摘要
目的研究基于加权统计模型和遗传算法的人像目标检测方法．方法提出用于检测非刚性图像目标的一种加权统计模型，该模型含有待检测目标样本的统计值，可有效解决图像目标发生形变和倾斜情况下检测器的适应性问题．结果给出人像目标检测应用中的处理结果，人像目标不发生倾斜和无噪声时，即使在局部发生形变和改变灰度对比度情况下也可以得到满意的结果，在低信噪比和目标发生倾斜时，正确识别率大于９０．０％（Ｓ／Ｎ≥１０ｄＢ）和７６．７％（倾斜度小于６°），采用改进的遗传算法，计算速度比单独应用顺序检测算法提高６０倍．结论将统计模型与改进的遗传优化算法结合可提高算法效率和精度．
Aim To propose a method for detecting and locating human face, based on the weighted statistical model and genetic algorithms. Methods A weighted statistical model(WSM) was presented to solve the problem of adaptability to deformation and tilting while detecting nonrigid image targets using a priori knowledge information and statistical knowledge based model. Results Applying the WSM to the practical human face detection problem, satisfactory results can be obtained even if there exists the local deformation and the gray level contrast changes in the noise free and tilting free case. The success ratio for detection is greater than 90% when S/N ratio is above 10 dB and greater than 76 7% when tilting degree is less than 6°. Adopting the modified GAs makes the computational time be 60 times saved compared to the original sequential detection mode. Conclusion\ By combining the WSM with GAs the efficiency and accuracy of the detection method can be enhanced.
出处
《北京理工大学学报》
EI
CAS
CSCD
1999年第2期199-202,共4页
Transactions of Beijing Institute of Technology
基金
河北省教委发展基金
关键词
加权统计模型
遗传算法
人像检测
人像识别
weighted statistical model(WSM)
genetic algorithms
face detection
