Research On Particle Filter Algorithm And Its Application In The Fingertip Tracking

In recent years, virtual machine-visio n-based human-co mputer interact ion techno logy has been great ly developed and applied, gesture interaction as a virtual interact ive techno logy in an interact ive way important that captures user gestures through the camera, for gesture recognit ion, according to Different gestures execute the appropriate command and the results back to the user, in this way makes it an interact ive process to rid itself of the shackles of touch sensors and equipment, great ly enhance the natural sense of human-co mputer interaction. Track your fingert ips as a machine based on the realizat ion o f complex gesture recognit ion and achieve accuracy and efficiency of the algorithm direct ly affect the accuracy of the co mputer user gesture recognit ion and real-t ime, thereby affect ing the ent ire interaction o f the user experience. Therefore, the study and implementat ion of high-precisio n, hig h efficiency fingert ip tracking algorithm, to promote the development o f the future o f gesture-based interact ive techno logy has very important significance. To so lve this problem, this paper based on the exist ing basis o f the extended Kalman filter algorithm particle on its proposed to improve the design and implementat ion o f a high-precisio n adaptive part icle filter algorithm, and the algorithm is applied to the fingert ips trail, effect ively raising the fingert ip t racking accuracy and t imeliness, the main contents o f this paper are as fo llows:1) For part icle filtering algorithm "part icle degradat ion" and algorithmic real-t ime problem, an adaptive part icle filter algorithm, the extended Kalman filter algorithm is improved, the extended Kalman filter, based on the finite difference method produce recommendat ions priorit y distribut ion function, to some extent, to avo id the "part icle degradation" phenomenon; and resampling particle filter, the introduct ion o f adapt ive sampling to determine the number of part icles in the sample thought, according to est imates by the difference between the predicted value and information to determine the number o f particles in the sample sampling, thus effect ively reducing the average part icle filter sampling time, improve the efficiency of the particle filter algorithm.2) Was proposed based on adaptive part icle filter algorithm fingert ips tracing method by obtaining video informat ion and video frame depth information fro m the Kinect sensor, then its hand region segmentat ion, fingert ip extract ion to obtain a two-dimensio nal posit ion of the fingert ip and direct ion, and finally get Kinect depth information based on their three-dimensio nal recovery, get the three-dimensio nal spat ial informatio n fingert ips, finally fingert ip mot ion model constructed on this basis, the use of adapt ive part icle filter algorithm proposed its mot ion tracking, predict their locat ion status, the realizat ion of the fingertip tracking.3) On the basis o f this paper and design tracking algorithm based on adapt ive particle filter fingert ips, based Kinect camera, Opencv and directshow techno logy to co mplete the design and implementat ion of a single fingert ip tracking system, through the actual running tests to verify the art icle fingert ip tracking algorithm design and algorithms in real-t ime tracking accuracy with respect to the conventional Kalman filter and extended Kalman filter algorithm fingert ips particle tracking has been significant ly improved.