| 引用本文: | 陆红飞,毛涵宇,周 豪,等.基于Canny-Hough的灌溉渠道边界快速检测算法研究[J].灌溉排水学报,2025,44(5):47-56. |
| LU Hongfei,MAO Hanyu,ZHOU Hao,et al.基于Canny-Hough的灌溉渠道边界快速检测算法研究[J].灌溉排水学报,2025,44(5):47-56. |
|
| 摘要: |
| 【目的】构建一种灌溉渠道边界快速检测算法。【方法】针对灌溉渠道边界的识别与检测问题，采用无人机采集句容东山河和盐城伍佑港的渠道影像，结合Canny边缘检测和Hough变换技术提取渠道边界线，提出了一种基于参照直线和垂直距离的边界归类方法，并分别采用线性拟合和二次多项式拟合方法，评估了边界线的拟合精度。【结果】相比原始图片，采用800×400尺寸图片进行边界提取效果最好，检测时间低于1.3 s，且基本能够描绘渠道边界走势；二次多项式拟合效果优于线性拟合，句容河右侧边界拟合时，二次多项式拟合和线性拟合决定系数（R2）分别为0.959 2、0.949 2；尤其是在处理因障碍物而形成的边界噪点时，能够更准确地描绘边界走势。此外，基于参照直线和垂直距离的边界归类方法，能够高效准确地获取渠道边界，二次曲线拟合时R2均超过0.99。【结论】在传统Canny检测和Hough变换基础上，基于参照直线和垂直距离的分组方法能够实现多条边界快速检测，本研究进一步推动了机器视觉算法在灌溉工程管理中的应用。 |
| 关键词: Canny；Hough；灌溉渠道；边界检测；无人机 |
| DOI：10.13522/j.cnki.ggps.2024375 |
| 分类号: |
| 基金项目: |
|
| Applying Canny edge detection and Hough transform algorithms to identify irrigation channel boundaries in irrigation districts |
|
LU Hongfei, MAO Hanyu, ZHOU Hao, ZHEN Bo, ZHONG Yao, YANG Bo
|
|
1. Institute of Farmland Irrigation , CAAS/ Key Lab of Water-saving Irrigation Engineering,
Ministry of Agriculture & Rural Affairs, Xinxiang 453002, China;
2. Jiangsu Vocational College of Agriculture and Forestry, Jurong 212400, China;
3. Jiangsu University, Zhenjiang 212013, China
|
| Abstract: |
| 【Objective】Airborne technologies have been increasingly used in agricultural sectors for various purposes. In this paper, we developed a fast algorithm for accurately detecting irrigation channel boundaries to support intelligent water resource management in irrigation districts. 【Method】Focusing on the boundary recognition of irrigation channels, unmanned aerial vehicles (UAVs) were used to collect high-resolution images of channels in the Dongshan River (Jurong) and Wuyou Port (Yancheng). The Canny edge detection algorithm was applied to extract edges, followed by the Hough transform to identify and segment channel boundary lines. A novel boundary classification method, based on reference lines and vertical distance metrics, was developed to distinguish channel boundaries. Linear and quadratic polynomial fitting techniques were used to evaluate the accuracy and stability of the extracted channel boundaries. 【Result】Resizing the images to 800 × 400 pixels yielded optimal results, reducing detection time to less than 1.3 seconds while maintaining accurate representation of the channel boundaries. Quadratic polynomial fitting was more accurate than linear fitting in estimating the boundaries. For example, when fitting the straight boundary of the Jurong River, the R2 value was 0.959 2 for the quadratic model and 0.949 2 for the linear model. The quadratic fitting was also more robust in handling boundary noise caused by obstacles. The proposed method can effectively extract channel boundaries, with an R2 greater than 0.99 when using the quadratic fitting method.【Conclusion】Combining traditional Canny edge detection with the Hough transform, along with a grouping method based on reference lines and vertical distances, can rapidly and accurately identify boundaries in multiple channel detections. Our results demonstrate the feasibility of using machine vision techniques and airborne images to effectively and automatically monitor irrigation infrastructures. |
| Key words: Canny; Hough; irrigation channels; boundary detection; UAV |
