中文摘要：

采用改进的区域生长法进行红外热图中遮挡目标的提取。首先利用阈值化方法提取出一个包含全部目标信息的矩形子区域，后续处理仅针对该矩形区域进行处理。然后证明了数学形态学的膨胀梯度定理（dila-tion gradient theorem，DGT）及其逆定理，并利用DGT逆定理查找图像的局部极大值点，将灰度值不为0的局部极大值点作为区域生长法的种子点。接着采用阈值条件与区域灰度差相结合的生长准则，分别在各种子点八个方向进行生长。最后采用二值形态学开运算进行去毛刺处理。理论与仿真结果表明，该算法适合于红外图中遮挡目标的提取。

英文摘要：

The infrared technology was used to detect the concealed object. An improved region growing method was proposed as the detection method. First, the approximate area of the object was extracted according the threshold method. In this step, a rectangle sub image which contained all the target pixels was segmented from the sample image. All the following process was done on the sub image. Second, the morphology gradient theorem （DGT） and its converse theorem were both proved. And the DGT reverse theorem was used to look for the regional maxima points of the sub image. In the improved region growing method, the seeds were those regional maxima points. Then the threshold term and region grayscale intensity difference term were combined as the growing rule. The growth was go on in four directions of all the seeds. At last, the binary morphological opening operation was used to clean the burrs. Both the theory and simulation results show that the algorithm can detect the concealed infrared object well.