With the growing adoption of robots across diverse domains spanning production processes and everyday life, investigations into this field have emerged as a prominent focus among scholars. The novelty of this research lies in the integration of a comprehensive system using ROS (Robot Operation System, ROS) in a hazardous chemical parcel sorting scenario. Through integration with the robot's precise environmental perception capabilities, the system enables accurate localization of optimal grasping points for target parcels—an advancement that remains underexplored in existing research on high-risk environment parcel sorting robotics. A prototype verification platform for the sorting robot has been constructed, which achieves sub-millimeter precision in sorting operations via parcel image recognition and spatial positioning algorithms. Experimental validation of the robotic sorting system demonstrates promising performance metrics: 96.31% adaptive recognition rate for heterogeneous parcel imagery with 0.12 s average processing latency per frame, 91.03% overall sorting success rate, and 1.12 s mean handling time per parcel. These outcomes confirm the successful implementation of automated parcel sorting. Notably, robotic deployment in hazardous environments substantially mitigates operational risk, while the proposed visual perception-based methodology exhibits significant efficacy in enhancing sorting efficiency, positional accuracy, and operational stability.