摘要
针对井下复杂受限环境下人脸、虹膜、指纹和掌纹等常常比较模糊,从而使得基于这些生物特征的井下人员身份识别率不高问题。在Warshall算法和最大最小判别准则的基础上,提出了一种最大最小判别映射的步态识别方法。该方法利用Warshall算法快速得到数据的类别关系,由此构建类内和类间散度矩阵。与经典的步态识别方法相比,该方法充分利用了数据的局部信息和类别信息,使得数据降维后在低维空间同类样本之间的距离减小,而异类样本之间的距离增大。与经典的监督子空间维数约简方法相比,该方法在构建类内和类间散度矩阵时不需要判别数据的类别信息,能够提高算法的性能。在真实步态数据库上进行了一系列实验。实验结果表明,利用该方法进行基于步态的煤矿井下人员身份鉴别是有效可行的。
Due to the complexity and constrained space in underground mine, the images of human face, iris, fingerprint and palmprint often become blurred, the recognition rates of the mine underground personnel identification based on these biological characters are not higher than that in the regular environment. Based on Warshall algorithm and maxi- rain criterion, a method of gait recognition, named maximin discriminant projection (MMDP), was proposed. In MM- DP, the label relationship of the data was quickly explored by the Warshall algorithm. The within-class and between- class scatter matrices were constructed by the label relationship. Compared with the traditional gait recognition meth- ods, the proposed method makes full use of the local information and class information of the gait data, so that in low- dimensionality projecting space, the distance between any pairwise samples belonging to the same class was reduced, while the distance between any pairwise samples coming from different classes was enlarged. Compared with the classi- cal subspace dimensional reduction algorithms, in the proposed method, it was not necessary to judge whether two sam- ples belong to the same class or not when constructing the within-class and between-class scatter matrices, which can improve the performance of the proposed algorithm. A series of gait recognition experiments were conducted on the real gait databases. Experimental results verify the proposed method is effective and feasible for mine underground person- nel identification by using gait.
出处
《煤炭学报》
EI
CAS
CSCD
北大核心
2013年第10期1894-1899,共6页
Journal of China Coal Society
基金
国家自然科学基金资助项目(61272333)
陕西省科技厅自然科学基金资助项目(2011JM8011)
陕西省科学技术研究发展计划资助项目(2011K06-36)