Hong Kong Cable Car Smart Passenger Counting System

Automatic counting of cable car passenger numbers using camera automation technology, thereby improving data accuracy and processing efficiency. This system uses image recognition and machine learning technology to analyze images to count passengers, which not only helps to optimize crowd management, but also improves safety and operational efficiency.

Introduction

A technical failure of the cable car on January 25, 2012, left many passengers stranded in the air for several hours. The incident raised public concerns about the safety and emergency response capabilities of the cable car system.

To address this situation, we have proposed an innovative monitoring solution. GeoXpert will leverage existing camera infrastructure and use computer vision technology to analyze captured image data in real-time. Our system can identify the license plate number of each cable car and accurately estimate the number of passengers inside using advanced object detection and tracking algorithms. This enables us to quickly and accurately grasp passenger information in the event of a technical failure or other emergency, thereby improving the efficiency and effectiveness of emergency response.

Our solution allows us to effectively utilize existing resources and provide cable car operators with a powerful monitoring and management tool without the need for additional surveillance equipment installation. This not only ensures passenger safety but also strengthens the ability to respond to potential future operational challenges, bringing comprehensive safety and efficiency enhancements to the cable car system.

Key Technical Features

This project's main technical features include using advanced computer vision and machine learning to get accurate counts and manage how many people use the cable car. The details include:

1. Object Detection Technology

Deep learning models empower our system to precisely identify visitors, staff, and cable car license plates in real-time video streams, even under varying lighting or long distances.

2. Real-time Data Processing Accelerated by Deep Learning

With GPU acceleration technology and optimized algorithms, real-time image analysis and data processing are achieved, ensuring real-time updates of passenger statistics and crowd dynamics.

3. Multi-Object Tracking (MOT)

This technology plays a critical role in enhancing the efficiency of crowd dynamics tracking and analysis, especially in the following key areas:

• Precise Object Detection and Tracking: The system can accurately detect every object in the scene and assign it a unique identification ID for real-time tracking. This capability enables the system to precisely track the dynamics of each object in motion, whether in a cable car or a dense crowd, and lock onto and track the movement trajectories of individual targets.
• Continuous passenger flow analysis: By tracking each object in real-time, the system can continuously gather and analyze data on passenger movements, offering deeper insights to enhance safety measures and improve operational efficiency.

Despite the advanced technology, tracking may still be inaccurate in real-world applications due to the following factors:

• Occlusion between people: In dense crowd movement situations, mutual occlusion between individuals may cause the system to mistakenly change the tracking ID of a target, which not only affects the continuity of tracking but also challenges the accuracy of overall crowd counting and analysis.
• Light reflection and reflection interference: Strong light reflection or reflection effects can cause the system to mistakenly identify non-target objects as targets, or cause the original targets to "disappear" and "reappear" in the system, confusing tracking IDs.

To solve the problems above, the system uses a new and better way to filter out short-term anomalies.

Time threshold setting: By setting a specific time threshold, the system can identify and filter out those abnormal targets that only appear for a short period, ensuring that only those targets that appear continuously and exceed this set time threshold will be identified as valid tracking targets. This strategy effectively reduces the misidentification caused by instantaneous occlusion or light changes, and greatly improves the accuracy of system tracking and the reliability of overall crowd flow analysis.

To demonstrate the actual effect of the time threshold setting, we provide a frame-by-frame slow-motion video example: In the video, the region of interest (ROI 1) initially detected four IDs, but there was a short-term misjudgment during which ID:2300 was detected. However, due to the time threshold setting, this misjudged ID was eventually filtered out in the population count, ensuring that the final population count returned to the correct value.

This process not only demonstrates the importance of time threshold setting in improving data accuracy, but also highlights the high flexibility and reliability of our system in handling instantaneous anomalies.

4. Real-time data processing and analysis

The system adopts real-time data processing technology to quickly analyze image data, update passenger count statistics and crowd flow dynamics in real time, and provide real-time data support for operation management.

5. Seamless integration with existing facilities

This solution uses the existing camera of the cable car, without the need to install additional hardware.

Benefits

Through the application of this intelligent crowd-counting system, cable car operations will experience the following significant benefits:

1. Enhanced Safety Management

‍The system significantly enhances the ability to grasp the dynamics of passenger flow by accurately counting and tracking passengers in real-time. This not only ensures passenger safety but also provides solid data support for rapid response in emergencies, effectively improving the overall safety management effectiveness.

2. Optimizing operational efficiency

‍With accurate passenger flow data, the operations management team can formulate more reasonable cable car dispatch plans based on actual needs. This data-driven decision-making process can not only reduce passenger waiting times and improve service quality, but also significantly improve overall passenger satisfaction.

3. Significant cost-effectiveness

‍By utilizing existing surveillance facilities, the system avoids the need to install additional new equipment, thereby reducing the cost of system upgrades. In addition, by optimizing operational processes and improving operational efficiency, the system also helps reduce unnecessary operational costs, achieving a significant improvement in cost-effectiveness.

With the implementation of this intelligent passenger flow statistics system, Cable Car will be able to achieve significant improvements in multiple dimensions, including safety management, operational efficiency, cost control, and passenger experience. This will provide visitors with a safer, more efficient, and more comfortable riding experience, while also bringing more flexible and efficient management capabilities to the operations team.

Demo