Wireless Multimedia Sensor Networks (WMSN), for object tracking, have been used as an emerging technology in different application areas, such as health care, surveillance, and traffic control. In surveillance applications, sensor nodes produce data almost in real-time while tracking the objects in a critical area or monitoring border activities. The generated data is generally treated as big data and stored in NoSQL databases. In this paper, we present a new object tracking approach for surveillance applications developed using a big data model based on graphs and a multilevel fusion. Our approach consists of three main steps: intra-node fusion, inter-node fusion, and object trajectory construction. Intra-node fusion exploits the detection and tracking of objects in each sensor, while inter-node fusion uses spatio-temporal data and neighboring sensors. Then, the fused data of all sensor nodes are combined to construct global trajectories of the detected objects in the monitored area on the WMSN. We implemented a prototype system and evaluated the performance of the proposed approach with both a real dataset and a synthetic dataset. The results of our experiments on the two datasets show that the use of third-level fusion in addition to inter-node and intra-node fusions provides significantly better performance for object tracking in the WMSN applications.