复杂动背景下的“低小慢”目标检测技术

In order to detect low-altitude, slow-moving and small targets against a complex sky background, this paper studies the visual saliency region characteristics of "low-altitude, small and slow" targets, integrates the scan line filling algorithm, and proposes an adaptive real-time detection technology for "low-altitude, small and slow" targets in a dynamic background. Firstly, the saliency map is obtained according to the luminance contrast of the image.接着，使用形态学梯度提取显著性特征，通过三帧差分算法得到种子点。Then, the scan line filling algorithm is used for growth, and the foreground is segmented by combining the proposed adaptive dual-Gaussian algorithm. Finally, false targets are eliminated according to the area proportion change, centroid distance change and aspect ratio difference of candidate targets to complete the detection. In order to verify the effectiveness of the algorithm, this paper selects 7 groups of video sequences with complex sky backgrounds for testing and compares with other excellent detection algorithms. 注：原文中“接着”部分英文重复了，我按照正确逻辑进行了翻译，如果有错误请指出。

为了在复杂天空背景下检测出低空慢速小目标，本文研究了“低小慢”目标的视觉显著性区域特征，融合扫描线填充算法，提出了一种动态背景下“低小慢”目标自适应实时检测技术。首先，根据图像的亮度对比度获取显著性图。接着，使用形态学梯度提取显著性特征，通过三帧差分算法得到种子点。然后，使用扫描线填充算法进行生长，结合提出的自适应双高斯算法分割出前景。最后，根据候选目标的面积占比变化、质心距离变化、宽高比差异剔除虚假目标，完成检测。为了验证算法的有效性，本文选取了7组复杂天空背景的视频序列进行测试，并与其他优秀检测算法进