MotionBlurredUAVsAndTheirTemperatureWaveformTimeSeries

We selected the infrared UAV image data with severe motion blur from the CVPR 3rd Anti-UAV 2023 dataset and extracted it as a temperature waveform sequence according to the method in the article. A total of 1095 images with sharp motion blur are selected. From the first image, every 5 images are from the same relative motion. For the group with less than 5 images in one relative motion, we have added them to 5 images by replication. Then, we performed six Gaussian filters with different parameters on these 1095 images. Finally, we extracted the brightest points of the drone targets in these images, and connected the brightest points of 25 images in the same scene into a sequence with a step size of 5.

我们从CVPR第三届Anti-UAV 2023数据集选取了存在严重运动模糊的红外无人机图像数据，并按文中方法将其提取为温度波形序列。共选取1095张存在明显运动模糊的图像。从首张图像开始，每5张图像对应同一相对运动；对于单组相对运动下图像数量不足5张的情况，我们通过复制将其补充至5张。随后，我们对这1095张图像施加了六种参数不同的高斯滤波（Gaussian filter）。最后，我们提取了这些图像中无人机目标的最亮点，并将同一场景下25张图像的最亮点按步长5连接成序列。