几何变化结合图像增强的航拍图像小目标检测算法

doi:10.3778/j.issn.1002-8331.2412-0294

摘要/Abstract

摘要： 航拍图像小目标检测场景复杂，检测指标下降，提出几何变化结合图像增强的航拍图像小目标检测算法。以YOLOv8n为基线，DCNv2-CA-GEO并行提取空间特征和通道特征，动态调整卷积核和池化核，快速适应几何变化；SPD-OK-CSP调整通道维度，捕捉细粒度特征，跨阶段全局特征感知，多尺度特征学习，提高图像增强；Dysample动态调整差分采样点，优化上采样计算；Dyhead动态调整特征张量维度，整合上下文信息，优化检测头性能；Inner-Wise-MPD-IoU多视角增益分配策略，平衡不同质量样本特征，准确回归真实框，优化泛化能力。评价指标mAP@0.5、mAP@0.5：0.95、Precision、Recall，消融和对比实验在VisDrone2021上分别提升6.1、4.4、6.0、5.0个百分点，在LEVIR-Ship上分别提升3.4、2.2、3.1、5.3个百分点，优于基线算法和次优算法；泛化实验在VOC2007+2012上分别提升2.0、4.3、1.5、3.1个百分点，鲁棒性良好。

关键词: 航拍图像, 小目标检测, YOLOv8n, SPDConv, Omni-Kernel, DCNv2, Dyhead

Abstract: Aerial image small target detection is complex, leading to a decline in detection metrics. An algorithm combining geometric transformations and image enhancement is proposed. Using YOLOv8n as the baseline, DCNv2-CA-GEO extracts spatial and channel features in parallel, dynamically adjusting the convolutional and pooling kernels to quickly adapt to geometric changes. SPD-OK-CSP adjusts channel dimensions, capturing fine-grained features and enhancing image quality. Dysample optimizes upsampling, while Dyhead improves detection head performance. The Inner-Wise-MPD-IoU strategy balances sample features and optimizes generalization. Evaluated by mAP@0.5, mAP@0.5:0.95, Precision, and Recall, experiments on VisDrone2021 show improvements of 6.1 percentage points in mAP@0.5, 4.4 percentage points in mAP@0.5:0.95, 6.0 percentage points in Precision, and 5.0 percentage points in Recall. On LEVIR-Ship, the improvements are 3.4 percentage points in mAP@0.5, 2.2 percentage points in mAP@0.5:0.95, 3.1 percentage points in Precision, and 5.3 percentage points in Recall. Generalization tests on VOC2007+2012 demonstrate enhancements of 2.0 percentage points in mAP@0.5, 4.3 percentage points in mAP@0.5:0.95, 1.5 percentage points in Precision, and 3.1 percentage points in Recall, indicating good robustness.

Key words: aerial image, small target detection, YOLOv8n, SPDConv, Omni-Kernel, DCNv2, Dyhead

齐向明, 李晓龙. 几何变化结合图像增强的航拍图像小目标检测算法[J]. 计算机工程与应用, 2025, 61(18): 218-230.

QI Xiangming, LI Xiaolong. Geometric Transformation Combined with Image Enhancement for Small Target Detection in Aerial Images[J]. Computer Engineering and Applications, 2025, 61(18): 218-230.

