通风机械仪表盘在复杂背景环境中视觉故障检测与定位研究

Research on Visual Fault Detection and Location of Ventilation Machinery Instrument Panel in Complex Background Environment

摘要:

通风机械仪表盘往往处于复杂的背景环境中,阴影或部分遮挡会在图像中引入不一致的颜色、亮度和纹理变化,使得故障区域与周围环境的对比度下降,导致人工方法难以正确定位故障区域。针对这些问题,设计一种通风机械仪表盘视觉故障检测与定位方法。首先,使用Kinect相机提取通风机械仪表图像,并进行直方图均衡化来调节图像的亮度和色调,增强故障轮廓与背景的局部对比度。然后,利用改进像素相关性分割算法分割图像特征,将图像中的仪表盘区域从复杂背景中提取出来。利用深度学习领域的深度卷积网络,对分割后的仪表盘图像进行故障轮廓检测。最后,计算定位目标(故障轮廓)的质心坐标,将质心位置作为目标点,映射到构建的投影成像空间坐标系中实现对仪表盘显示故障区域的高精度定位。实验结果显示:应用该方法后,故障区域与周围环境的对比度区分显著增强,具有较高的检测和定位精度。

Abstract:

Ventilation mechanical instrument panels are often in complex background environments, where shadows or partial occlusion can introduce inconsistent color, brightness, and texture changes in the image, resulting in a decrease in the contrast between the faulty area and the surrounding environment, making it difficult for traditional methods to accurately locate the faulty area. To address these issues, a visual fault detection and localization method for ventilation machinery instrument panel is designed. First, use Kinect camera to extract the image of ventilation machinery instrument, and carry out Histogram equalization to adjust the brightness and hue of the image, so as to improve the discrimination between fault contour and background. Then, an improved pixel correlation segmentation algorithm is used to segment the image, extracting the dashboard area from the complex background. Using deep convolutional networks in the field of deep learning to detect fault contours in segmented instrument panel images. Finally, calculate the centroid coordinates of the positioning target (fault contour), use the centroid position as the target point, and map it to the constructed projection imaging spatial coordinate system to achieve high-precision positioning of the instrument panel fault area. The experimental results show that after applying this method, the contrast distinction between the fault area and the surrounding environment is significantly enhanced, with high detection and positioning accuracy.