This study proposes an improved YOLO-GQ model and an optimized A* algorithm, which respectively enhance the efficiency and accuracy of object detection and trajectory planning for mobile grasping robots. By integrating autonomous navigation, detection, and trajectory planning algorithms, a physical prototype was successfully developed, and the grasping experiments were validated. The main contributions are as follows: (1) The YOLOv5 model was improved by incorporating an attention mechanism, MBConv module, and angle loss function. (2) The A* algorithm was optimized using a dynamic weight heuristic function, bidirectional search, and Bezier curve path smoothing, which improved the planning speed and path smoothness. (3) A physical prototype integrating a mobile chassis and robotic arm was built, verifying the feasibility of the pose detection and grasping algorithms, meeting the requirements for mobile grasping tasks.