摘要
针对红外地面固定目标无直接可用基准图,目标边缘模糊,不利于目标识别检测等问题,提出一种新的基于形状特征的红外目标检测方法。首先在依据红外图像形状特征的基础上,引入图像的灰度形态学梯度,扩展对比度、增长图像边缘特征;其次进行多子区划分,并设计像素相似性计算,有效地结合了像素点的灰度信息以及空间位置;最后在考虑实时图中非真实边缘影响时,加入了Canny算子检测边缘,分离目标与背景,在红外实时图中检测出所需的目标。实验结果表明,本文所提算法检测率能达到80%以上,与直方图检测方法、Hausdorff算法、Nprod算法相比,分别平均提高了近10%,11%,20%,算法花费时间缩短2/3。对于红外固定目标,该方法具有检测率高、速度快、精度高等优点。
For the infrared image of fixed target without available base image,it is difficult to recognize the target due to the blurry target edge. A new target detection algorithm based on shape characteristics matching is proposed. First ly, based on shape characteristics of infrared image, the mathematic morphological gradient algorithm is introduced in order to expand contrast and strengthen edge character. Secondly, multi seedregions are designed, and similarity cal culation of pixels is introduced. Pixels' gray information and spatial location are integrated efficiently. Lastly, consider ing the impact of unreal edge, Canny operator is added into the edge detection for separating the target and back ground. The requisite target is detected in the real infrared image. Experiment results show that the detection probabili ty can reach up to 80%. Comparing to histogram detection algorithm, Hausdorff distance algorithm and Nprod algo rithm,the probability increases by 10% , 11% and 12% respectively and the spent time is shortened to 2/3. This method has better performance at detection probability ,computing speed and recognition precision.
出处
《激光与红外》
CAS
CSCD
北大核心
2013年第1期49-53,共5页
Laser & Infrared
关键词
形状特征
多子区
像素相似性
空间位置
目标检测
shape character
multiple seed-region
similarity of pixels
spacial location
target detection