Multi-object Detection and Tracking (MODT) Machine Learning Model for Real-Time Video Surveillance Systems

摘要

Recently, video surveillance has garnered considerable attention in various real-time applications. Due to advances in the field of machine learning, numerous techniques have been developed for multi-object detection and tracking (MODT). This paper introduces a new MODT methodology. The proposed method uses an optimal Kalman filtering technique to track the moving objects in video frames. The video clips were converted based on the number of frames into morphological operations using the region growing model. After distinguishing the objects, Kalman filtering was applied for parameter optimization using the probability-based grasshopper algorithm. Using the optimal parameters, the selected objects were tracked in each frame by a similarity measure. Finally, the proposed MODT framework was executed, and the results were assessed. The experiments showed that the MODT framework achieved maximum detection and tracking accuracies of 76.23% and 86.78%, respectively. The results achieved with Kalman filtering in the MODT process are compared with the results of previous studies.