| The design idea of this paper is to extract the objective effect of the booth,residence time and attention degree in real time from video signals which will be used as the basic data of audience attention and effect evaluation model of the booth.Considering the serious lateral occlusion problem,two cameras are used,one is placed on the top of the booth and the other on the side.The top camera is mainly responsible for the detection and tracking of moving targets.The side camera is mainly responsible for face detection and tracking.The core of the whole algorithm is moving target detection and tracking,which is the basic research content of computer vision.Considering that insects can quickly capture moving objects with relatively simple nervous systems,this paper used the principle and characteristics of insect compound eye vision system for reference and designs two motion detection algorithms:the method based on visual electrophysiology principle and the method based on functional simulation of compound eye structure.Both of them use the compound eye structure of non-overlapping visual field to divide the image into hexagon,as the minimum unit of subsequent processing.Based on the principle of visual electrophysiology,a set of Reichardt correlators are used to calculate local motion information around each small eye.In this paper,the length of low-pass filter in the Reichardt correlator is analyzed in detail.Then,based on the space integral idea of the dendritic tree,several groups of local motion detector array masks were selected,each group corresponded to different main directions of motion,and the motion information of different directions in the field of vision was obtained through the space integral.Next,according to the main direction of motion obtained,the motion region is determined through clustering to provide intermediate information expression for motion tracking.Finally,the recognition of simple motion patterns can be realized according to the distribution of motion information in visual field.The simulation method based on structural function used the geometric distribution characteristics of compound eyes to detect the motion by detecting the correlation between the gray level and motion of adjacent small eyes.It takes each hexagon as a single eye,and each single eye is mostly idle.At this time,only color mutation detection,background update and other simple processing are required.After color mutation,single eye entered the activated state,and the matching relation of adjacent single eye in the adjacent frames was used to determine the local motion.Finally,the image is segmented into facets by watershed algorithm,which is integrated with local motion information,and the target tracking is achieved by counting the matching degree between adjacent single eyes in different frames.Through the experiments test,both the method based on the visual electrophysiology principle of compound eye and the simulation method based on the structure function of compound eye can perform the motion detection effectively.The former has the advantages of high computational efficiency,insensitivity to light,anti-random jitter,abundant local motion information and convenience in motion pattern recognition.The latter has the characteristics of anti-local interference and good spatial consistency,and improves the precision of regional edge with the help of watershed results.The computational complexity is low in static background,but high in complex motion,and it is sensitive to parameter selection.With the face detection and tracking algorithm,the detection and tracking of pedestrians and faces in front of the booth are finally realized,and the basic data for the objective evaluation of booth effect such as total number of people,number of people interested and duration of attention can be counted,realizing the design goal.The motion detection and tracking algorithm based on compound eye of insect has certain features and advantages,which is worth of further research. |
PDF Full Text Request