Scholarlink

<p>Confronted with the twin challenges of deteriorating image quality and subpar target detection accuracy in the realm of intelligent ship driving, this research introduces a target detection model that integrates wavelet-based image enhancement techniques with an optimized version of You Only Look Once version 8n (YOLOv8n). This model first uses adaptive color restoration, multi-scale detail enhancement and dual-branch wavelet fusion to improve image quality. This research uses convolutional block attention module, C2f-Faster lightweight module, DySample dynamic upsampling, normalized Wasserstein distance, and Inner-IoU fusion loss function to build a target detection network. Experiments revealed that the maximum PSNR value of the image enhancement model was 19.8. It was 3.2 higher than the second best generative adversarial network fast image enhancement model, with a maximum SSIM value of 0.74. The improved YOLOv8n model could achieve a detection accuracy of 85.9% and a calculation speed of 32&#xa0;ms. In conclusion, the approach suggested in this study can enhance navigation safety and offer effective visual perception solutions for intelligent ship driving.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Robust Maritime Visual Perception Framework via Dual-Branch Wavelet Fusion and Feature-Enhanced YOLOv8n

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • Guan Luo

摘要

Confronted with the twin challenges of deteriorating image quality and subpar target detection accuracy in the realm of intelligent ship driving, this research introduces a target detection model that integrates wavelet-based image enhancement techniques with an optimized version of You Only Look Once version 8n (YOLOv8n). This model first uses adaptive color restoration, multi-scale detail enhancement and dual-branch wavelet fusion to improve image quality. This research uses convolutional block attention module, C2f-Faster lightweight module, DySample dynamic upsampling, normalized Wasserstein distance, and Inner-IoU fusion loss function to build a target detection network. Experiments revealed that the maximum PSNR value of the image enhancement model was 19.8. It was 3.2 higher than the second best generative adversarial network fast image enhancement model, with a maximum SSIM value of 0.74. The improved YOLOv8n model could achieve a detection accuracy of 85.9% and a calculation speed of 32 ms. In conclusion, the approach suggested in this study can enhance navigation safety and offer effective visual perception solutions for intelligent ship driving.