Automated detection and extraction of highlight moments in football games is critical, yet existing computer vision(CV) approaches predominantly focus on global events and often neglect region-specific event detection. This paper introduces a CV-based framework designed to automatically detect and extract highlight moments specifically within the penalty area in football video analysis. The proposed method leverages the YOLOv12 object detector, pre-trained on MS COCO without additional fine-tuning, combined with Observation-Centric SORT (OC-SORT) to reliably track players and the football. A homographic transformation is calibrated using known field correspondences to generate a normalized top-down view, within which the penalty region is explicitly defined to enable precise spatio-temporal localization of highlight events. When the football first enters the defined penalty region (detected via a rising-edge trigger), our system records a temporal buffer consisting of 6 s prior to and following this event, ensuring contextually complete yet concise highlight clips. Evaluated on a dataset comprising 90-minute broadcast football game video, the proposed framework achieves a precision of 86.5%, recall of 91.2%, and F1-score of 88.8%, while operating at around 20 FPS on a system with an Intel i7-12700 CPU and NVIDIA RTX 3060 GPU. These results highlight the effectiveness and efficiency of our region-specific event detection approach in football game analysis, presenting a viable solution adaptable to highlight extraction tasks across diverse sports video analytics scenarios.