Tunnel Image Mosaic Method Based On Geometric Calibration And Feature Detection

With the rapid development of China's transportation industry,the number of tunnels is increasing,and the detection of tunnels is an important project to ensure the safe operation of tunnels.Therefore,the tunnel disease detection technology is paid more and more attention.The current method of tunnel disease detection is based on the panoramic view of the tunnel.The method is to install a row of cameras in the direction of the vehicle in the direction of the vehicle and the camera needs to be straddled in the direction of the road which takes pictures of the tunnel at different times.And then spliced into a tunnel panorama.However,due to the characteristics of the tunnel itself,most of the collected images are with few features,and distinction is not high.Therefore the conventional feature matching method is difficult to achieve splicing and the method of splicing the tunnel image is worthy of further study.In order to overcome the difficulty of splicing caused by the characteristics of the tunnel image,this paper proposes a tunnel image splicing method.Firstly,all the tunnel images are concatenated by geometric calibration and feature detection.Then,for the non-characteristic image,the geometric relationship between the detected tunnel and the calibrated detection system is used to reduce the error caused by driving bumps through the inertial navigation system data,and the adjacent area of the adjacent image is calculated.The image coincides with the number of pixels superimposed and the splicing,so as to complete the tunnel geometric splicing panorama;For the feature image,the improved SIFT algorithm is used to detect the feature.Tunnel detection system Camera auxiliary light source is used in high brightness parallel light LED lights,shooting out the middle of the tunnel image area bright,dark areas around the direct splicing will be detrimental to the late tunnel disease to find and identify.Therefore,this paper improves the image fusion method in the original SIFT algorithm,and increases the gray value of the image in the coincidence region after splicing.It can smooth the splicing area.Finally,the result of the feature detection method is overlaid on the geometric splice panorama,and the geometric calibration method is replaced by the result of splicing the characteristic tunnel image to form the tunnel panorama.In this way,thereare tunnel images of the image stitching exactly,no tunnel disease images can be proportional to restore the actual tunnel.Combined with the actual examples to verify the proposed method of this paper is practical and effective,to meet the engineering requirements,the proposed splicing method for the late tunnel disease identification to provide support.