Search and rescue (SAR) operations typically entail fast response times and the need to cover a vast and complicated region. Several sensors are typically needed when using an autonomous robot for auxiliary tasks, and ensuring real-time performance is challenging. This study aims to improve tracking in scenarios with occlusions, deformations, and varying environmental conditions by proposing two novel algorithms for target detection and tracking. These algorithms are initialized by the result of the detection of the YOLOv5 algorithm. Our proposed algorithms incorporate the Kalman Filter into the Kernelized Correlation Filter (KCF) tracking component and Tracking-Learning-Detection (TLD) algorithms to obtain robustness and best accuracy. The performance of these algorithms is tested through challenging sequences and compared with classic algorithms, and the results prove that the algorithms can meet the requirements we need in search and rescue operations.