摘要
将Itti模型应用于海洋监视卫星图像舰船目标的检测中。简要阐述了Itti模型的算法处理过程,并将视觉注意点的提取转移过程建立为电容阵列充电模型。针对Itti模型的诸多问题,比如所提取的显著区域形状大小固定、小半径检测实时性差、大半径检测包含背景区域多等,提出了改进算法:引入离散矩变换,增强了图像纹理特征响应;采用阈值分割的方法由显著点搜寻显著区域,提高了检测精度和实时性。运用Matlab对算法进行测试,实验结果表明,改进算法所提取的显著区域形状大小基本与目标一致,实时性好,且显著区域包含背景少。与Itti模型相比,改进算法更适合应用于海洋监视卫星图像舰船目标检测提取。
An improved Ittirs model is applied on the ship targets detection of ocean surveillance satellite images. We illustrate the algorithm process of Ittirs model, and introduce a capacitor array charging model to describe the extracting and transferring process of the focus of attention. To solve the problems existing in the traditional Itti＇s model such as the fixed shape and size of the extracted salient region, the poor real-time detecting performance when the radius of salient region goes too small, and excessive background areas contained in the salient region when the radius is set too large, the algorithm is improved in some aspects in this paper. Firstly, the discrete moment transform is introduced to the algorithm to enhance the response of image texture features. Then, the threshold segmentation method is chosen to extract the salient region with the focus of attention, and thus both the detection accuracy and real-time performance are improved greatly. According to the Matlab test results of the improved algorithm, it is verified that both the shape and size of the salient region are consistent well with the ship targets; the background contained in the salient region is also reduced significantly. Moreover, the improved algorithm has a good real-time performance. It comes to the conclusion that compared with Itti＇s model, the improved algorithm is more effective and suitable for the extraction of ship tarlzets detection of ocean satellite images.
出处
《激光与光电子学进展》
CSCD
北大核心
2013年第12期57-65,共9页
Laser & Optoelectronics Progress
基金
国家863计划(2008AA121803)
国家自然科学基金(6110066)
关键词
图像处理
视觉注意
显著特征
海洋监视卫星图像
舰船目标识别
阈值分割
image processing visual attentiom salient feature ocean surveillance satellite image ship targetsdetection threshold segmentation
