YOLOv8-LCNET: An Improved YOLOv8 Automatic Crater Detection Algorithm and Application in the Chang’e-6 Landing Area

Sensors, Год журнала: 2025, Номер 25(1), С. 243 - 243

Опубликована: Янв. 3, 2025

The Chang’e-6 (CE-6) landing area on the far side of Moon is located in southern part Apollo basin within South Pole–Aitken (SPA) basin. statistical analysis impact craters this region crucial for ensuring a safe and supporting geological research. Aiming at existing crater identification problems such as complex background, low accuracy, high computational costs, an efficient automatic detection model named YOLOv8-LCNET (YOLOv8-Lunar Crater Net) based YOLOv8 network proposed. first incorporated Partial Self-Attention (PSA) mechanism end Backbone, allowing to enhance global perception reduce missed detections with cost. Then, Gather-and-Distribute (GD) was integrated into Neck, enabling fully fuse multi-level feature information capture information, enhancing model’s ability detect various sizes. experimental results showed that performs well task, achieving 87.7% Precision, 84.3% Recall, 92% AP, which were 24.7%, 32.7%, 37.3% higher than original model. improved then used CE-6 (246 km × 135 km, DOM resolution 3 m/pixel), resulting total 770,671 craters, ranging from 13 m 19,882 diameter. catalogue has provided critical support site selection characterization mission lays foundation future lunar studies.

Язык: Английский

---

A systematic literature review of quantum object detection and recognition: research trend, datasets, topics and methods

Intelligent Systems with Applications, Год журнала: 2025, Номер 26, С. 200499 - 200499

Опубликована: Март 17, 2025

Язык: Английский

---

REFNet: reparameterized feature enhancement and fusion network for underwater blur target recognition

Robotica, Год журнала: 2025, Номер unknown, С. 1 - 18

Опубликована: Май 16, 2025

Abstract The underwater target detection is affected by image blurring caused suspended particles in water bodies and light scattering effects. To tackle this issue, paper proposes a reparameterized feature enhancement fusion network for blur object recognition (REFNet). First, the gathering (REG) module, which designed to enhance performance of backbone network. This module integrates concepts reparameterization global response normalization network’s extraction capabilities, addressing challenge posed blurriness. Next, cross-channel information (CIF) neck combines detailed from shallow features with semantic deeper layers, mitigating loss detail blurring. Additionally, replace CIoU function Shape-IoU improves localization accuracy, difficulty accurately locating bounding boxes blurry images. Experimental results indicate that REFNet achieves superior compared state-of-the-art methods, as evidenced higher mAP scores on robot professional competitionand objects datasets. surpasses YOLOv8 approximately 1.5% $mAP_{50:95}$ URPC dataset about 1.3% DUO dataset. achieved without significantly increasing model’s parameters or computational load. approach enhances precision challenging environments.

Язык: Английский

---

Less is more: A semi-supervised fine-grained object detection for satellite video

Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 139, С. 109609 - 109609

Опубликована: Ноя. 18, 2024

Язык: Английский

---