Study On The Technique Of Motorway Speed Limit Sign Detecting And Recognition

With the progress of technology,the development of city planning and the number of running cars and traveling individuals are greatly increasing,the problems of traffic safety are becoming more and more serious.In this kind of condition, research about the intelligent transportation system (ITS) is started,which involves image processing,digital signal processing,electronics technique,artificial intelligence,information technique,pattern recognition, communication technique and system engineering technique etc. ITS is significant to guarantee road safety.The road sign recognition (RSR),an important component of intelligent vehicle,has become a crucial part of semi-automatic or automatic vehicle.By collecting and recognizing the road signs information in the vehicle driving process,then giving alert or warnings to drivers,it can help keep the transportation smoothing and avoid traffic accidents.Therefore,there are important theories and practical values in the research of the road signs recognition. In the paper,the research objects are located in speed limited signs,considering the reliability and validity.According to the results from the experiment research on algorithms,we can see that the algorithm is very valid,and amount of computation is very small.The main contents of the paper include:1.It causes photo's color distortion,photo's size incongruity and photo's fuzziness,due to environmental factors such as weather condition and traffic information.In the paper,pretreatment is done to enhance the quality of the images,including using histogram equalization enhancement algorithms to enhance color based on I channel of HSI color model,utilizing scaling techniques to adjust the size and utilizing restoration techniques to clear the fuzzy images.2.According to the color featheristics of the objects,the paper uses the threshold segmentation algorithm in RGB color space and in H(hue)channel of HSI color space respectively,and the object area and the background area will be separated into binary images.In RGB space,it makes objects separated from the background distinctly,and saves time,while there are many noises which are similar to the color of objects in the images.The segmentation of image in H channel of HSI space can segment red very distinctly with few noises because black does not have a hue,but it can't segment black objects area from background and since there is a course of RGB to HSI space conversion process, it takes more time comparing with RGB space segmentation.Through comparison analysis,the RGB color space segmentation is used,and isolated points are eliminated by median filtering in the paper.3.According to the circle featheristics of the objects from the paper studying,the circle targets are extracted utilizing the circle algorith.The calculation of the circle is obtained by the size and perimeter of each region,which are simple in principle,small in calculation, good real-time and good extracting effecs.After extracting the circle,the circle is eliminated through projection method,and the remains are processed by mathematical morphology in order to fill the holes,eliminate the burrs on edge and make the object region smooth.4.Recognition is the most important work in the paper.The objects are recognized through template matching and BP neural network after extracting the feather by the template method.Through comparing,BP neural network is better than template matching in recognizing the objects.