The Study Of Identifying And Tracking The Garment Packing Based On Theimproved Camshift Algorithm

Clothing enterprise logistics in the presence of large amounts of video monitoring to avoid stolen goods, and industrial factory is equipped with a large number of garment enterprises probe to ensure factory security.However, the traditional video monitoring system can only be used for later forensics, cannot have the effect of prevention of early warning, a new generation of video surveillance system will introduce computer vision computing, its main function is to automatically identify anomalies and issue a report to the police.Clothing identification and tracking of the wooden case study to prevent the lost packet loss in the process of warehousing and logistics transportation as well as realization of intelligent video surveillance and the integration of RFID technology to realize intelligent monitoring system is of great significance.The main research contents and theoretical innovation points are as follows:Intelligent video monitoring is discussed in the clothing enterprise in the field of logistics, this paper introduces the preliminary image processing technology in the mainstream of filtering algorithm and morphological processing algorithm, do pretreatment for garment packing identification and tracking.Extract the clothing boxes based on color feature vector---BGBRGRBGR],,,,,[, packing color feature vector distribution were analyzed.Using support vector machine(SVM) classification algorithm is implemented for packaging of sample training, training results and use the training results are given for packaging discriminant result of pixels.MeanShift theory is discussed and its tracking algorithm, this paper introduces the Camshift tracking and give the two tracking algorithms under the same target model and probability graph actualization of contrast experiments, through the quantitative analysis of the experiment parameters change in the process of qualitative evaluation of these two kinds of tracking algorithm.Further processing of the SVM recognition results are clothing packing size and location and initialize the tracking target, discussed the process of image binarization threshold of recognition results, the influence of the size of the different scenarios are given the recognition results of different boxes.Kalman filtering theory is introduced.Combined with Camshift algorithm is proposed and an improved algorithm based on Kalman filter :(1) the target is sheltered Camshift algorithm cause the failure of tracking problem, with smaller goals are sheltered area information to determine whether the target is blocked, with the current goal and target model two histogram of coefficient of correlation detection target whether leave keep out area.Packing not obscured by Kalman forecast as a result the output, using Camshift tracking update Kalman filter, the result of the packing cases are occluded with Kalman filter prediction as a follow up on the results, with the result update Kalman filter and detection packing whether leave keep out area.(2) the target by large area of similar color interference Camshift tracking target leakage problems, with goals by color interference target area increases when the information to determine whether the target color disturbance, the number of iterations using Camshift algorithm to detect whether leave color interference target area.Not by color interference with Camshift tracking results update Kalman filter and Kalman prediction of the current frame as a result output;Target color disturbance using Kalman filter prediction as the tracking results, with the result as observed value update Kalman filter, and determine whether to leave color interference area.