【摘要】 提出了基于均匀网格重采样算法的原型三维人脸对应算法．基于人脸特征实现原型三维人脸之间的对应,克服了传统对应算法对应效果差,算法精度低的缺陷;提出了基于改进遗传算法的形变模型匹配算法．新的匹配算法不依赖于目标函数的梯度信息和初值,全局搜索能力强．优化过程中交叉和变异概率的调节机制,有效提高了算法的收敛速度和精度．实验结果表明,新的对应算法可有效实现原型三维人脸之间的对应,提高形变模型的精度．新的匹配算法能有效提高模型匹配的效率和精度,缩短模型匹配时间．更多还原

【Abstract】 A uniform mesh resampling based alignment algorithm is proposed to align prototypical 3D faces．This algo- rithm enables us to achieve aligning of 3D prototypes based on facial features．It is free of the weaknesses of conventional ones and precision．Improved genetic algorithm based model matching method is able to match morphable model to 2D facial images independently of initial values and gradient of object function,and is capable of global searching．Regulation of crossover and mutation probabilities during optimizing process effectively improves the convergent speed and precision of the algorithm．Experimental results show that this novel alignment algorithm effectively applied to align the prototypes, and improves precision of morphable model．The novel matching method effectively improves the efficiency and precision of model matching,and shortens the time for the matching process．更多还原