视频监控图像高速运动物体稳定特征识别仿真
|
摘要
在智能视频监控系统中,运动目标稳定特征提取是实现高速运动物体行为理解、识别以及跟踪的前提条件。传统的运动目标识别方法难以精确提取出稳定的运动目标,以及某些干扰物体中有部分区域与运动目标形状相同时,识别率低、实时性差等问题。针对上述问题,提出基于动态模板匹配的高速运动物体稳定特征识别方法。采用帧间差分算法确定高速运动目标物体可能存在的区域,获取较为清晰的高速运动物体目标轮廓,并基于运动目标轮廓信息构造稳定特征不变量。以稳定特征不变量为中心选择指定大小的矩形区域,将其作为下一帧模板匹配算法的模板,将待测图像的特征描述与运动目标模板特征进行匹配,完成对高速运动目标的识别。实验结果表明,所提方法能够实现对高速运动目标物体的实时分析,同时满足实际应用中,对高速运动目标物体识别实时性和准确性的要求。
Traditional method of moving target recognition is difficult to accurately extract the stable moving target. Meanwhile,the recognition rate and the real-time performance are low. Therefore,a method to recognize stability feature of high-speed moving object based on dynamic template matching was presented. At first,the inter-frame difference algorithm was used to determine the possible area of high-speed moving object and obtain clear contour of high-speed moving object. Then,the stable feature invariant was constructed based on the contour information of moving object. Taking the stable feature invariant as the center,a rectangular area with a specified size was selected,which was used as the template of the next frame matching algorithm. Finally,the feature description of the image to be tested was matched with the feature of moving target template. Thus,we completed the recognition of high-speed target. Simulation results show that the proposed method can achieve real-time analysis of high-speed object. Meanwhile,this method satisfies the real-time and accuracy requirements for high-speed object recognition.
Traditional method of moving target recognition is difficult to accurately extract the stable moving target. Meanwhile,the recognition rate and the real-time performance are low. Therefore,a method to recognize stability feature of high-speed moving object based on dynamic template matching was presented. At first,the inter-frame difference algorithm was used to determine the possible area of high-speed moving object and obtain clear contour of high-speed moving object. Then,the stable feature invariant was constructed based on the contour information of moving object. Taking the stable feature invariant as the center,a rectangular area with a specified size was selected,which was used as the template of the next frame matching algorithm. Finally,the feature description of the image to be tested was matched with the feature of moving target template. Thus,we completed the recognition of high-speed target. Simulation results show that the proposed method can achieve real-time analysis of high-speed object. Meanwhile,this method satisfies the real-time and accuracy requirements for high-speed object recognition.
引文
[1]刘洪彬,常发亮.权重系数自适应光流法运动目标检测[J].光学精密工程,2016,24(2):460-468.
[2]宋涛,李鸥,刘广怡.基于空时多线索融合的超像素运动目标检测方法[J].电子与信息学报,2016,38(6):1503-1511.
[3]蔡娟,李东新.基于优化k均值建模的运动目标检测算法[J].国外电子测量技术,2016,35(12):20-23.
[4]董俊宁,杨词慧.空间约束混合高斯运动目标检测[J].中国图象图形学报,2016,21(5):588-594.
[5]张应辉,刘养硕.基于帧差法和背景差法的运动目标检测[J].计算机技术与发展,2017,27(2):25-28.
[6]曹义亲,程威,黄晓生.关于视频图像运动目标跟踪识别仿真研究[J].计算机仿真,2017,34(1):191-196.
[7]郭伟,高媛媛,刘鑫焱.改进的基于混合高斯模型的运动目标检测算法[J].计算机工程与应用,2016,52(13):195-200.
[8]王刚.基才Oracle的运动身体指标监测系统设计与实现[J].电子设计工程,2017,17(25):148-151.
[9]孙凯,谢林柏.结合W4算法和LBP模型的运动目标检测方法[J].计算机工程与应用,2017,53(5):187-191.
[10]周同雪,朱明.视频图像中的运动目标检测[J].液晶与显示,2017,32(1):40-47.
[2]宋涛,李鸥,刘广怡.基于空时多线索融合的超像素运动目标检测方法[J].电子与信息学报,2016,38(6):1503-1511.
[3]蔡娟,李东新.基于优化k均值建模的运动目标检测算法[J].国外电子测量技术,2016,35(12):20-23.
[4]董俊宁,杨词慧.空间约束混合高斯运动目标检测[J].中国图象图形学报,2016,21(5):588-594.
[5]张应辉,刘养硕.基于帧差法和背景差法的运动目标检测[J].计算机技术与发展,2017,27(2):25-28.
[6]曹义亲,程威,黄晓生.关于视频图像运动目标跟踪识别仿真研究[J].计算机仿真,2017,34(1):191-196.
[7]郭伟,高媛媛,刘鑫焱.改进的基于混合高斯模型的运动目标检测算法[J].计算机工程与应用,2016,52(13):195-200.
[8]王刚.基才Oracle的运动身体指标监测系统设计与实现[J].电子设计工程,2017,17(25):148-151.
[9]孙凯,谢林柏.结合W4算法和LBP模型的运动目标检测方法[J].计算机工程与应用,2017,53(5):187-191.
[10]周同雪,朱明.视频图像中的运动目标检测[J].液晶与显示,2017,32(1):40-47.