基于多特征融合的弱小目标检测方法研究

提出了一种基于多特征距离图的红外弱小目标检测方法。弱小目标的许多特征，如局部熵、平均梯度强度等，不但刻画了弱小目标的特点而且易于提取。通过特征融合技术，可以将弱小目标检测问题转化成在一个多特征空间的极值求取问题。该方法利用已经提取的多个特征，采用特征融合技术构造一个距离图像，再对该图像进行二值化处理，达到目标检测的目的。通过对实际的红外图像序列进行小目标检测，验证了所提方法的可行性和有效性。

Detection of small infrared targets based on multi-feature fusion

A novel method for small weak target detection based on multi-feature distance map (MFDM) in image sequences is proposed. Small weak targets have many features like local entropy, average gradient strength etc. These features not only describe the characteristics of small infrared targets, but also are easy to be extracted. Multi-feature-based fusion techniques are applied to detect weak targets by converting the problem of detecting small targets to the search for peak values in specified feature space where multi-feature vectors space (MFVS) is considered. Target detection is performed in DM which can be derived according to feature vectors. The targets are detected in complex backgrounds via binarizing a DM image constructed by multi-feature fusion. The proposed approach is validated using actual infrared image sequences with sea-sky backgrounds. Experimental results demonstrate the robustness and high performance of the proposed method.