In contemporary society, individuals increasingly rely on self-directed learning to enhance their skills and knowledge, with dance education receiving particular attention. However, without professional guidance, learners often struggle to master the intricacies of dance movements and to accurately assess the gap between their performance and standard demonstrations. To address this challenge, this study employs the OpenPose human pose estimation algorithm to capture dancers' joint points and compares videos with different background settings. Using this technology, we successfully developed an automated scoring system for dance, focusing on standard movements, strength, and fluidity. This system incorporates principles of manual scoring to ensure that the results align with human aesthetic understanding, thereby preserving the ability to judge dance aesthetics despite reliance on data. By leveraging human pose estimation technology, we can conduct an in-depth analysis of dance movements, enabling learners to compare their performance with standard dance moves and identify learning discrepancies. This system can dissect individual movements and provide detailed explanations of the execution of various body parts, allowing learners to gain a deeper understanding of each dance element and offering methods for improvement. When learners use this scoring system in the context of learning standard dance movements, it can evaluate their dance performance and provide clear suggestions for improvement. Through this research, we also hope to enable learners to create more effective and engaging self-directed learning environments using online platforms, even in the absence of professional guidance.