Deformation Detection Of Air-rail Structures Based On Machine Learning

The future of the city's population growth,whether in life or travel to give people a great deal of inconvenience.And hanging monorail(air rail traffic)as a new way of transport to a certain extent,to solve people's travel problems.For the safety of people travel,air track structural deformation detection is an urgent need to solve the problem.In this paper,the detection of the deformation of the empty track structure will be a word red line laser and machine vision combined.Although the contours of the camera's field of view can only be constructed at a time,the detection errors obtained in the experiment are in the millimeter-level accuracy and have good accuracy and recall rate for the detection of defects.The main work of the paper has the following two aspects:(1)First with a red line laser in the air box beam beam irradiation,and then use the camera to shoot a word,red line laser light red light.In order to detect the red line,this paper first according to the unique attributes of red,for the picture feature map conversion.The feature graph suppresses these non-red areas while keeping the red pixel attributes as much as possible.In order to eliminate the influence of noise,this paper uses cascade median filter to filter the image.Intercept the region of interest(ROI)of the filtered image.For the region of interest,this paper uses the weighted centerline extraction algorithm to obtain the expression of the box section,the extraction result can not meet the smoothing requirements.So the results of the weighted average are more accurately expressed using the guided filter smoothing.(2)For camera calibration,the world coordinate system maps the image coordinate system.To correct the glossy light emitted by the red line laser,you must use the camera to take an image without a red line,and then use the camera's internal parameters to get the spatial coordinates of each pixel in the image to the camera coordinate system.Keep the board,and then check the grid with a red line laser.Repeat this step several times to get the coordinates of the point on the smooth surface of the camera coordinate system.The parameters of the smoothing equation can be obtained by using the machine learning method.In order to detect the structural deformation,we first extract all the extrapolated points of the three-dimensional coordinates,and then the second-order least squares fit for the neighborhood of each point.In this paper,we can get the three-dimensional coordinates of the track segment.When the fitting error is less than a certain threshold,it can be judged as a defective part.Experiments show that the algorithm used in this paper is effective and feasible for detecting structural defects of air orbit.