Khashawi, Fadhel
(2011)
Detecting motorcycles in road scenes: differences between car drivers and motorcyclists.
PhD thesis, University of Nottingham.
Abstract
Many motorcycle “right of way” violations are believed to be a result of low conspicuity and a failure to perceive the motorcycle. Drivers with previous experience of riding motorcycles tend not to commit this type of violation, indicating an influence of awareness or expectancy. This thesis investigates the way these drivers with motorcycle experience inspect traffic scenes in the appearance of a motorcycle, and compares their performances to drivers with no motorcycle experience. It investigates also the effect of motorcycle awareness promotion signs such as “Think Bike” that are used in safety campaigns across the United Kingdom. It also tries to develop a design that could be used in eye movement studies to compare eye movements patterns of drivers with motorcycle experience and those who do not have motorcycle experience.
In experiment 1, a group of drivers with motorcycle experience were tested on how readily they perceive motorcycles in traffic still pictures. Their performances were compared with another two groups of drivers without motorcycle experience, one of them were shown warning signs promoting motorcycle awareness used in a safety campaigns. Still pictures of real traffic environments were used as stimuli, allowing control over several variables: appearance and visual saliency of the motorcycle, danger of the situation, and the presence of warning signs promoting motorcycle awareness. The subjects were asked wither they think it was safe to cross the road or not. Then the motorcycle was digitally edited and was made less salient to make it hard to be detected, or it was removed or replaced by a car. Motorcycle saliency was determined using the Itti and Koch (2000) saliency map program that calculates in which order that motorcycle is likely to be spotted compared to other objects in the scene. The results showed slower inspection times when the scene was safe, indicating a greater extent of the search for hazards. Also the presence of the motorcycle had the effect of extending search times for drivers with motorcycle experience, again reflecting a more extensive search associated with the type of driving experience. Prolonged inspection times appeared with car drivers after presentation of the warning signs used in a safety campaign in the UK.
In experiment 2, the same traffic pictures were used but with a searching task instead of a hazard detection related task. A group of drivers with no motorcycle experience were giving a searching task about an aspect that may appear in the picture. The aspect was either about a motorcycle or other vehicle in the scene, and the motorcycle was either salient and easy to detect, non salient, or absent. The results showed high accuracy in the searching task, with no effect whether the searching was on an aspect related to the motorcycle or not, and motorcycle saliency did not appear to have any impact. In experiment 3, the same experiment was repeated, but viewing time of the pictures was reduced to 500 milliseconds. The results showed a small decrease on accuracy compared to experiment 2, and it failed to spot any difference across the motorcycle presentation. These results highlight the effect of Top-down processing of the scene rather than a bottom-up processing.
Experiment 4 continued the same method of using still pictures of traffic scene; but with a more control over the traffic aspects in the scene. Pictures of approaching motorcycle to an intersection from several distances were used in this experiment. The motorcycle was then digitally edited to be either removed or replaced by a car. This experiment was a replication of Crundall, Humphrey, & Clarke (2009) with the addition of the saliency factor. Two groups of drivers without motorcycle experience were tested, one of them was shown warning signs promoting motorcycle awareness used in a safety campaigns and the other was not. The task was a simple searching task of spotting an oncoming vehicle, with a limited viewing time of 500 milliseconds for each picture. The results showed that warning signs did increase accuracy in spotting the oncoming vehicle. Saliency also has a significant impact, especially with motorcycles approaching from far distance. In experiment 5, the same experiment repeated with only one group of drivers. The task was to evaluate each picture on wither they think it was safe to pull in front of the approaching vehicle or not. The variation of saliency showed an effect on decision and increasing viewing time. The results of these two experiments confirmed the effect of motorcycle warning signs. The effect of saliency also started to come out after controlling some aspects of the pictures such as the location of the oncoming vehicle.
Experiment 6, a further modifications were added to the pictures to insure fully controlled about all objects that appears on the scene. On this experiment, saliency was changed with the amount of traffic density of the road. Instead of making the vehicle difficult to spot, a different number of vehicles appear on each scene to act as distracters. This method allows for more realistic pictures, and to have traffic related objects to compete with the approaching vehicle in attracting attention. A group of drivers with motorcycle experience were asked wither they think it was safe to pull out on front of the oncoming vehicle or not. Their performance was compared with another two groups of drivers without motorcycle experience, one of them were shown motorcycle’s warning signs. Eye movements were recorded in this study to see if there are any differences between groups on how they spot the motorcycle compared to cars. Results replicated the effect of the distance in the previous experiment. It also found an effect on the number of the distracters that appears in the scene resembling the saliency effect in the previous experiments. The effect of the safety campaign signs appears also. Regarding the eye movement pattern, results showed a slightly different pattern between groups that indicates that motorcycle awareness affects the way drivers inspect the scene. This awareness could be achieved by either having motorcycle riding experience, or simply by priming the appearance of motorcycle appearance using appropriate road warning signs.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
