I. THE PROPOSED ALGORITHM
A real-time, calibration-free and AUTOSAR-compliant lane departure warning system algorithm is proposed. First, the region of interest (ROI) is extracted to reduce the outlier lines in the image. Then, pre-processing stage is carried out using Gaussian pyramid to smooth the image and reduce its dimensions, which decrease the unnecessary details in the image. A lane detection stage is then developed based on Edge Drawing Lines (EDLines) algorithm that is a real-time line segment detector, which has false detection control. After that, basic machine learning (ML) concepts are employed in the lane filtering and clustering stage to reject the lines with low probability of being lane boundaries. Based on these lines, an advanced reference-counting algorithm is introduced to track the lanes, between consecutive frames taken by a single front camera, and predict the missing lanes.
II. EXPERIMENTAL RESULTS
According to ISO 17361:2017 standard, the environment conditions required to test LDWS are flat and dry asphalt, lane markings are directly visible by the driver and horizontal visibility range shall be greater than 1 km. The stated conditions in the standard describe an ideal environment, but real life is not ideal. To guarantee that the driver safety is accomplished by our system, the system is tested by a various challenging weather and illumination conditions: clear, cloudy, rainy, day, sunset and night. ISO 17361:2017 standard tests, which are Warning generation in a curve, Repeatability and False alarm, are also performed on the proposed system successfully. The proposed system achieves detection rate of 99.41%, departure rate of 100% and average processing time of 100 milliseconds milliseconds on ECU with TC397 microcontroller up to 300 MHz and 6.9 MB RAM.
Calibration phase consumes a lot of time and effort for getting its knowledge, configuring calibration modules and performing the calibration itself to the system parameters. In addition, calibration process costs a lot of money for buying both calibration tools and AUTOSAR XCP calibration module.
In this paper, a new reliable and robust algorithm to implement an LDWS that is AUTOSAR-compliant and do not require calibration, as it does not have parameters that need to be tuned and adjusted to make the system work properly, is introduced. Hence, the proposed system saves time, effort and money required by calibration process. Moreover, it is efficient to be used in the self-driving systems in the Original Equipment Manufacturers (OEMs) cars that are composed of either AUTOSAR-based or non-AUTOSAR-based ECUs
Mr. Islam Gamal, Mentor Graphics, EGYPT; Mr. Ahmed Hamed, Mentor Graphics, EGYPT