Computer vision video analytics solution for automatically processing security records
About the Customer
Exadel is a software engineering company that delivers the digital platforms, products, and applications our clients need to run and grow their businesses.
The Strangers Recognition project was born in 2018 to optimize processes for Exadel’s security department. Previously, Exadel employees had to manually inspect all off-hour footage to discover unauthorized access to the office security perimeter. Surveillance records were analyzed post factum during the next working day to detect security incidents. No immediate, instant actions were performed live by security team members when motion in video records is detected. For inspection, a security team member accessed the DVR web-interface and selected the date of records. The portions of the record where motion was detected were highlighted on a visual timeline on the UI. In range after 18:00 and before 9:00, all records with detected motion were visually inspected to ensure the motion only came from employees. This task took up 5-8 billable hours each week, or 20-30 hours monthly and 260-420 hours annually.
The Artificial Intelligence mechanism of the ESR system analyzes all available surveillance records and information, classifies incidents, and notifies a security team member instantly about incidents that require human processing and attention.
The computer vision video processing framework is built with a highly-modifiable independent feature, which makes it usable in almost every computer-vision or video analysis project. The solution can:
- Read and process local files. Video is captured from the office security system
- Decompose record to frames and filter and pre-process frames for high-quality recognition in CompreFace(Exadel’s face recognition system)
- Train the system with images of known persons
- Categorize people found on video record into known and unknown persons
The Strangers Recognition project is an internal project initiated by our group of experts. It was initially designed to meet one of the security department’s vital business needs: detecting security incidents automatically. The group of experts also wanted to develop their expertise in Node.js and video analysis as part of Exadel’s continual efforts to help all employees with professional and personal growth.
The computer vision video analytics framework is a highly modifiable independent feature, which makes it usable on almost every computer-vision and video analysis project. The steps and architecture of analysis and processing can be easily applied for specific goals.
The computer vision solution helped automate security records processing and detect and classify Incidents. This reduced the time that the security department spent on manual incident detection from 4-5 hours weekly to a maximum of 2 hours, while maintaining 99% accuracy.