Research And Construction Of Concise Line Style Portrait Generation System

Line style portrait painting is a way of depicting characters.It vividly outlines the gestures and expressions of characters with simple and abstract lines.It is a unique form of artistic expression.it Has always been enthusiastically sought after by people.However,the drawing of line-style portraits often requires certain professional knowledge and professionally trained personnel to complete,and the drawing process requires careful observation by the painter,which is time-consuming and labor-intensive.With the continuous advancement of computer vision technology methods,the use of computers to simulate human painting has gradually become a hot issue.The line drawing effect generated by the current mainstream algorithm is more complicated,which is not conducive to the drawing of the robot manipulator.In response to this problem,this article has done the following work:1.This paper proposes a Concise Line Portrait Generation Based on Semantic Segmentation(CLPG-SS)method.First,perform semantic segmentation on the face image to divide the face For different areas,extract edge contours and facial features based on different areas,optimize the edge tangential flow,thereby enhancing the direction information;on this basis,use the line drawing to generate a harmonic image,and use the optimized edge tangential flow,Semantic face segmentation results and reconciled images,adjust the parameters of the line extraction method for different segmentation regions,achieve line filtering in detail-independent areas and line enhancement in detail key areas,and generate clean line portraits.The experimental results show that the CLPG-SS method proposed in this paper can effectively extract the main contour lines of the face,and this method realize the targeted adjustment of the detailed lines for different regions,which improves the drawing efficiency of the robot manipulator.2.Research the use of neural networks for face line extraction,and this article propose a simple line portrait generation method DA-UNIT based on the attention residual module to Generative Adversarial Networks.This method combines VAE,GAN and attention mechanism,adds spatial position attention mechanism and channel attention mechanism to the VAE encoding and decoding network,which improves the feature representation ability of the network,and the addition of the attention mechanism expands the convolution kernel Feeling wild.The comparison with the Gatys neural style transfer method and the Pix2 Pix method,and the ablation experiment of the attention mechanism show that the DA-UNIT method can better ensure the continuity of the extracted lines.3.Combining the two methods proposed in this article,a set of clean line style portrait generation system is developed.The system uses C++,Lib Torch and Open CV to develop line style portrait generation,and connects the robotic arm to the drawing board.Draw it out and design a more anthropomorphic robotic arm trajectory plan,making the painting process more in line with people’s drawing habits,and finally achieve the purpose of real-time collection of face images and rapid drawing of simple line style portraits.