On The Algorithm Of Electronic Image Stabilization For The Helmet Camera Video

Helmet Camera is widely used in mines, railway corridors and other construction sites. It is worn on the head of the maintenance and overhaul personnel to get real-time video for experts’analysis and processing. The blur and jitter phenomenon of video caused by environmental and application scenarios hamper experts’ watching and analysis. However, there are few video image stabilization algorithms special for helmet camera.In this thesis, a helmet camera video image stabilization algorithm is investigated. The main contents of this thesis are listed as follows:(1) The situation and principles of electronic image stabilization is analyzed in details, and then an image stabilization solution especially for the helmet camera video.(2) We experimentally compare the state-of-art motion estimation methods, and select out ORB as the corner detection algorithm in the video sequence of helmet camera. Since helmet camera is always very close to the target, projection transformation is the most suitable model for global motion transformation.(3) We propose a fast fuzzy frame processing algorithm to process the numerous seriously blurred and shake frames in the video sequence of helmet camera. To solve the problem of frame discontinuous, an interpolated frame kalman filter algorithm is proposed.(4) Lastly, we design and implement the prototype of helmet camera video image stabilization on the basis of algorithms mentioned above with Visual Studio2010and Opencv2.4.9.Experimental results show that methods proposed in this thesis achieve the accuracy and real-time of image stabilization.