- About this Journal ·
- Abstracting and Indexing ·
- Aims and Scope ·
- Annual Issues ·
- Article Processing Charges ·
- Author Guidelines ·
- Citations to this Journal ·
- Contact Information ·
- Editorial Board ·
- Editorial Workflow ·
- Free eTOC Alerts ·
- Publication Ethics ·
- Reviewers Acknowledgment ·
- Submit a Manuscript ·
- Table of Contents
Advances in Mechanical Engineering
Volume 2013 (2013), Article ID 790813, 7 pages
Study on the Influence of Driving Experience on Visual Characteristic
College of Transportation, Jilin University, Changchun 130022, China
Received 9 September 2013; Revised 15 October 2013; Accepted 3 November 2013
Academic Editor: Wuhong Wang
Copyright © 2013 Shi-wu Li et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
In order to study the influence of driving experience and traffic flow conditions on driver's visual search mode, an experiment platform was built. Driver's eye movement data was collected through a large number of real vehicle tests. By analyzing the visual features variations of different experience drivers at peak and peaceful peak period, we found out the change laws of different experience drivers in their driving process. To study different drivers’ fixation areas in different traffic flow conditions, the clustering method of -means was selected to cluster plane coordinates of drivers’ fixation point with driver's interesting regions in the process of driving, and driver's vision plane was divided into eight regions finally. The results of the study showed that the clustering method of -means could divide driver's vision plane more accurate, and driving experience and traffic flow conditions could influence driver's visual search mode obviously.
When driving on urban road, about 80%–90% safe driving information of drivers was from eyes, and 95% information was dynamic [1, 2]. Numerous studies showed that safe driving was most closely related to the drivers’ dynamic visual features. The drivers’ dynamic visual features are mainly fixation features, including fixation time, fixation area, and fixation direction. So drivers’ dynamic visual feature was an important part of studying the drivers’ behavioral characteristics. Studies on the field of driver’s visual behavior were quite broad on a broad, specific researches including visual search pattern and search strategy comparison between skilled drivers and unskilled drivers, the relationship between visual behavior and vehicle control, work load effects on driver’s visual features, and so forth [3–5].
Underwood et al. from England Nottingham University studied the reasons that visual search range of nonskilled drivers is less than that of skilled drivers in horizon. At the same time, they researched the visual search characters of the nonskilled drivers and skilled drivers through watching a scene video obtained from real traffic situation in the laboratory . By introducing different kinds of streets into different rode, Crundall and Underwood studied the differences of visual search strategy between nonskilled drivers and skilled drivers . Underwood et al. had studied the scan order (scan path) of non-skilled drivers and skilled drivers driving in the country road, suburban road, and city two-way street. Tania Dukic from Sweden Chalmers Technology University, L. Hanson from Lund University, and Torbjörn Falkmer from Linköping University researched the visual behavior characters and handle characters when manipulating vehicles . Falkmer and Gregersen researched the difference of visual search strategies between the non-skilled drivers and skilled drivers using eye tracking system .
Different traffic conditions could bring different influence for different experience drivers and also cause driver’s fixation objective, interesting fixation area different. So, it was necessary to study the visual features of different experience drivers in various traffic environments.
2. The Construction of Experiment System
In order to collect driver’s visual parameters, an experiment system was constructed based on eye movement tracking device-Smart Eye, GPS, accident recorder equipment, and car camera. Smart Eye was used to collect driver’s visual features, such as fixation orbit, fixation range, fixation time, and interesting fixation point. GPS was used to collect vehicle speed and location. The GPS data can characterize traffic flow conditions. Car camera was used to record driver’s manipulation to vehicle. The experiment system was showed in Figure 1.
3. The Selection of Experimental Program
3.1. Experiment Route
In order to cover all types of urban roads, we selected the following route in Changchun: West gate of Jilin University-Renmin Street-Changbai Road-Yaitai Street-Nanhu Road-Yanan Street-Ziyou Road-Renmin Street-West gate of Jilin University. This route included intersections, roundabouts, urban ordinary roads, and urban expressways and passes railway station. Volume of traffic was large, and pedestrians and vehicles caused more horizontal interference. These were typical characteristics of urban roads.
According to the driving experience, three drivers were selected, marked with driver 1, driver 2, and driver 3, represented as drivers of skilled, generally skilled, and unskilled. The main information of drivers was showed in Table 1.
3.3. Experiment Method
Selecting testing time in peak and peaceful peak period, drivers drove the test car as their driving habits on experiment route (the drivers have been told before experiment). The eye movement data was collected by Smart Eye. These data not only included driver’s eye movement data at normal driving but also included driver’s eye movement data at traffic congestion and red light.
4. The Analysis of Fixation Time
4.1. Calculation of Fixation Time
We imported the data into data analysis software SPSS and selected the fixation data. According to fixed sample frequency of Smart Eye, we could calculate the total fixation time by (1) as follows: where was the total fixation time, was the number of fixation data, and was the sample frequency of Smart Eye.
