CAJ | 학술논문

安防系统输入数据质量各异，部分质量差的照片会引起误报率升高，如何将这部分数据优化排除，筛选用于人脸建模的安防系统数据源是本文的目的。采用量化照片评价指标的方法，通过对照片中人脸部分的清晰度、对比度等质量因素进行评价，将其量化并排序从而得到数据源质量序列。在软件工程化实践中对随机抽取的100张照片质量评价验证，其评价结果准确率可达95％。以文中所提到的方法对采集数据优化后，其建模后的数据能有效地降低比对过程中误报率，提高安防系统识别的准确率。
안방계통수입수거질량각이，부분질량차적조편회인기오보솔승고，여하장저부분수거우화배제，사선용우인검건모적안방계통수거원시본문적목적。채용양화조편평개지표적방법，통과대조편중인검부분적청석도、대비도등질량인소진행평개，장기양화병배서종이득도수거원질량서렬。재연건공정화실천중대수궤추취적100장조편질량평개험증，기평개결과준학솔가체95％。이문중소제도적방법대채집수거우화후，기건모후적수거능유효지강저비대과정중오보솔，제고안방계통식별적준학솔。
The quality of input data is quite different, some poor quality pictures will increase the rate of false alarm in the safety defense system. The purpose of this paper is to optimize these data and choose the source data for mod?eling the face detection system. Using the method of quantifying the picture evaluation indexes to measure the pic?tures with the factors influencing quality of face, such as resolution, contrast ratio, etc., then quantify and se?quence these factors, and thus derive the quality sequence of data source. In the software engineering practice, 100 images were selected randomly to evaluate their quality, proving that the accuracy of evaluation reaches 95%. The data collected by the method mentioned above were optimized and then used for modeling, proving that the false a?larm in the process of comparison can be reduced effectively and the accuracy of identification of safety defense sys?tem increased.