Research On Image Recognition And Tracking Algorithm Based On The High-Dimensional Biomimetic Neural Network

Artificial neural networks provide a new access to the research of imageclassification recognition and object tracking. And relative to the traditional artificialneural networks theory, the high-dimensional biomimetic pattern recognition theorycan depict the curve of the sample in the feature space more precisely.Firstly, this thesis studies the existing theory of image recognition and objecttracking. And according to the high-dimensional geometrical methods, thehigh-dimensional biomimetic neural network is applied to the field of imagerecognition and object motion tracking. Secondly, this thesis designs a imageclassification discriminator in use of the high-dimensional biomimetic neurons, andanalyses the correct recognition rate when the gray pixel of image is treated as theinput of the image classification discriminator. A large number of simulations andexperiments show that the image classification discriminator based onhigh-dimensional biomimetic neurons is feasible and the correct recognition rate ishigher. At last ,a new object tracking algorithm based on high-dimensional biomimeticNeural Network is mentioned, this algorithm combine the PCA algorithm with thehigh-dimensional biomimetic neurons. Simulation results show that this algorithm hasbetter robustness and has a good effect in object tracking with complex background.