MIRSat-QL

High-quality annotated datasets from diverse scenarios play a crucial role in the development of deep learning algorithms. However, due to the strict access limitations of space-based infrared satellite platforms, space-based infrared small target datasets are scarce. Therefore, we have developed the MIRSat-QL dataset, based on a space-based infrared satellite platform, for space-based dynamic scene infrared target detection. Our data is synthesized from space-based infrared satellite images and ground-based infrared cameras capturing airborne targets. The specifics are as follows。背景图像来自上海技术物理研究所 （SITP） 开发的 QLSAT-2 核心有效载荷，其中包括 QLSAT-2 高分辨率光学相机。QLSAT-2 长波红外 （LWIR） 相机在 14 公里的高度提供 500 米的地面分辨率。背景序列包含复杂的场景，例如城市区域、海洋表面、陆地和云。运动目标图像由中波红外 （MWIR） 相机捕获，该相机记录飞机等机载目标，形成用于航空观测的红外运动目标数据集。对数据集进行注释以获得目标掩码。然后将掩码图像和背景图像融合以生成新的图像序列。具体来说，我们首先对掩码图像应用高斯平滑以柔化目标边缘，确保背景和目标之间更平滑的过渡。接下来，我们根据目标区域的像素强度对背景图像和掩码进行加权处理。在非目标区域中，背景被完全显示，而在目标区域中，像素强度被控制以调整目标信噪比。背景加权完成后，目标区域和背景区域被合并以产生最终的融合图像。目标区域的像素通过平滑的蒙版图像直接添加到加权背景图像中，确保两者之间自然平滑的过渡。

High-quality annotated datasets from diverse scenarios play a crucial role in the development of deep learning algorithms. However, due to the strict access limitations of space-based infrared satellite platforms, space-based infrared small target datasets are scarce. Therefore, we have developed the MIRSat-QL dataset, based on a space-based infrared satellite platform, for space-based dynamic scene infrared target detection. Our data is synthesized from space-based infrared satellite images and ground-based infrared cameras capturing airborne targets. The specifics are as follows. The background images are sourced from the core payload of QLSAT-2 developed by the Shanghai Institute of Technical Physics (SITP), which includes the high-resolution optical camera of QLSAT-2. The QLSAT-2 long-wave infrared (LWIR) camera provides a ground resolution of 500 meters at an altitude of 14 kilometers. The background sequences cover complex scenarios such as urban areas, ocean surfaces, landmasses, and clouds. The moving target images are captured by a medium-wave infrared (MWIR) camera, which records airborne targets like aircraft, forming an infrared moving target dataset for aviation observation. The dataset is annotated to obtain target masks. Subsequently, the mask images and background images are fused to generate new image sequences. Specifically, we first apply Gaussian smoothing to the mask images to soften the target edges, ensuring a smoother transition between the background and the targets. Next, we perform weighting processing on the background images and masks based on the pixel intensity of the target regions. In non-target regions, the background is fully displayed, while in target regions, the pixel intensity is controlled to adjust the target signal-to-noise ratio (SNR). After the background weighting is completed, the target regions and background regions are merged to produce the final fused images. The pixels in the target regions are directly added to the weighted background images via the smoothed mask images, ensuring a natural and smooth transition between the two.

搜集汇总

背景与挑战

背景概述

MIRSat-QL是一个用于天基动态场景红外目标检测的数据集，基于天基红外卫星平台开发，数据合成自卫星红外图像和地基红外相机捕获的空中目标。该数据集包含100组图像序列，提供了训练和测试路径，适用于深度学习算法开发，旨在解决天基红外小目标数据稀缺的问题。