Particle Filter Algorithm And Its Software Library Implementation

Due to development of modern mathematics theories such as statistics, stochastic process, and great innovations of computer science, novel filters (Particle Filter, Unscented Kalinan Filter, Gaussian-Hermite Filter) are proposed and further developed respectively since the latter of 1990', Because of their outstanding performance in precision, adaptability, convenience when compared to traditional filters (say, Kalman Filter), they are successfully applied in fields of signal/image/voice processing, intelligent system, data fusion, fault diagnosis, biology engineering, industrial process control and etc.Algorithm software library techniques are developed with modern object-oriented software engineering, which are used to encapsulate certain algorithms into conviniently usable libraries. However, few filters related implementations are available to provide these modern solutions, which set obstacles for speading and progressing.In this thesis, sampling based filters (mainly Particle Filter) and algorithm software library technology are studied and developed. The main contributions of the paper are summarized as follows:1. Using Google Scholar and Google Trends to investigate particle filters and others filters development situations, and comparatively statistics and analysis are compactly provided.2. Particle Filter, UKF and GHF are synthetically unified in sampling based filters shceme. With different theory bases, they show distinctness in precision, complexity, and extension capability, which are demonstrated by signal processing simulations.3. Apply particle filter to infrared images sequence denoising instead of traditional Kalman filter applictions. For non-linear heat equation based infrared images scenario, upper precision and nonlinear adaptability are demonstated in practical simulation results.4. Aiming at overcome the defects of common Particle Filter (fixed sampling interval, fixed particle number), an adaptive particle filter is proposed to give flexibility in controlling particle number and sampling interval, which can remarkably reduce complexity of fixed particle filter. For instance, In a target...