational efficiency for real-time systems. The roadside LiDAR system has great potential to significantly reduce vehicle-to-pedestrian crashes both at intersections and non-intersection areas, either used as a stand-alone system or in conjunction with the connected V2I and I2V technologies.

LiDAR-Enhanced Connected Infrastructures Sensing and Broadcasting High-Resolution Traffic Information Serving Smart Cities

Connected-vehicle system is an important component of smart cities. The complete benefits of connected-vehicle technologies need the real-time information of all vehicles and other road users. However, the existing connected-vehicle deployments obtain the real-time status of connected vehicles, but without knowing the unconnected traffic since there are still many unconnected vehicles and pedestrians on the roads. Therefore, it is urgent to find an approach to collect the high-resolution real-time status of unconnected road users. When it is difficult for all vehicles, pedestrians, and bicyclists to broadcast their real-time status in the near future, enhancing the traffic infrastructures to actively sense and broadcast each road user's status is an intuitive solution to fill the data gap. This paper introduces a new-generation LiDAR-enhanced connected infrastructures that can actively sense the high-resolution status of surrounding traffic participants with roadside LiDAR sensors and broadcast connected-vehicle messages through DSRC roadside units. The system architecture, the LiDAR data processing procedure, the data communication, and the first pilot implementation at an intersection in Reno, Nevada are included in this paper. This research is the start of the new-generation connected infrastructures serving connected/autonomous vehicles with the roadside LiDAR sensors. It will accelerate the deployment of the connected network for the smart cities to improve traffic safety, mobility, and fuel efficiency.

Automatic Background Filtering Method for Roadside LiDAR Data

The high-resolution micro traffic data (HRMTD) of all roadway users is important for serving the connected-vehicle system in mixed traffic situations. The roadside LiDAR sensor gives a solution to providing HRMTD from real-time 3D point clouds of its scanned objects. Background filtering is the preprocessing step to obtain the HRMTD of different roadway users from roadside LiDAR data. It can significantly reduce the data processing time and improve the vehicle/pedestrian identification accuracy. An algorithm is proposed in this paper, based on the spatial distribution of laser points, which filters both static and moving background efficiently. Various thresholds of point density are applied in this algorithm to exclude background at different distances from the roadside sensor. The case study shows that the algorithm can filter background LiDAR points in different situations (different road geometries, different traffic demands, day/night time, different speed limits). Vehicle and pedestrian shape can be retained well after background filtering. The low computational load guarantees this method can be applied for real-time data processing such as vehicle monitoring and pedestrian tracking.

Driving cycles for measuring passenger car emissions on roads with traffic calming measures

Although local authorities in the UK need to be aware of any air quality impacts resulting from their traffic calming operations, there is little information relating to the effects of different traffic calming measures. The effects on air quality on this scale are complex, and so TRL is providing guidance by developing performance indices for different measures based on their effects on vehicle emissions. The emissions indices for passenger cars are based on tests conducted on a chassis dynamometer, and this paper describes the development of the methodology for constructing the driving cycles to be used. The technique involves the measurement of the speed profiles of a large number of vehicles using a roadside LIDAR system, and the determination of typical gear selections using three-instrumented cars.