Domestic agriculture is facing a decline in strawberry cultivation area due to the aging population, necessitating the development of automation technologies for agricultural tasks. Research on agricultural robots has largely focused on clearly defined harvesting tasks. In automating these processes, effective detection technologies play a crucial role. However, because a strawberry consists of multiple components including the fruit, calyx, and stem precise pose estimation of these elements is essential for determining the optimal cutting point. Therefore, in this study we suggested a pose estimation method to provide detection information for strawberry harvesting robots. Our suggested method defines three classes essential for strawberry harvesting (fruit, calyx, and stem) and detects each object using a deep learning model. Subsequently, a matching algorithm is employed to form coherent groups by associating the detected fruit, calyx, and stem that belong to a single strawberry, and pose estimation is performed based on these groups. The evaluation of the detected classes and keypoints was conducted using mean average precision (mAP), keypoint pixel error, and the percentage of detected keypoint (PDK) index. Results showed an average 0.89 mAP@0.5, an average pixel error of 16.00 ± 10.42 for the x-coordinate and 13.18 ± 6.92 for the y-coordinate, and an average 59.33% PDK@0.5. Although some false positives occurred for the calyx and stem in certain images, the proposed method successfully detected most objects and estimated their poses. Results of this study, the proposed approach is expected not only to detect the fruit but also to comprehensively estimate the poses of all strawberry components, thereby contributing to the development of robotic detection systems capable of precise manipulation and efficient harvesting.