Turning Parking Cameras into AI‑Powered Safety Guardians

Background

In February 2021 the "New Content New Interaction" Global Video Cloud Innovation Challenge, co‑hosted by Intel and Alibaba Cloud, attracted over 4,600 participants worldwide. One of the pressing social issues highlighted was the high rate of child injuries in parking‑lot blind spots, accounting for 17.1% of accidents.

Student Project

Liu Baoyin, a senior computer‑science student at Qingdao University, developed a smart parking‑lot system that focuses on eliminating these blind‑spot hazards. By repurposing existing surveillance cameras, his solution captures video streams, processes them intelligently, and provides real‑time parking status and danger alerts without requiring additional hardware.

Technical Solution

The system runs on an Intel Xeon SG1 GPU and utilizes the Intel OpenVINO toolkit together with an optimized Mask R‑CNN instance‑segmentation model. It segments and matches people and vehicles in the video feed, enabling accurate vehicle counting and pedestrian‑danger detection (e.g., children playing near vehicle blind spots).

Beyond safety alerts, the platform can supply public‑parking space information to major map applications, visualize parking occupancy, and integrate with public facilities such as broadcast speakers or warning lights to issue proactive safety signals.

Student Profile and Achievements

Liu, previously a debate champion, has earned multiple AI‑related awards, including first place in the 16th "Challenge Cup" Shandong University competition, a third‑place prize in the 5th Shandong Science‑Innovation Contest, and a bronze medal in a Kaggle data‑mining challenge. He emphasizes a learning approach that starts with official documentation, runs demo samples, and iteratively builds custom solutions.

He has been admitted to the graduate program at Northeastern University, where he plans to continue advancing AI technologies.

Images