Research On Estimation Of Crowd Density And Application In Hospital

With the rapid development of the national economy, people’s social activitieshave become increasingly rich, various large-scale events such as sporting events,cultural activities and shopping promotions held in public places so that more andmore frequent crowd gathered, the crowd scale is also growing, generated great publicsafety risk. Therefore, how to effectively use computer and image processingtechnology, real-time monitoring the activities of people in public places, timely warnthe crowd density anomalies and take appropriate measures, respect to themaintenance public order, protection of the population have very importantsignificance.In this paper, It has analyze and research the current mainstream crowd densityestimation method, and improve the relevant algorithms, and use a population-basedblock density estimation method to estimate population density in hospital applicationscenarios.Using the average background and frame difference method to extracting thecrowd image background, and use the whole histogram threshold method to extractforeground object movement of people after binarization processing. Against to thelow density crowd, using a least-squares curve fitting features and pixel-basedapproach, this method using binary image approach to extract target edge. And countthe pixel area of foreground and the number of edge pixels respectively. Then useleast squares to fit the curve of the number of people with two pixel characteristics.According to the fitting curve to quantitative estimates the population density. Usethis method to estimate low-density populations image have high degree accuracy, butthe estimation of high-density image have much mistake. To solve the problem of thedense crowd density estimation, use a GLCM based SVM method. First, calculatedthe GLCM of image and extract from the GLCM, then put the texture features intoSVM classifier based on the RBF kernel function and use one multi-classclassification method to training. According to the density of the resulting classifier toestimate population density.Against to the requirement of the Real-time monitoring of the number of scenesfor hospital application, this paper uses a block-based method, use a perspectiveprojection camera model to divided image into a number of equal-sized area sub-images. Then the two methods were used for quantitative estimation of thenumber of and qualitative estimate of density, resulting in the density distribution ofthe monitoring area, to monitor the density of the local area and can find anomaliesand accurate positioning. It can provide effective help to protect the medicalenvironment, maintain medical and hospital security.To estimation the crowd density in a hospital outpatient lobby capture fromHangzhou, experiment shows that this method can effectively estimate the number ofpeople, and have high accuracy, and the average time to process each frame image0.197s, can meet the real-time requirements. Therefore, the population density of theblock-based approach can be applied to estimate the actual scene.