基于SVM分类的红外舰船目标识别

Infrared ship-target recognition based on SVM classification

针对海天背景下红外舰船目标识别提出了一种基于机器学习的分类算法。该算法首先利用分割算法提取红外图像中的连通区域,并对原图相应的位置进行标记和归一化处理,然后利用HOG特征提取标记区域的高维特征向量,用线下样本库训练得到的SVM分类器对所提取的HOG特征进行高维特征空间的分类,识别目标和干扰。仿真实验表明,该算法具有良好的性能,在复杂海天干扰背景下能够有效地识别红外舰船目标。

Aiming at the ship-target recognition of sea-sky background, an classification algorithm based on machine learning was proposed. In the method, the segmentation algorithm was firstly adopted to extract connected region in infrared image. Then, the corresponding position of the original image was marked and normalized. Afterwards, the high-dimensional feature vector of branded region by using the HOG algorithm was extracted. Finally, the high-dimensional feature vector that came form suspected target area was classified by the SVM classifier which was trained by sample library. Simulation experimental result indicates that the algorithm not only can effectively recognise the infrared ship-targets in complex sea-sky background of interference, but also have good performance.