Detection And Recognition Of Speed Limit Sign Based On Radial Symmetry Transform

According to public information,the number of cars in China has accounted for one-tenth of the amount of cars,but the number of deaths due to traffic accidents is higher than the amount of cars,reaching 16%.WHO 2015 statistics show that China's road traffic accidents caused by the death toll has more than210,000,the vast majority of the driver due to speeding caused.In this context,the intelligent driving system(including unmanned systems)concept came into being,its main goal is to be use of computer vision technology to collect and automatically analyze the road information,and effectively improve the safety of road traffic.Timely and accurate identification of speed limit signs can remind drivers to maintain secure driving speed.In this paper,an improved radial symmetry algorithm is proposed to detect the speed limit sign.At present,the location of speed-limiting signs is mainly based on the characteristics of color and shape.In this paper,the shape feature is adopted to use in the detection of speed-limiting signs,because the detection algorithm of shape feature is more stable than that of color,and is not easily affected by natural environment and other factors.So firstly,with the prior knowledge of location information,we determine the regions of interest to detection the speed-limiting signs to avoid unnecessary calculation loss.Then in the regions of interest,the detection algorithm based on the radial symmetry is proposed,which has lower computational complexity and is more accurate for detecting the position of the signs.In the calculation of the edge gradient of the image,we choose the Scharr operator to make gradient map to contain more edge features,so the corresponding center of the candidate area has more votes,the region located is more accurate.According to the maximum circumscribed circle,the region,which only contain the characters of the speed-limiting sign,is segmented from the candidate region.Then the Gaussian filter and Otsu algorithm are used to obtain the binary image.We compute the number of all the connected binary regions tojudge whether the candidate region is speed limit sign.We preserve the largest connected area in the binary image.From the largest connected area,we extracte the complete characters,which are recognized by the projection segmentation method.After normalizing the characters,the feature vectors of the coarse mesh feature are extracted.Finally,the trained BP neural network is used to classify and identify the characters.Experiments show that the proposed algorithm has good robustness and is real-time.