Lane markings detection based on e-maxima transformation and improved hough
According to Malaysians Unite for Road Safety (MUFORS) online survey, human error, for example, improper vehicle deviation or unintentional lane change is one of the main causes of traffic accident. Lane shift in traffic can be complex and dangerous. This study aims at developing a fast, low-cost, a...
| Summary: | According to Malaysians Unite for Road Safety (MUFORS) online survey, human error, for example, improper vehicle deviation or unintentional lane change is one of the main causes of traffic accident. Lane shift in traffic can be complex and dangerous. This study aims at developing a fast, low-cost, and sophisticated system with the ability to detect unexpected lane changes that may reduce the probability of a vehicle straying out of lane. Various road models to identify the lanes have been explored including straight-line, B-snack, linear-parabolic model, and deformable model. Most lane models, either simple or lack of flexibility or complex, may cause heavy computation in processing the time needed. The feature of roadway has certain degree of curvature and constraints, for instance, no sudden road turn is the design for road safety driving. A short segment of a long curve with a relatively low curvature is approximated as a straight line, based on this point, the important contribution of this thesis presents a lane detection algorithm using E-MAXIMA transformation and improved Hough transform which is the algorithm with great efficiency, high robustness and also at low cost to detect road lane markings. First of all, the region of interest from input image to reduce the searching space is defined; then the image into near field-of-view and the far field-of-view is divided. In the near field-of-view, Hough transform will be applied to detect lane markers after image noise filtering and lane features extraction by E-MAXIMA. The experimental results based on collected video data under complex illumination conditions had proved that the proposed algorithm is able to detect the road lane marking efficiently achieving a correction rate of 95.33%. The process time on average is 32 ms/f, namely every second can deal with 31.25 frames that demonstrate superior and robust results compared to other existing methods. To conclude, the work done in this thesis may apply to autonomous driving navigation and driving security assistance. The potential of such a system is further linked to the system with the vehicles’ turn signal, whereby the system will be able to detect an unintentional drift out of the lane. |
|---|