Study On Engineering Drawings Vectorization

Engineering drawings were scanned into a raster image into the computer, the graphics file to save the form of pixels, you need to take a lot of memory, and do not facilitate the timely update of information, raster images therefore become the research focus of vector, raster image is a vector Species of the raster image processing, analysis, identification, will be transformed into the ultimate computer-aided design software identify the vector graphics file format of a method. Vector raster image is related to the processing of image analysis, artificial intelligence and pattern recognition has important theoretical significance of a commercial value and great technology. This paper is the process for carrying out vector drawing some discussion and research.Vector drawing process, including image binarization, noise reduction, smoothing, thinning, vectorization and vector output. In this paper, features and analysis of computation of wavelet denoising based on mathematical morphology, wavelet transform and the improved denoising method combining mathematical morphology, using this method has very good denoising effect, and to ensure that the image Completeness and accuracy. As lines, arcs, circles are vector drawings of image recognition and processing of the three major basic image elements, a straight line in the identification process, for the current calculation of straight line Hough Transform to Identify large, the accuracy was not high on the slope of this article Fractional look-up table was modified Hough transform to construct the gradient threshold, the classification of pixels, by setting the gradient threshold to filter out the pixels do not belong to a straight line to address the shortcomings of large computing Hough transform; the least square method Hough transform and improve this combination to improve the recognition accuracy, so that both the direct use of Hough transform to overcome the identified shortcomings when accuracy is not high, but also to overcome the least squares fitting a straight line at the point when susceptible to noise interference shortcomings. By analyzing the geometric characteristics of circles, and arcs, using circle Hough transform and geometric properties of a combination of methods to effectively solve the traditional Hough transform circle generated when the three-dimensional parameter identification problem, which reduces the computation and complexity. Finally, the results by the vector system in accordance with AutoCAD DXF file format to record, to achieve the vector of the output process.Experiments show that the algorithm results in denoising significant follow-up to meet the needs of vector processing, improved recognition algorithms identify the lines and circles in time with high accuracy to identify the scanned drawings of lines and circles, to the program was originally designed Ideas.