Aiming to address the challenges of limited space, leaf occlusion, and fruit fragility in ridgecultivated strawberry harvesting, a four-wheel drive ( 4WD ) dual-arm picking robot system was developed. The system integrated a 4WD mobile platform, 6-DOF robotic arms, an RGB D camera, an air pump-driven flexible three-finger gripper, and a low-voltage brushless ducted blower, with the blower specifically designed to disperse occluding leaves and expose fruits. A three-stage framework (YOLO v8 seg PCA ICP) was proposed to enable accurate fruit instance segmentation and pose estimation. For trajectory planning, the rapidly-exploring random tree star ( RRT?) algorithm was integrated with Bspline interpolation to generate smooth, vibration-mitigated picking trajectories. Experimental results demonstrated that YOLO v8 seg achieved an mAP@ 0.5 of 92.7% , with 87.5% accuracy for occluded fruits. The overall accuracy of pose estimation reached 71.8% , with an average angular error of 9.7°±6.6°. After B-spline interpolation, the path generated by RRT?exhibited no abrupt variations and effectively mitigated joint vibrations. In greenhouse harvesting trials, the robot achieved an overall harvesting success rate of 86.5% , with 88.1% for non-occluded fruits and 71.4% for leaf-occluded fruits. The average time per fruit was 9. 6 s, and the undamaged picking rate for isolated fruits was 86. 8% . These results demonstrated that the proposed system delivered excellent operational performance in complex ridge cultivation environments, providing a feasible solution for the mechanized harvesting of ridge cultivated strawberries.