Technologies & Tools:
ASP.NET MVC, ASP.NET Web API , .NET Framework 4.5, Entity Framework, JQuery, Materialize CSS, MS VS2015,MS SQL Server, Android Studio, Gradle, Webrtc
The Australian division of an international chain of convenience stores.
Most major retail chains face the problem of tracking the salespeople’s activity. The larger a retail store is, the more difficult it becomes for the manager to keep an eye on each employee and control their timely presence and due performance. Our client was no exception; the chain representatives reached us to discuss whether it is possible to increase the efficiency of employee monitoring and reduce the direct involvement of managers in the process.
Upon scrutinizing the issue, we found out the chain stores were equipped with CCTV cameras for tracking employees and spotting the unauthorized activity. However, the store managers themselves admitted the system functionality was limited as it required extensive human participation and lacked accuracy in identifying a certain employee’s personality.
Thus, we enhanced the existing solution by developing an image recognition algorithm to automate the process of tracking the time salespeople spend at the cash desks. Additionally, we upgraded to the new system of cameras conducting the incessant video surveillance within the store to help resolve other critical issues, such as shoplifting and employee theft.
First, we abandoned the existing CCTV camera sets the client had utilized before. Instead, we opted for the system-on-chip devices powered by the Snapdragon 410/410E processors, for their strong embedded radio modules and ability to process high-definition multimedia files. Thus we enabled seamless connection between the cameras and the central server, and ensured good quality of the recorded video.
We managed to reach almost fully automated functioning of the system due to the embedded image recognition algorithm. When a camera identifies a new face, it takes a picture and transmits it to the server. The manager who has access to the server marks the picture with an employee’s name and sends it back to the device. Afterwards, when a camera recognizes a certain face, it also records the time at which the face was recognized.
To make the devices more user-friendly, we installed an Android application on them. The application gives managers access to the settings of each device and any recorded video. Besides, it allows setting the following camera parameters:
- recognition accuracy;
- the time intervals between the consecutive camera operations;
- recognition algorithm;
- the limiting angle of face turn;
- video definition and quality, etc.
The application interface allows switching between the Google Maps mode showing the location of stores across the country and floor plan mode displaying the location of cameras within the store. In case a store contains several rooms equipped with cameras, each room is assigned a separate floor plan. The number of rooms is not limited.
We designed the solution from scratch, successfully integrated it with the company’s ERP and passed all the accompanying documents to the client. Today we provide maintenance and technical support for the delivered application.