基于凸壳检测的红外船舶轮廓提取技术

Contours Extraction of Infrared Boat Based on Convex Hull Detection

利用星载红外探测系统获取的图像进行船舶目标识别,具有隐蔽性强、抗干扰能力强、全天候工作以及智能化程度高等优点,已经广泛应用于军事领域.红外图像船舶识别的前提条件是准确提取目标的特征,其中最重要的就是几何特征,而这需要首先确定船舶目标区域的轮廓.提出基于凸壳检测的红外船舶目标区域轮廓提取技术,借助凸壳检测算法,确定目标区域的凸集,进而描绘并提取目标的轮廓.该技术可以修补由于图像分割造成的目标区域残缺,同时不受目标区域内部孔洞的影响,比现有的基于边缘检测、二值图像轮廓提取,以及基于轮廓跟踪的轮廓提取方法更能描绘船舶的真实外形.实验表明,提出的算法可以有效提取船舶目标区域的轮廓,为准确提取船舶目标的几何特征,以及提高船舶目标识别的准确率和稳定性奠定了基础.

Infrared image boat target detect and recognition by using satellite infrared detection system has the advantages of strong concealment, anti-interference ability, all-weather work, high intelligence and so on. It has been widely used in the military field. The prerequisite of infrared image boat recognition is to accurately extract the features of the boat target. Geometric features are the most significant features. However, it is necessary to first determine the contours of the boat target. Contours extraction technology of infrared boat based on convex hull detection was presented. The technology can repair the boat target region due to image segmentation, and it is not affected by the target region within the hole. Compared with the edge detection based, binary image contour extraction and contour tracking based, the proposed technology can better describe the real shape of boat. Experiments show that the proposed method can extract boat target region contour effectively. The technology is the foundation for the accurate extraction of the geometric features and the accuracy and stability of boat target recognition.