一种基于多视图拼接的集装箱残损自动检测方法及系统
By combining optical character recognition and an improved adaptive elliptic Gaussian kernel, splicing errors caused by strong textures on container surfaces are solved, achieving high-precision container damage detection and reducing the false detection rate.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA MERCHANTS HARBOR DIGITAL TECH (LIAONING) CO LTD
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, container damage detection methods based on multi-view stitching are prone to ghosting, misalignment, or artifacts when dealing with strong textures on container surfaces, resulting in inaccurate stitching results. Furthermore, standard Gaussian filtering can excessively blur the edges of damage, increasing the difficulty of detection.
Optical character recognition is used to remove box number regions, an improved adaptive elliptical Gaussian kernel is used for directional smoothing, and a dynamic scale factor is combined to construct feature matching. An adaptive elliptical Gaussian kernel is constructed through the main direction of local texture to perform local directional smoothing, thereby eliminating interfering features and preserving damaged edges.
It effectively avoids splicing ghosting and misalignment, improves detection accuracy, reduces false detection rate, and ensures high geometric consistency of damaged areas and quality of feature points.
Smart Images
Figure CN122175990B_ABST