Visual fuzzy recognition of the object nameplate in video sequences

This paper presents an approach of the fuzzy recognition of the object nameplate in video sequences based on computer vision. After the nameplate character image is layered, the critical points and critical line part can be extracted in the layers of the images. According to the stroke extracting regularity, the critical points and line parts can be merged into strokes. The stroke structure is used to describe the spatial relation of the conjoint strokes. The text image can be mapped into the feature space of the stroke and stroke structure. In the visual feature space, a fuzzy classifier is extensively used. The visual fuzzy recognition can much improve the recognition of the rotated, distorted, deformed and defiled text image in the video sequences. Experiments carried out on large data sets of video sequences, automatically recognized the licenses of the automobiles in real time, show very promising results.