【摘要】 针对运动平台上的红外热成像仪拍摄目标出现偏离,模板与目标无法匹配,提出结合MSRCR图像增强的AKAZE算法红外图像目标提取的方法。该方法采用MARCR增强红外图像,采用频域低通滤波去除图像噪声,使用AKAZE进行快速多尺度特征提取,生成特征点描述子。采用FLANN快速最近邻逼近搜索算法,快速匹配特征点。利用RAN-SC随机抽样一致性算法,剔除误匹配特征。该算法避免通常计算量大、匹配错误、目标轮廓不清晰等不足,有助于移动平台上定点红外检测。更多还原

【Abstract】 Aiming at the problem that the infrared thermal image taken on mobile platform is deviated,the template and the target cannot match,pro-poses the method of extracting the infrared image object with the AKAZE algorithm and MSRCR image enhancement.The method usesMARCR to enhance the infrared image,uses frequency domain low pass filter to remove the image noise,uses AKAZE to carry out fastmulti-scale feature extraction,and generates the feature points descriptors.FLANN fast nearest neighbor proximity search algorithm isused to quickly match the feature points.Using the random sampling consistency algorithm of RANSC,the mismatched feature is eliminat-ed.This method avoids the shortcomings such as large computational complexity,matching error and unclear target contour,which can helpto realize fixed-point target infrared detection on mobile platform.更多还原