Aiming at the problem that thermal comfort prediction, which is a complicated nonlinear process, can not be applied to real-time control of air conditioning directly, this paper proposed a thermal comfort prediction model based on the improved Particle Swarm Optimization-Back Propagation (PSO-BP) neural network algorithm. By using PSO algorithm to optimize initial weights and thresholds of BP neural network, the problem that traditional BP algorithm converges slowly and is sensitive to the initial value of the network was improved in this prediction model. Meanwhile, for the standard PSO algorithm prone to premature convergence, weak local search capabilities and other shortcomings, this paper put forward some improvement strategies to further enhance the PSO-BP neural network capabilities. The experimental results show that, the thermal comfort prediction model based on the improved PSO-BP neural network algorithm has faster algorithm converges and higher prediction accuracy than the traditional BP model and standard PSO-BP model.

In order to improve the robustness of visual tracking algorithm when the target appearance changes rapidly, a particle filter tracking algorithm based on adaptive subspace learning was presented in this paper. In the particle filter framework, this paper established a state decision mechanism, chose the appropriate learning method by combining the verdict and the characteristics of the Principal Component Analysis (PCA) subspace and orthogonal subspace. It not only can accurately, stably learn target in low dimensional subspace, but also can quickly learn the change trend of the target appearance. For the occlusion problem, robust estimation techniques were added to avoid the impact of the target state estimation. The experimental results show that the algorithm has strong robustness in the case of illumination change, posture change, and occlusion.

Most existing methods for detecting plane in point cloud cost long operation time, and the result of detection is susceptible to noise. To address these problems, this paper put forward a kind of multi-plane detection algorithm based on geometric statistical characteristics of the point clouds. The proposed method coarsely segmented point clouds according to the change rate of the volume density firstly, then used the Multi-RANSAC to fit planes, at last the authors proposed a new merge-constraint condition to combine and optimize the initial fitted planes. The experimental results show that the method in this paper is easy to realize, can effectively reduce the influence of cumulative noise to the detection results, improve the plane detection accuracy and also greatly reduce the computing time.