基于马尔科夫随机场模型的多视角异质多模图像的目标检测

摘要

研究复杂背景中不同视角的不同光质图像中的特定目标检测问题.利用马尔可夫随机场模型,提出了一个基于地面区域匹配和空间约束关系的目标检测方法.在可见光俯视参考图像和红外光侧视观测图像的实验数据集上的检测结果表明,区域匹配能够有效提高召回率,而空间约束能够有效降低虚警率,获得了比一般异质光图像检测中基于边缘的方法更好的检测结果.该方法降低了不同视角带来的影响,同时能够克服图像间光质不同带来的检测困难,能够有效处理复杂背景下不同光质图像的匹配及其中目标的准确检测定位.%The paper study the problem of target detection in multiview and multimodal (multisensor) images. To solve it, a new target detection method is proposed by using Markov Random Field and based on the ground region matching and spatial constraints. The detection result in the data sets including visual reference images in top views and infrared sensed images in side views shows that the region matching can improve the recall rate and the spatial constraints can reduce the false alarm rate. The detection result is better than that by the common methods using edges to detect in multimodal images. The target detection can effectively reduce the impact of multiview, overcome the difficulties of detection in multimodal images and solve the problems of multimodal image registration and target detection in multiview and multimodal images against complex backgrounds.