Detection-Driven Gaussian Mixture Probability Hypothesis Density Multi-Target Tracker for Airborne Infrared Platforms DOI Creative Commons
Min Hong, Jiarong Wang, Ming Zhu

и другие.

Sensors, Год журнала: 2025, Номер 25(11), С. 3491 - 3491

Опубликована: Май 31, 2025

Recent advancements in the unmanned aerial vehicle remote sensing field have highlighted effectiveness of infrared sensors detecting and tracking time-sensitive ground targets, particularly within domain early warning surveillance. However, limitations inherent airborne platforms can lead to irregular imaging inadequate textural features. This study presents a multi-object system specifically designed for weak-textured aimed at enhancing detection accuracy stability. Initially, improvements are made YOLOv10 model through incorporation modules such as DSA, c2f_fasterblock, NMSFree, which collectively enhance robustness targets. Subsequently, results employed conjunction with GM-PHD tracking, enabling rapid stable target tracking. The proposed methodology demonstrates 2.3% improvement 3.8% increase recall when assessed using publicly available datasets. Notably, key metric, MOTA, achieves value 90.7%, while IDF1 score reaches 94.6%. findings from experiments indicate that algorithm surpasses current methodologies regarding effectiveness, accuracy, context multi-target tasks, thereby meeting requirements associated tasks.

Язык: Английский

InSAR-YOLOv8 for wide-area landslide detection in InSAR measurements DOI Creative Commons
Ruopu Ma, Haiyang Yu,

Xuejie Liu

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 10, 2025

InSAR monitoring technology is widely used in investigating landslide hazards. Leveraging object detection algorithms to quickly extract information from Wide-Area measurements of great significance. Our InSAR-YOLOv8, an algorithm that automatically detects landslides measurements, addresses the low accuracy and suboptimal performance existing network models. In this method, we first design add a head specifically targeting small-scale objects. This improvement enhances model's ability features across different scales strengthens its capability detect varying sizes. We also replace original C2f module with lighter C2f_Faster process more efficiently, making model efficient. Finally, SIoU loss function replaces CIoU improve bounding box regression enhance accuracy. results show proposed achieves 97.41% mAP50, 66.47% mAP50:95, 92.06% F1 score on dataset, while reducing number parameters by 25%. Compared YOLOv8 other advanced models (YOLOvX, Faster R-CNN, etc.), our exhibits distinct advantages possesses wider range potential applications measurement for detection.

Язык: Английский

Процитировано

1
Detection-Driven Gaussian Mixture Probability Hypothesis Density Multi-Target Tracker for Airborne Infrared Platforms DOI Creative Commons
Min Hong, Jiarong Wang, Ming Zhu

и другие.

Sensors, Год журнала: 2025, Номер 25(11), С. 3491 - 3491

Опубликована: Май 31, 2025

Recent advancements in the unmanned aerial vehicle remote sensing field have highlighted effectiveness of infrared sensors detecting and tracking time-sensitive ground targets, particularly within domain early warning surveillance. However, limitations inherent airborne platforms can lead to irregular imaging inadequate textural features. This study presents a multi-object system specifically designed for weak-textured aimed at enhancing detection accuracy stability. Initially, improvements are made YOLOv10 model through incorporation modules such as DSA, c2f_fasterblock, NMSFree, which collectively enhance robustness targets. Subsequently, results employed conjunction with GM-PHD tracking, enabling rapid stable target tracking. The proposed methodology demonstrates 2.3% improvement 3.8% increase recall when assessed using publicly available datasets. Notably, key metric, MOTA, achieves value 90.7%, while IDF1 score reaches 94.6%. findings from experiments indicate that algorithm surpasses current methodologies regarding effectiveness, accuracy, context multi-target tasks, thereby meeting requirements associated tasks.

Язык: Английский

Процитировано

0