参考文献

[1] LIU Y, SUN P, WERGELES N, et al. A survey and performance evaluation of deep learning methods for small object detection[J]. Expert Systems with Applications, 2021, 172: 114602.
[2] YANG Q, MA S, GUO D Q, et al. A small object detection method for oil leakage defects in substations based on improved faster-RCNN[J]. Sensors, 2023, 23(17): 7390.
[3] MA X J, JI K F, XIONG B L, et al. Light-YOLOv4: an edge-device oriented target detection method for remote sensing images[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2021, 14: 10808-10820.
[4] YU J, ZHENG H, XIE L, et al. Enhanced YOLOv7 integrated with small target enhancement for rapid detection of objects on water surfaces[J]. Frontiers in Neurorobotics, 2023, 17: 1315251.
[5] HUANG Z C, WANG J L, FU X S, et al. DC-SPP-YOLO: dense connection and spatial pyramid pooling based YOLO for object detection[J]. Information Sciences, 2020, 522: 241-258.
[6] WANG Z Z, XIE K, ZHANG X Y, et al. Small-object detection based on YOLO and dense block via image super-resolution[J]. IEEE Access, 2021, 9: 56416-56429.
[7] JU M R, LUO H B, WANG Z B, et al. The application of improved YOLOV3 in multi-scale target detection[J]. Applied Sciences, 2019, 9(18): 3775.
[8] AHMED M, WANG Y, MAHER A, et al. Fused RetinaNet for small target detection in aerial images[J]. International Journal of Remote Sensing, 2022, 43(8): 2813-2836.
[9] ZHONG X. CAL-SSD: lightweight SSD object detection based on coordinated attention[J]. Signal, Image and Video Processing, 2024, 19(1): 31.
[10] 黄健宸, 王晗, 卢昊. 结合轻量化骨干与多尺度融合的单阶段检测器[J]. 中国图象图形学报, 2022, 27(12): 3596-3607.
HUANG J C, WANG H, LU H. One-stage detectors combining lightweight backbone and multi?scale fusion[J]. Journal of Image and Graphics, 2022, 27(12): 3596-3607.
[11] 钟映春, 郑海阳, 张文祥, 等. 面向航拍图像中工程车辆检测与识别的改进胶囊网络[J]. 中国图象图形学报, 2022, 27(8): 2380-2390.
ZHONG Y C, ZHENG H Y, ZHANG W X, et al. Improved capsule network method for engineering vehicles detection and recognition in aerial images[J]. Journal of Image and Graphics, 2022, 27(8): 2380-2390.
[12] 王一旭, 肖小玲, 王鹏飞, 等. 改进YOLOv5s的小目标烟雾火焰检测算法[J]. 计算机工程与应用, 2023, 59(1): 72-81.
WANG Y X, XIAO X L, WANG P F, et al. Improved YOLOv5s small target smoke and fire detection algorithm[J]. Computer Engineering and Applications, 2023, 59(1): 72-81.
[13] 刘树东, 刘业辉, 孙叶美, 等. 基于倒置残差注意力的无人机航拍图像小目标检测[J]. 北京航空航天大学学报, 2023, 49(3): 514-524.
LIU S D, LIU Y H, SUN Y M, et al. Small object detection in UAV aerial images based on inverted residual attention[J]. Journal of Beijing University of Aeronautics and Astronautics, 2023, 49(3): 514-524.
[14] 赵鹏飞, 钱孟波, 周凯琪, 等. 改进YOLOv7-Tiny农田环境下甜椒果实检测[J]. 计算机工程与应用, 2023, 59(15): 329-340.
ZHAO P F, QIAN M B, ZHOU K Q, et al. Improvement of sweet pepper fruit detection in YOLOv7-tiny farming environment[J]. Computer Engineering and Applications, 2023, 59(15): 329-340.
[15] 杨慧剑, 孟亮. 基于改进的YOLOv5的航拍图像中小目标检测算法[J]. 计算机工程与科学, 2023, 45(6): 1063-1070.
YANG H J, MENG L. A small target detection algorithm based on improved YOLOv5 in aerial image[J]. Computer Engineering & Science, 2023, 45(6): 1063-1070.
[16] 邱昊, 钟小勇, 黄林辉, 等. 面向航拍小目标的改进YOLOv5n检测算法[J]. 电光与控制, 2023, 30(10): 95-101.
QIU H, ZHONG X Y, HUANG L H, et al. An improved YOLOv5n detection algorithm for aerial photography of small targets[J]. Electronics Optics & Control, 2023, 30(10): 95-101.
[17] 龙燕, 杨智优, 何梦菲. 基于改进YOLOv7的疏果期苹果目标检测方法[J]. 农业工程学报, 2023, 39(14): 191-199.
LONG Y, YANG Z Y, HE M F. Recognizing apple targets before thinning using improved YOLOv7[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39(14): 191-199.
[18] 王春艳, 张成谦, 王祥, 等. 改进YOLOv7-tiny网络的多尺度无人机航拍小目标检测[J]. 测绘科学, 2023, 48(11): 189-199.
WANG C Y, ZHANG C Q, WANG X, et al. YOLOv7-tiny improved network multiscale UAV aerial photography small target detection incorporating Swin-Transformer[J]. Science of Surveying and Mapping, 2023, 48(11): 189-199.
[19] 赖杰, 彭锐晖, 孙殿星, 等. 融合注意力机制与多检测层结构的伪装目标检测[J]. 中国图象图形学报, 2024, 29(1): 134-146.
LAI J, PENG R H, SUN D X, et al. Detection of camouflage targets based on attention mechanism and multi-detection layer structure[J]. Journal of Image and Graphics, 2024, 29(1): 134-146.
[20] 齐向明, 严萍萍, 姜亮. 基于YOLOv8n的航拍图像小目标检测算法[J]. 计算机工程与应用, 2024, 60(24): 200-210.
QI X M, YAN P P, JIANG L. Small target detection algorithm for aerial images based on YOLOv8n[J]. Computer Engineering and Applications, 2024, 60(24): 200-210.
[21] 梁燕, 何孝武, 邵凯, 等. 改进YOLOv8的无人机航拍图像目标检测算法[J]. 计算机工程与应用, 2025, 61(1): 121-130.
LIANG Y, HE X W, SHAO K, et al. Target detection algorithm for UAV images based on improved YOLOv8[J]. Computer Engineering and Applications, 2025, 61(1): 121-130.
[22] ZHU X Z, HU H, LIN S, et al. Deformable ConvNets V2: more deformable, better results[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019.
[23] HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 13708-13717.
[24] ZHANG H, XU C, ZHANG S. Inner-IoU: more effective intersection over union loss with auxiliary bounding box[J]. arXiv:2311.02877, 2023.
[25] TONG Z, CHEN Y, XU Z, et al. Wise-IoU: bounding box regression loss with dynamic focusing mecharenism[J]. arXiv:2301.10051, 2023.
[26] SILIANG M, YONG X. MPDIoU: a loss for efficient and accurate bounding box regression[J]. arXiv:2307.07662, 2023.
[27] ZHENG Q H, TIAN X Y, YU Z G, et al. Robust automatic modulation classification using asymmetric trilinear attention net with noisy activation function[J]. Engineering Applications of Artificial Intelligence, 2025, 141: 109861.