Recognition And Prediction The Trajectories Of Flyers In Periscope Video

The characteristic of periscope video is that water ripples may cause some effectson the image of the object, such as irregular distortions, or even broken images and soon. This thesis focuses on two calculation methods: one is used to restore the originalimage of target from periscope video, and the other is used to identify the trajectories ofthe moving object. The relative research is as follows:Firstly, the thesis discusses the three characteristics of water ripples: light justbrings about irregular refraction on the waves; the sum of vector light on the watersurface is zero; waves are high-frequent components in the system.Secondly, the thesis studies the algorithm of center localization to design andimplement the algorithm of RGB weighted center localization based on the algorithm ofgray weighted c center localization. At the same time, the thesis verifies the zero meancharacteristics of wave in single-frame images by calculating the center coordinates ofthe moving target in single image.Third, the thesis makes a research on the algorithm of multi-frame imagerecognition based on digital filter, which combines the advantage of flexible and reliableof digital filter and the advantage of small amount of calculation and wide range ofapplications of frame difference algorithm. The thesis uses differential equations withconstant coefficients of digital filter to achieve inter-recursive and the algorithm ofmulti-frame image recognition. The thesis establishes the algorithm of multi-frameimage recognition based on low-pass of digital filter in periscope video to verify thefeasibility of multi-frame recognition algorithm.Finally, the thesis discusses the algorithm of tracking the moving target objects inperiscope video. The thesis combines the algorithm of RGB weighted centerlocalization and multi-frame image recognition in the image sequence to achieve theprecise center coordinates of the moving target object. The thesis also analyzes themethod about trajectory prediction of moving target objects in periscope video andintroduces the method of linear prediction and parabolic prediction.