The Research Of On-line Recognition Technology Of Hand-drawn Sketch

The recognition of hand-drawn sketches is an important subject in the field of humancomputer interaction.It allows designers to express their ideas freely in the conceptual design stage of the product as the interaction between pen and paper,which realizes the efficiency and nature of human-computer interaction.In this paper,a method of recognition based on geometric features of graphic and parametric curve fitting is proposed to solve the problem of recognition rate of feature extraction based on element recognition method.In this paper,the sketch of the pen interactive input is used as the research object.After the stroke data is collected,the noise such as sampling error and jitter is eliminated by means of implementing the preprocess such as clustering elimination,resampling and endpoint closure error elimination;According to the geometric features of the sketch and the online recognition process,the primitive elements of the graphs are fitted and identified by the least squares method and the quadratic curve invariant theory;In the stroke division,the feature points are divided into the corner points where the local angle changes dramatically and the tangent points of the smooth transition between primitives,For the corner point,the discrete curvature maximum is selected as the basis for the extraction,the relation between the angle and distance are used to eliminated the pseudo-corner,improving the recognition rate of the corner;the vector product and the piecewise parametric polynomial curves fitting are as the basis for extraction when finding the tangent points;In the pattern recognition,the spatial geometric position relation between the elements in the graph is analyzed,after extracting the eigenvectors representing the graphs,the classifier is constructed,the pattern recognition method is used to identify the graphics,and then the graphs are regularizedThrough the verification of contrast experiment for the proposed algorithm,the results show that the algorithm can effectively perform the stroke segmentation and the pattern recognition.