Journal of Modern Technology and Engineering

Single object tracking is widely used in real-time computer vision systems due to its effectiveness in applications such as surveillance, human-computer interaction, and robotics. However, when dealing with occlusions, conventional trackers of this type often require pretraining which can be time-consuming and computationally expensive. To address this issue, a shape-adaptive colour-based single object tracker capable of recovering from occlusion without relying on deep learning techniques is developed. The proposed framework builds upon established colour segmentation, blob analysis, and motion prediction techniques while introducing first-frame shape bootstrapping, shape-adaptive candidate scoring, confidence-gated appearance updating, and a lightweight post-occlusion rescue mechanism. The target shape is automatically determined from the first frame and the candidate scoring model is subsequently adapted using circularity, rectangularity, vertex count, edge anisotropy, and dominant edge-direction peaks. The proposed tracker was evaluated on synthetic stress-test scenarios and representative sequences from the OTB50 benchmark and compared with CAMShift and Staple trackers using standard tracking metrics. Experimental results demonstrate reliable post-occlusion recovery, improvements over the classical CAMShift tracker, and competitive performance relative to the Staple tracker while maintaining computational efficiency.