Crowd Safety Assessment Based On Fuzzy Reasoning

Crowd analysis is a research hotspot in the field of intelligent monitoring.Which can judge the safety of the crowd by analyzing the extracted characteristics of the crowd.In this thesis,we combine the method of crowd analysis with the theory of fuzzy control,and take the crowd safety assessment method of fuzzy reasoning as the main research content.First of all,it briefly explain the advantages and disadvantages of a single feature to judge the crowd safety status;Next,this thesis consider the crowd safety from the perspective of the distribution status of the crowd,and propose the algorithm to measure the crowd distribution status.Finally,fuzzy inference theory is introduced to construct a crowd safety assessment system with a single feature and a combination of the two features,and a crowd safety assessment method based on fuzzy inference is discussed.The main research work and innovations of the thesis are as follows:1.A algorithm for judging the uniformity of crowd distribution based on entropy is designed.This method first divides the scene into several areas of the same size,calculates the ratio of the number of foregrounds in each small block to the total number of foreground pixels in the entire image;then,combined with the theoretical knowledge of entropy,finds the distribution of pixels in each video sequence Entropy value;Finally,according to the size of the entropy value to determine the uniformity of the crowd distribution in the scene.The simulation experiment is carried out on the video scene with obvious crowd dispersion.By comparing the obtained entropy with the actual scene crowd dispersion analysis,the conclusion that this method can measure the change of crowd distribution in the scene is drawn.2.A crowd safety assessment system based on fuzzy reasoning is built.Firstly,the number of crowd and the distribution of the population,which reflect the safety of crowd,are fuzzed by a certain membership function respectively as input.Then,the fuzzy inference system based on based on a single feature and a combination of two features is constructed respectively.At the same time,the fuzzy rules based on experiments and expert's experience are combined with certain fuzzy inference reasoning methods to carry out reasoning operation on the constructed system.Lastly,the fuzzy results of crowd safety obtained by reasoning is defuzzied,and the accurate values of crowd safety are obtained.Simulation experiments show that the method in this thesis can evaluate the safety of the crowd.At the same time,the results obtained from the single feature and two features are compared with the real values by using the average absolute error and the average relative error.It shows that the combination of the two features is more suitable for the real situation.