| In the past two decades,the construction of motorway has been vigorously developed in China.The number of motorways in the country has increased rapidly.The social transportation becomes more convenient.The traffic facilities are becoming more and more perfect.The motorway lane marker plays an important role in regulating vehicle driving and reducing the accident rate.However,as time goes by,the facilities becomes obsolete.Due to the characteristics of motorway,it is easy to cause the defect of the lane marker,which has safety hazards.At present,the defect detection of the motorway lane marker is mainly observed by the relevant staff through the naked eye,assisting with some testing equipment.It takes a huge labor cost and it is difficult to guarantee the detection effect.Therefore,it is urgent to research the intelligent detection of motorway lane marker defects.Using intelligent and automated methods for lane marker defects detection is of great significance for promoting the level of motorway intelligent construction and improving the motorway maintenance system.The main work and research of the paper are as follows:1.The vehicle detection system was designed,which equipped with visible camera,infrared camera and GPS.The industrial computer was used as the central processor.The system contains four modules: environmental awareness,defect location,humancomputer interaction and power supply.The real-time detection,defect diagnosis and driving assistance functions could be realized.2.The algorithm combined with double threshold and Hough transform was proposed to detect lane marker defects.The algorithm includes preprocessing,Adaptive threshold segmentation,Hough transform and double threshold judgment.The video acquired by the vehicle detection system is analyzed by the algorithm to detect the possible lane marker defects.The experiment proved that the proposed algorithm’s average detection rate was 82.86% and the running time was under 0.1 second,which could realize the lane marker detection function.3.The algorithm using SSD(Single Shot Multibox Detection)model for lane marker detection was proposed.The LM-3 motorway lane marker data set was constructed and the model for detection was trained.After the verification of experiment,the average detection rate was 94.87% and the running time was under 0.01 second,which had nice detection effect.4.Infrared and visible image fusion based on BEMSD(Bidimensional Empirical Mode Shearlet Decomposition)and improved fuzzy set was proposed.The algorithm combined bidimensional Empirical Mode Decomposition and non-subsampled shearlet transform to achieve the strong adaptability.The fusion rules adapted to the fusion framework were designed to achieve better result in subjective vision and objective indicators.The algorithm could be used to realize the safety assistance function. |
