Key Technology Of Vision Measurement And It’s Application In The Monitoring Of Switch Rail Expansion Displacement

Vision measurement is the application of computer vision in the field of automatic detection and precision measurement, due to its automation, non-contact, non-destructive, high accuracy and intelligent features, vision measurement has become one of the hot issues in the research field of computer vision, and has been widely used in public security, transportation, military and other fields.In this paper, the key technology of visual measurement is studied and analyzed deeply, in the light of the urgent needs of automatic monitoring for the rail telescopic displacement in rail transit, in order to solve the limitations and disadvantages of manual interpretation for rail telescopic displacement image, it focus on the automatic monitoring method for rail displacement based on vision measurement technology, and constructs of the rail displacement image data set, a large number of experimental data verified the method is effective and can meet the actual demand of rail expansion displacement automatic monitoring. The main work and research results of this paper are as follows:(1) According to the limitations and disadvantages of the manual interpretation for rail telescopic displacement image, firstly this paper proposed an automatic image recognition algorithm based on hierarchical integration gradient. The algorithm adopts hierarchical approach to the target area, the results of the feature matching method for the guidance, combined with the two values and the integral gradient method, and gradually extract the precise region of the interpretation of the target. Based on the interpretation of the gradient integral and extreme value point accurate positioning of feature point positions, combined with reliability test method, the method realizes the automatic interpretation of rail telescopic displacement image, and the validity of the algorithm is proved by experiments.(2) Aim at the exist problems of the traditional feature matching method to deal with detection and location of scale in rail telescopic displacement image which contains complex noise, this paper puts forward a depth convolution neural network scale positioning method based on spatial support. This method is guided by the theory of depth study, and the convolution neural network is the model, based on idea of the bracing. First of all, it uses the integral gradient like regional recommended method for recommendation, combined with fast RCNN model training method, obtains robust detection model, according to the test results of the model, it adopts the target location method based on spatial support, finally, the accurate coordinate position of the scale is obtained, and the effectiveness of the proposed calibration method is proved by experiments.(3) For the scale interpretation problem in reduced quality of rail displacement telescopic image, this paper studies and proposes a robust interpretation method to reduce image quality. In this method, the tilt image is corrected firstly, and uses method which based on digital matching to face the noise influence in scale ulnar region, combined with a prior knowledge to interpret the low quality fuzzy image, and realizes the automatic interpretation of the degraded image, finally summarizes full text method and the validity of the method is proved by experiments.In order to verify theoretical research and technology, this paper collected elevated station of the south Tianjin high-speed railway track turnout monitoring system nearly a year about 10000 images constructed a rail telescopic displacement data sets, developed a data set of tagging software, classification and annotation and interpretation of the data set image is carried out and constructed a ground truth. The methods proposed in this paper were verified on the data set, and reached the project requirements of rail telescopic displacement automatic monitoring.