The sample frequency of Smart Eye was 60 HZ; according to (1), we calculated three drivers total fixation time and mean fixation time. The results were shown in Table 2. Table 2 showed that the drivers’ fixation numbers and mean fixation time decreased with the increasing the driving experience. The drivers’ mean fixation time at peaceful peak period was longer than that at peak period, because traffic condition was good at peaceful peak period and the drivers can drive easily with little traffic information. So the drives’ eye movement speed was slow and mean fixation time was long.
4.2. The Analysis of Duration Fixation
Fixation behavior occurred frequently in an experiment; it was impossible to calculate every fixation time. So we cut apart the duration fixation time as the following subparagraphs: [0, 100 ms), [100, 200 ms), [200, 300 ms), [300, 400 ms), [400, 500 ms), [500, 600 ms), [600, 700 ms), [700, 800 ms), [800, 900 ms), [900 ms, +∞).
From Figures 2 and 3, we could conclude that the drivers’ duration fixation time mainly locates at the three subparagraphs of [0, 100 ms), [100, 200 ms), and [200, 300 ms) at peaceful peak period and at the two subparagraphs of [0, 100 ms) and [100, 200 ms) at peak period. It was because that the traffic condition was complicated at peak period. Drivers needed fast eye movement speed to collect more traffic information.
So the drivers’ duration fixation time was shorter at peak period than that at peaceful peak period. With the driving experience increasing, the changes of drivers’ duration fixation time became stable. So the influence of traffic flow changes on drivers decreased with the increasing the driving experience.
5. The Analysis of Fixation Area
To study different drivers’ fixation areas in different traffic flow conditions, -means clustering  was selected to cluster plane coordinates of drivers’ fixation point with driver’s interesting regions in the driving.
5.1. The Design of -Means Clustering
If were numbers in the space of , before the start of clustering, the number of was needed to be taken as the initial number of clustering. There were many methods for us to select the parameter of , such as selecting randomly or according to the number of sample. The basic steps of -means clustering were as follows [10–12].
Step 1. Select objects from numbers as the clustering center; other objects were respectively allocated to the best similar classes according to their cluster center similarity (distance). Equation (2) was the calculation of similarity (distance) as follows: where was the clustering center of category and was the distance between sample and clustering center . Euclidean distance, Manhattan distance, and Minkowski distance were the most used distance calculation methods. Consider(1)Euclidean distance(2)Manhattan distance(3)Minkowski distance
Step 2. Recalculate clustering center of every category that has been updated. If was the sample of category , is the clustering center of category . When is the th characteristic of clustering center , can be obtained from (6) as follows:
Step 3. Do Steps 2 and 3 repeatedly until the standard detection function become convergence; the updated cluster centers and cluster center before the update were consistent on the surface. Generally mean square deviation was used as the detection function of standard clustering, as following Equation (7):
Figure 5 showed the process of -means clustering analysis.
5.2. -Means Clustering of Gaze Point Coordinates
When using -means clustering to divide driver’s vision plane [13, 14], category numbers were the parts of dividing driver’s vision. So category numbers determined the parts of driver’s vision plane. How to select appropriate category numbers was the foundation and key of further research [15, 16]. If driver’s vision plane was divided into too many parts, the correspondence between distribution area of the fixation point and the fixation target was difficult to determine. On the contrary, the target characteristics in the distribution area of fixation point and the characteristics of information process could not be reflected fully.
Combining road conditions, vehicle arrangement and driver’s visual search features in driving, this paper used 6, 7, and 8 categories of -means clustering to cluster driver’s fixation point coordinate. Through comparing the advantages and disadvantages of the three clustering results, we selected the appropriate categories to divide driver’s vision plane. Table 3 showed the clustering result using the software SPSS.
5.3. The Division of Driver’s Vision Plane
As the fixation points collected through the experiment were space points, to divide driver’s vision plane conveniently, we firstly clustered driver’s fixation points and then did projection for these points to vertical plane. The division results of driver’s vision plane were shown in Figure 6.
Drivers could choose the traffic information which was helpful to safe driving; different drivers gaze at different areas when they drove. To research their interesting fixation area and the law of fixation area changes, We analyze three division results of driver’s vision plane according to driver’s interest area and view transfer characteristic in actual driving, was the best clustering results conforming to the actual situation; therefore, driver’s vision plane was divided into eight parts. Figure 7 showed the 8 parts of driver’s vision plane. They were respectively 1—front lane, 2—left lane, 3—right lane, 4—car dashboard, 5—right rearview mirror, 6—the outside of left lane, 7—left rearview mirror, and 8—the outside of right lane.
Drivers’ fixation time in the seven parts and percent of total time were shown in Table 4. From Table 4, the skilled drivers’ fixation time was longer than that of unskilled drivers at same time and route.
