Research On Moving Target Recognition Method Based On Data Fusion

Intelligent video surveillance is now widely used in security system, and mainly installed in important monitoring area such as warehouses, banks, airports, traffic intersections and so on. Therefore, how to improve the accuracy of moving target detection and recognition in video surveillance system has become the focus of our current research.In order to improve the moving target detection and recognition rate of intelligent monitoring system, this paper has analyzed and researched the existing target detection and recognition technology, then proposed some effective improvement schemes. Meanwhile,data fusion has been used to improve the accuracy of target recognition. The main contents are as follows:An improved two-dimensional OTSU image thresholding segmentation method based on adaptive inertia weight optimized firefly algorithm(IFA) is proposed in this paper which improved the operational efficiency of two-dimensional OTSU thresholding segmentation method under the premise of ensuring segmentation effect. Then, a moving target detection method that combines the improved three-frame difference method based on IFA two-dimensional OTSU thresholding method(ITFD) and optical flow method is proposed so as to obtain relatively complete information of moving objects, which makes the detection process to meet the request of real-time and achieve better target detection effect at the same time. On the basis of the motion area detection, a recognition method of single target object is used in this paper based on the physical characteristics of detected target and HOG feature combined with support vector machine classifier. And "One-against-one" multi-classification method is used in order to achieve a multi-target classification, thus effectively improve the recognition performance. Finally, to solve the misjudgment problem with in single-view target recognition, this paper proposed a target recognition method based on multi-view image decision fusion technology to obtain a more accurate category judgment.In summary, the detection and recognition technology of moving target has beenin-depth studied in this paper. The improved algorithm and implementation strategies of threshold segmentation, moving target detection, target recognition and misjudgment correction have been proposed, so as to improve the recognition accuracy of single feature targets in multiple moving targets domains. Experiments show that the proposed method in this paper improved the target recognition accuracy and robustness when compared with the traditional method, which can obtain accurate detection and recognition rate effectively.At the same time, multi-view data fusion method can effectively correct the error discrimination caused by single-view target recognition and effectively improve the recognition accuracy of the system.