陈晗 周强



摘要:针对传统水域救生过程无落水人体检测先验知识、检测效率低的现状,提出了一种基于倒影图像检测的水域落水人员判断方法。该方法以水面倒影检测为核心,将倒影与真实落水目标进行比对,通过水纹特征实现对两者的区分检测,并采用方向梯度直方图(histogram of oriented gradient,HOG)特征以及支持向量机(Support Vector Machine,SVM)的人体检测算法对真实落水目标进行最终人体识别检测。实验表明,该方法对落水人员检测准确率为98.2%,达到了对真实落水人员的有效检测,提升了水面救援检测的效率和速度。
关键词: 水域救生;落水人员检测;倒影图像检测;HOG-SVM
中图分类号:TP311 文献标识码:A 文章编号:1009-3044(2018)26-0175-04
Abstract:Focusing on the problems that caused from no prior knowledge of human detection and the low detection efficiency in the traditional waters lifesaving process, a method for judging the persons falling into the water basin Based on reflection image is proposed. This method uses water surface reflection detection as the core, compares the reflection with the real falling water target, realizes the distinguish detection between the two through the water pattern feature, and uses the HOG and SVM human body detection algorithm to perform the final human body recognition detection on the real falling target. Experiments show that the accuracy rate of this method is 98.2% for the drowning personnel, which can achieve effective detection of real drowning personnel and improve the efficiency and speed of water surface rescue detection.
Key words:Water life-saving; Detection of falling personnel; Reflection image detection; HOG-SVM
我国国土幅员辽阔,对外开放的公园人工湖、水库和湖泊数量巨大,但实际生活中由于安全措施不到位、监控存在盲区和死角,以及湖泊区域水上应急救助设施配置较为薄弱等原因[1],导致溺水死亡事故频繁发生,产生了恶劣社会影响,故而落水人员的救生问题就显得尤为重要。
目前,我国水上救生市场主要存在人员设备不足、研发体系不完善以及技术含量較低等问题,整体处于有待开发的状态。常见的针对人员落水问题的救生装置主要包括:传统救生装置和新型救生装置,其中传统救生装置主要包括救生筏和救生圈等[2],主要依靠人力进行救援行为,并存在浮力不足、体积大、救援投掷准确性不高等问题,而新型装置则主要包括潜航器应急救生装置、智能救生网、弹……