5.4. The Analysis of Fixation Time in Eight Parts
Figures 8 and 9 showed the percent of drivers’ fixation time in eight parts at peak and peaceful peak period. We concluded that drivers’ most interesting fixation area was front lane at both peak and peaceful peak periods. Front lane was the main area of getting traffic information for drivers. Skilled and generally skilled drivers’ second interesting fixation area was right lane, but that of unskilled drivers’ was left lane. Furthermore, skilled and generally skilled drivers could gaze at the other parts expect front lane; unskilled drivers nearly did not gaze at the other parts. So with the increase of driving experience, drivers’ fixation area got wider. It was beneficial for safe driving.
From above on, the following conclusions were obtained.(1)The drivers’ fixation numbers and mean fixation time decreased with the increasing of driving experience. The drivers’ mean fixation time at peaceful peak period was longer than that at peak period. Drivers’ duration fixation time of peaceful peak period mainly located at the three subparagraphs of [0, 100 ms), [100, 200 ms), and [200, 300 ms) and in the two subparagraphs of [0, 100 ms) and [100, 200 ms) at peak period. The drivers’ duration fixation time was shorter at peak period than that at peaceful peak period. With the driving experience increasing, the changes of drivers’ duration fixation time became stable.(2)-means clustering method was selected to cluster drivers’ fixation point coordinate. Combining road conditions, vehicle arrangement, and drivers’ visual search features in driving, driver’s vision plane division was divided into eight regions finally.(3)The most interesting fixation area of drivers was front lane at both peak and peaceful peak. Heading lane was the main area of getting traffic information for drivers. Skilled and generally skilled drivers gazed at right lane more often than the other areas except front lane; however, unskilled drivers gazed at left lane more frequently. So with the increasing of driving experience, drivers’ fixation area got wider. It was beneficial for safe driving.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
This paper was sponsored by the National Natural Science Foundation of China (Project Approval no.: 51308250), Infrastructure projects of Jilin Provincial Science and Technology Department (201105014), and the Specialized Research Fund for the Doctoral Program of Higher Education (20110011120045).
- A. Smiley, T. Smahel, and M. Eizenman, “Impact of video advertising on driver fixation patterns,” Transportation Research Record, no. 1899, pp. 76–83, 2004.
- H. Kosaka, H. Koyama, and H. Nishitani, “Analysis of driver's behavior from heart rate and eye movement when he/she causes an accident between cars at an intersection,” SAE Technical Paper 01-0432, NARA Institute of Science and Technology, Nara Prefecture, Japan, 2005.
- M. Sivak, “The information that drivers use: is it indeed 90% visual?” Perception, vol. 25, no. 9, pp. 1081–1089, 1996.
- E. C. B. Olsen, S. E. Lee, and B. G. Simons-Morton, “Eye movement patterns for novice teen drivers: does 6 months of driving experience make a difference?” Transportation Research Record, no. 2009, pp. 8–14, 2007.
- R. R. Mourant and T. H. Rockwell, “Mapping eye-movement patterns to the visual scene in driving: an exploratory study,” Human Factors, vol. 12, no. 1, pp. 81–87, 1970.
- G. Underwood, D. Crundall, and P. Chapman, “Selective searching while driving: the role of experience in hazard detection and general surveillance,” Ergonomics, vol. 45, no. 1, pp. 1–12, 2002.
- D. E. Crundall and G. Underwood, “Effects of experience and processing demands on visual information acquisition in drivers,” Ergonomics, vol. 41, no. 4, pp. 448–458, 1998.
- G. Underwood, P. Chapman, N. Brocklehurst, J. Underwood, and D. Crundall, “Visual attention while driving: sequences of eye fixations made by experienced and novice drivers,” Ergonomics, vol. 46, no. 6, pp. 629–646, 2003.
- T. Falkmer and N. P. Gregersen, “A comparison of eye movement behavior of inexperienced and experienced drivers in real traffic environments,” Optometry and Vision Science, vol. 82, no. 8, pp. 732–739, 2005.
- J. Maequeen, “Some methods for classification and analysis of multivariate observations,” in Proceedings of the 5th Berkeley Symposium on Mathematical Statistics and Probability, vol. 1, pp. 281–297, 1967.
- D. J. McRae, “A Foritran in iterative K-means cluster analysis program,” Behavioural Seience, no. 16, p. 423, 1971.
- P. S. Bradley, “Refining initial Points for K-Means clustering,” in Proceedings of the 15th International Conference on Machine Learning, pp. 91–99, 1998.
- W. Yuan, Study on dar driver’s dynamic visual characters test on city road [Ph.D. thesis], Chang’an university, Xi’an, China, 2008.
- T. Falkmer and N. P. Gregersen, “Fixation patterns of learner drivers with and without cerebral palsy (CP) when driving in real traffic environments,” Transportation Research F, vol. 4, no. 3, pp. 171–185, 2001.
- B. S. Mark and W. Ben, An Exploratory Exam Ination of Eye Movement in High Speed Driving, Transportation Research Board, Washington, DC, USA, 2004.
- G. Underwood, P. Chapman, K. Bowden, and D. Crundall, “Visual search while driving: skill and awareness during inspection of the scene,” Transportation Research F, vol. 5, no. 2, pp. 87–97, 2002.