一个多尺度的物体检测网络,集成空间通道协作注意力,用于远程传感图像
Lijun Ma1, Chengjun Xu2, Kun Jiao1
1College of Energy (College of Modern Shale Gas Industry), Chengdu University of Technology, Chengdu 610059, China.
Sensors (Basel, Switzerland)
|February 27, 2026
概括
这项研究引入了一种用于遥感图像的新型多尺度物体检测网络,提高了小型和大型目标的准确性. 新模型增强了特征表示,降低了计算成本,优于现有方法.
科学领域:
- 计算机视觉 计算机视觉
- 遥感 遥感 遥感 遥感
- 机器学习 机器学习
背景情况:
- 目前的遥感物体检测模型在规模变化,杂乱的背景和小物体检测方面扎.
- 固定大小的卷积内核在特征提取中模糊物体轮和极限表示.
- 现有的注意力机制缺乏空间和通道特征的深度整合,增加了计算开销.
研究的目的:
- 开发一个用于远程传感图像的多尺度物体检测网络.
- 增强特征的感知和代表多个规模的目标,特别是小的.
- 为了提高检测准确度,同时降低计算成本.
主要方法:
- 引入了一个跨通道的多尺度特征提取模块 (CC-MSFE),以增强特征表示.
- 开发了一个通道空间交叉注意力机制 (CSCA),集成通道注意力 (CA),空间注意力 (SA) 和交叉注意力融合 (CAFM).
- 设计了跨空间和通道维度的动态交互和联合优化,以改进检测.
主要成果:
- 在 DIOR 上实现了 78.1% 的 mAP,在 HRRSD 上实现了 90.6% 的 mAP,在 RSOD 数据集上实现了 96.5% 的 mAP.
- 在DIOR和HRRSD分别表现比YOLOv11高0.7%和1.4%.
- 在RSOD上超过YOLOv8的2.1%,参数数量和计算复杂性较低.
结论:
- 拟议的网络有效地解决了遥感物体检测方面的挑战.
- 集成的空间通道协作注意力显著提高了多尺度对象的检测精度.
- 该模型在检测性能和计算效率之间实现了卓越的平衡.
相关概念视频
02:16Light Acquisition
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
01:15Depth Perception and Spatial Vision
Depth perception is the ability to perceive objects three-dimensionally. It relies on two types of cues: binocular and monocular. Binocular cues depend on the combination of images from both eyes and how the eyes work together. Since the eyes are in slightly different positions, each eye captures a slightly different image. This disparity between images, known as binocular disparity, helps the brain interpret depth. When the brain compares these images, it determines the distance to an object.