Abstract: Digital photography has very good application prospect in automatic observation of precipitation phenomenon, in which how to accurately detect the edge of precipitation particle from the digital image is a key technology. Snake model has the ability to merge prior knowledge and image processing algorithms, which can be used to accurately identify the outline of target. Considering the characteristics of raindrop image, an improved method to automatically select the initial contour of Snake model based on target shape heart automatic calibration method is presented, and the iteration processing of the greedy algorithm is used in the detection process. The raindrop detection method based on Snake model can be used to accurately detect the raindrop edge profile and has good stability. Compared with the traditional edge detection operators, the proposed method has better edge detection effect.

Key words: raindrop image, edge detection, Snake model, greedy algorithm

摘要： 数字摄像技术在降水粒子的自动观测中具有非常好的应用前景，如何在数字图像中准确地进行降水粒子的边缘检测是其中的一项关键技术。Snake模型具有很好的融合图像上层知识和底层特征的能力，能够实现目标轮廓的准确定位。结合雨滴图像自身的特点，提出了目标形心的自动标定方法，在此基础上改进了Snake模型初始轮廓点的选取方法，并通过贪婪算法进行迭代处理，实现了基于Snake模型的雨滴边缘检测算法。算法能够准确地对数字图像中的雨滴边缘轮廓进行检测，且具有较好的稳定性。与传统的边缘检测算子相比，该方法对雨滴图像获得了更好的边缘检测效果。

关键词: 雨滴图像, 边缘检测, Snake模型, 贪婪算法