基于水平集模型的电气设备紫外图像放电区域分割

电气设备电晕放电时产生紫外光,紫外成像检测技术通过检测紫外光的方法检测电气设备外绝缘状况，是发现电气设备电晕放电故障隐患的重要手段。图像处理在紫外成像检测技术中有非常重要的作用。基于简化的Mumford-Shah水平集(Level Set)图像分割模型，提出一种新的紫外图像放电区域分割方法，采用该方法对采集到的紫外图像进行图像分割处理，得到了电晕放电区域边界曲线和放电面积，有利于进一步的模式识别。为检验该方法的抗扰性，进行了多次的实验，结果表明该方法可以在不进行滤波预处理的情况下，有效地抑制泊松噪声和高斯噪声引起的干扰。针对现场采集的含有较多噪声的紫外图像，提出改变模型权重的措施，进一步提高了放电区域分割的准确度。该电气设备放电区域分割方法对于不同类型、不同背景的紫外图像具有良好的适应性。

Discharge Area Segmenting of Power Equipment in UV Image Based on Level Set Technology

Ultraviolet (UV) light radiates when corona discharge occurs on the power equipment. UV imaging tests the insulation of power equipment by detecting the ultraviolet light emission, and is an important method to locate the corona discharge. Image processing technique is essential for UV detection. Based on the simplified Mumfoud-Shah Level Set image dispose model, a new segmentation method on discharge area for UV images is proposed. By segmenting UV images using this method, the area and contour of the corona discharge are obtained, and it is useful for further pattern recognition study. To verify the anti-jamming (antinoise) capability of this method, a series of experiments were conducted. The results showed that using this method Poisson noise and Gaussian noise were effectively reduced without filtering precondition. By changing the weights of model parameters as proposed in this paper, the accuracy of discharge area segmentation can be further improved, especially for heavily noised UV images. The new discharge area segmentation method is adaptable to UV images of a wide range of categories and background.