Project Description

Accurate occupancy counts are essential to reduce the energy consumption of commercial buildings. The tried and true method is to use beam break sensors at entrance/exits. While these are fairly accurate in certain situations, they are easily fooled by multiple people passing at the same time. Video plus range sensors, such as the Microsoft Kinect, may be a useful tool to create an advanced people counter.

Project Goals

Develop an advanced people counter using one or more video plus distance (or more) sensors, such as the Microsoft Kinect. The group will deploy one sensor at each entrance/exit of academic west (or other campus building). The group will compare the performance to commercial people counters as well as ground truth data (manual counts).

Constraints

Budget, limitations of Kinect SDK (or openkiect).

Criteria

  • Reliable and accurate people counting on a reasonable budget.
  • Should work in all weather conditions (day (bright)/ night (dark), rain, snow, etc).
  • Robustly handle unusual cases such as large backpacks, carrying boxes, pushing a cart, large groups (e.g., campus tour, fire alarm).

Resources

  • I can help purchase sensors.

Licensing

  • MIT / GPL

Impact

This system will be put into production and used for my research. Front page material for the Bucknell website. Potential for a scholarly publication. Potential startup opportunity.

Point of Contact

Prof. Alan Marchiori, Department of Computer Science

 

Leave a Reply