摘要
身体部位可划分为头部、胸部、腹部和盆腔,对各部位的自动识别可以通过其包含的代表性器官的检测实现。对原始二维Haar-AdaBoost方法进行改进,根据器官的实际物理大小使用固定图像分辨率,有效地选择关键Haar特征,减少计算量,采用一种基于正态分布拟合的方法,将二维检测结果转换为三维器官的感兴趣区域边界。测试结果表明,该方法对胸部和腹部识别的敏感度达到97%以上。
The body part is divided into head, chest, abdomen and pelvis. Automated localization of body part was implemented by detecting representative organs. Based on the original 2D Haar-AdaBoost method, the fixed image resolution based on the physical size of the organ was used. Haar features were selected to improve the efficiency of algorithm. A Gaussian fitting method was applied to obtain the 3D volume of interest of organs by the use of the 2D detection results. Preliminary results show that the sensitivity of chest and abdomen localization is higher than 97%.
出处
《计算机工程与设计》
北大核心
2017年第1期247-252,共6页
Computer Engineering and Design
基金
国家自然科学基金青年科学基金项目(81301282)
上海科学技术委员会重大基金项目(13411950108)
上海科学技术委员会医学影像及放射治疗工程技术研究中心基金项目(13DZ2250300)
