An Algorithm Of License Plates Detection And Character Segmentation In Real Scene

The License Plate Recognition System has important significance in the intelligent Transportation system. The most related researches are focused on the single license plate. However, multiple vehicles may appear in the detection region at the same time in the real scene. The more important thing is that the correct recognition of multiple license plates can save the using cost largely because it can reduce the number of the cameras vastly. The image preprocessing, the license coarse detection, the license plate tilt correction and character segmentation are studied.1. Image preprocessing. Image enhancement methods are firstly adopted to enhance the image contrast and weaken the impact of uneven illumination. And then Canny operator is used to detect the edges of the image.2. License plate coarse locating. A series of morphological processing is employed to eliminate the interference edges and obtain the closed and "solid" regions made up by candidate edges. And then the edge dot density and plate ratio are applied to eliminate some interference regions and select candidate regions.3. License plate accurate locating. To solve owed segmentation and eliminate the non-license plate regions further, the character connected domain analysis and nearest neighbor chain are proposed to obtain exact plate boundaries.4. The license plate tilt correcting and character segmenting. Firstly, an improved character connected domain analysis method is presented to remove the connected domains of the adhesive or fractured characters. Then the default character location information and the tilt angle of the license plate are calculated by the leaved ones. Finally, the characters are segmented. The method removes the difficulty of the character segmentation from the deficient license plate characters and omits the steps of removing the license plate frame and rivets.114 images of multiple license plates with poor quality captured from different real scenes are processed. The results of locating license plate and segmenting characters show that the proposed method can meet the requirements of the license plate location and character segmentation in the real scene because it has the small calculation and real-time performance.