基于多向梯度背景预测的红外目标检测算法

为解决复杂空中背景下红外弱小目标的检测，提出一种新的基于多向梯度的背景预测方法。该方法根据云层边缘区域、平缓背景区域及弱小目标所呈现的不同梯度特点，采取不同方法分别进行预测；基本保留云层边缘区域和平缓背景区域的点，而对弱小目标区域采用邻域低灰度值点进行预测。然后经过背景消除和阈值分割，将弱小目标检测出来。仿真结果表明，该算法对复杂空中背景预测有很高的准确性，能够更加有效地抑制云层边缘引起的虚警，将红外弱小目标点检测出来。

Dim Infrared Targets Detection Based on Multi-gradient Background Prediction

In order to solve the detection problems of dim infrared targets in the complicated air background, a new algorithm of background prediction based on multi-orientation gradient is presented. The different prediction methods are respectively adopted according to the different characteristics of gradients for the areas of cloud＇s edge, flat background and dim infrared targets. For the points in the areas of the cloud＇s edge and flat background, the most similar point is sought as a predic rive point. For the dim infrared targets, the point with lower gray value in neighborhood is used as a predictive point. Then the dim infrared targets are detected after the background elimination and threshold segmentation. The simulation results show that this algorithm has a high accuracy of prediction for the dim IR targets in the complicated air background, can effectively suppress false alarms caused by the edge of cloud, and can detect dim infrared targets.