研究目的
To develop a new algorithm for detecting chemical spills from ultraviolet images by combining adaptive thresholding and region filling to address uneven illumination issues.
研究成果
The UV image is more effective than visible images for chemical spill detection due to better distinction from background water. The proposed algorithm combining adaptive thresholding and region filling outperforms Otsu and FCM in handling uneven illumination, reducing false alarms and omissions, and is efficient in time consumption. However, improvements are needed for edge linking.
研究不足
The dataset has limited universality as it only covers low and middle wave conditions; morphological operations may cause edge offset; future work should focus on linking discontinued edge lines.
1：Experimental Design and Method Selection:
The study involved comparing UV and visible images for chemical spill detection, developing a segmentation algorithm using local adaptive thresholding and region filling, and comparing it with Otsu and FCM algorithms.
2：Sample Selection and Data Sources:
10 UV images of xylene spills captured in an outdoor experiment at Qizhen Lake, Zhejiang University, with a resolution of 2016×1296 pixels.
3：List of Experimental Equipment and Materials:
UV camera, visible camera (Sony α6000), ultraviolet narrow bandpass filter, xylene (analytic grade from Kebo company), and a laptop (Intel i5 7600U CPU, 8GB DDR4 memory, Intel HD Graphics 620).
4：0). Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Xylene was released into the lake, images were taken 30 seconds after release using cameras at specific angles and distances, and the algorithm was implemented in MATLAB R2017b.
5：7b. Data Analysis Methods:
5. Data Analysis Methods: Performance evaluated using commission error, omission error, averaged error, and running time, with manual image interpretation for reference.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
