基于光谱成像系统的小目标识别技术的研究

Research on Small Object Recognition Based on Spectral Imaging System

针对天地复杂背景下红外弱小目标的检测,提出了一种基于光谱成像融合top-hat变换、光谱识别的新方法。首先将改进的top-hat变换应用于红外弱小目标图像,将图像中与目标特性相似的像素进行增强,并去除云层背景对目标的干扰,得到若干疑似目标的增强结果,并结合阈值分割方法对增强结果进行筛选,剔除大部分不符合弱小红外目标特性的背景干扰点。然后采用光谱匹配算法,通过三种测度进行加权匹配,去除图像处理中对干扰物、噪声等因素的干扰,得到筛选过的目标识别结果,通过实验证明,本文提出的算法相较于传统弱小目标检测有更好的检测效果。

We proposed a new method based on spectral imaging fusion top-hat transform and spectral recognition for the detection of infrared dim and small targets in the complex background of earth and sky. First this paper applies the improved top-hat transform infrared dimsmall target image, similar to the target characteristics of the pixel in the image enhancement, and removes the interference of background on the target, the clouds get some suspected targets enhanced as a result, and combining the threshold segmentation method to enhance to filter results, eliminating most do not conform to the small and weak infrared target characteristics of the background noise points. Then, the spectral matching algorithm is adopted to make weighted matching through three measures to remove the interference of distractor and noise and other factors in the image processing, and the filtered target recognition results are obtained. The experimental results show that the algorithm proposed in this paper has a better detection effect than the traditional weak and small target detection.