摘要
利用图像识别节制闸混凝土表观病害时,受图像增强敏感性差异影响,病害分类结果较粗糙,最终识别结果IOU交并比偏低。因此,提出基于无人机图像增强的节制闸混凝土表观病害识别方法。通过Context Capture软件进行传输,无人机识别采集规避空间,得到病害识别集中采集点,生成增强矩阵;利用拉普拉斯图像增强算法完成滤波处理,获取增强处理算子,得到图像增强函数,输出病害识别增强图像,提取高质量表观病害特征;以该特征为基础设置分类阈值,调整多尺度特征差值,进行长距离回归计算,引入语义融合模块获取精细分类结果,完成节制闸混凝土表观病害识别。实验结果表明,应用目标方法后,不同迭代轮数下的混凝土表观病害识别IOU交并比均较高,不存在重叠异常,具有集中可靠性。
When using image recognition to control the apparent defects of concrete gates,the difference in sensitivity to image enhancement results in rough disease classification,leading to a lower IOU intersection ratio in the final recognition result.Therefore,a method for identifying apparent defects in concrete for regulating gates based on drone image enhancement is proposed.Using Context Capture software for transmission,the drone identifies and collects data to avoid space,obtains centralized collection points for disease identification,and generates an enhancement matrix;Using the Laplacian image enhancement algorithm to complete filtering processing,obtaining enhancement processing operators,obtaining image enhancement functions,outputting disease recognition enhanced images,and extracting high-quality apparent disease features;Based on this feature,a classification threshold is set,multi-scale feature differences are adjusted,long-distance regression calculation is performed,and a semantic fusion module is introduced to obtain fine classification results,completing the identification of concrete surface diseases in the control gate.The experimental results show that after the application of the proposed method,the IOU intersection to union ratio of concrete apparent defects identification under different iteration cycles is relatively high,and there is no overlapping anomaly,indicating centralized reliability.The experimental results show that after the application of the proposed method,the IOU intersection to union ratio of concrete apparent defects identification under different iteration cycles is relatively high,and there is no overlapping anomaly,indicating centralized reliability.
作者
王诗佳
唐福来
Wang Shijia;Tang Fulai(Guangdong Construction Engineering Group Co.,Ltd.,Guangzhou511340,China)
出处
《吉林水利》
2025年第7期12-17,共6页
Jilin Water Resources
关键词
无人机图像增强
节制闸
混凝土
表观病害
病害识别
Drone image enhancement
Control gate
concrete
Apparent diseases
Disease identification
作者简介
王诗佳(1994-),女,汉,湖南南县人,工程师,研究方向为水工施工。
