Augmented reality (AR) technology is a technique that overlays virtual digital information onto the real world, enhancing users’ perception of real-world games. In the field of augmented reality, improving players’ cognitive experiences is a critical task. This paper extracts feature spaces from game images, selects the optimal model to fit their changes, and matches the game image data. Using the SIFT algorithm and ORB algorithm, the overall contrast of the images is enhanced. The G-AR algorithm is employed to enhance players’ immersive experience in object grasping within the game. Additionally, the object grasping algorithm is optimized using Reinforcement Learning with Deep Deterministic Q-Network (RGRL) based on DDQN. A Unity scene is constructed, and through a series of scripts and UI interface designs, the overall game scene is designed. Through a questionnaire survey, the correlation between cognitive challenges and immersive experience is analyzed. AR technology and immersive experience. The significance coefficient for cognitive challenge and immersion is 0.4855, and the Pearson correlation coefficient for AR challenge and immersion is 0.7188, with a significance coefficient of 0.0248. The average scores for the metrics in scenes 1–4 are 3.8501, 2.0899, 3.1095, and 3.5531, respectively. Scene 2’s score is below 3, indicating room for improvement in Scene 2.
