支持向量机的红外目标自动检测与识别

Infrared target detection and cognition based on the Support Vector Machine

针对红外图像提出一种基于支持向量机的目标检测和识别算法.首先运用数学形态学方法对背景进行滤波,突出候选目标;选取适当的阈值和边缘检测算子对候选目标进行图像二值分割和边缘提取;最后以候选目标的边界不变矩作为特征,用支持向量机方法进行目标的识别,确定目标的位置.实验表明,该方法能够有效地实现对红外目标的检测和识别,并具有较高的抗噪声和抗复杂背景的能力.

Support Vector Machine is a new machine learning technique based on the structure risk minimization principle of statistic learning theory.An Automatic Infrared Target Recognition algorithm based on the Support Vector Machines is presented.First,a mathematical morphology filter is used to suppress the noise in the infrared image and enhance the potential targets.Then,the proper grayscale threshold and edge-detector are selected to obtain the binary image and the target boundary.Finally,the infrared targets are recognized by using the Support Vector Machines from the potential targets based on the feature of boundary moment invariants.The experimental results prove that the algorithm can effectively detect and recognize the infrared target and possess the preferable properties of counter-noises and counter-complex background.