【摘要】 为了保证大型储罐的正常运行,必须对储罐表面进行定期检查。现有方法通常采用攀附罐体表面的机械设备,借助涡流或漏磁进行缺陷检测,存在安全隐患和损害罐体和等问题。本文提出一种基于无人机视觉的缺陷检测方法,无人机携带相机按规划路径环绕罐体飞行以采集储罐表面图像,通过图像处理算法在线判断储罐表面是否存在缺陷。由于储罐表面的缺陷具有视觉显著性的特征,本文采用简化的Itti视觉显著性算法对缺陷图的显著区域进行提取从而分割出缺陷区域。为了解决非缺陷图像可能出现的误判问题,本文基于图像颜色通道求取显著区域的统计均值,设定阈值后降低了误判率。基于室内模拟储罐的实验结果表明,本文提出的缺陷检测方法具有良好的实时性和准确性。更多还原

【Abstract】 In order to ensure the normal operation of large tanks, the surface of the tank must be inspected regularly. Existing methods usually use mechanical equipment climbing to the surface of the tank and detecting defects by eddy current or magnetic flux leakage, there are problems such as damage to the tank body, as well as safety hazards. In this paper, a defect detection method based on unmanned aerial vehicle vision is proposed. The unmanned aerial vehicle carrying camera collects the surface image of the tank around the tank according to the planned path, and determines whether there is a defect on the surface of the tank by the image processing algorithm. Because the surface defects of the tank are visually significant, the simplified Itti visual saliency algorithm is adopted to extract the significant area of the defect image to segment the defect area. In order to solve the misjudgment problem that may occur in the non-defective images, this paper calculates the statistical mean of the salient regions based on the image color channel, and sets the threshold to greatly reduce the false-positive rate. The experiment results based on indoor simulated storage tank show that the proposed defect detection method has good real-time applicability and accuracy.更多还原