Vehicle Counter and Classifier

Automation in the traffic control and automobile sector is essential to keep the count of all vehicles and monitor real-time traffic flow. For this, many international companies have developed traffic counter devices that count, classify and measure the speed of vehicles passing by.

A vehicle classifier is another popular product that identifies different types of vehicles through video images. As the name suggests, the device helps classify vehicles as per their predefined classes. 

Overview

Our client was a leading intelligent transport system (ITS) provider based in South Africa. They offered end-to-end traffic management solutions to markets in the African continent and had significant experience in providing transport-based solutions.

The client wanted to automate their vehicle counting and classification process and wanted a product that could efficiently keep track of vehicles and aid their classification. Since their existing system was manual, it was more prone to errors and susceptible to the associated problems.

We at Alpha ICT have decades of experience in providing vehicle-related solutions and developing intelligent systems. Therefore, we were successfully able to deliver comprehensive solutions to our clients according to their requirements.

Approach

Our team first conducted a thorough examination of the client’s current solution. It was a basic product that had a lot of scope for improvement. The next step was to highlight the areas that could be improved and include enhanced technology to develop an optimized product.

After an in-depth analysis, we concluded that designing and developing an automated solution is our top priority. We decided to include numerous new and advanced features in the product like error-free counting and classification. We also recognized the need for monitoring vehicles in real-time and therefore decided to base the vehicle counter on image analysis.

We aimed to transform the existing solution by enabling automatic classification based on the previously provided configuration history. We also felt the need to make vehicle logs accessible to users, so that they can be revisited and checked for anomalies.

Now that we had detected the areas of improvement, our next step was to devise cohesive and realistic solutions to resolve the client issues.

Results

We designed a counter and vehicle classifier that could take the video feed from a camera at the counting site. It would then analyse the video feed to detect vehicles passing through regions of interest and provide all details of the vehicle based on configurations. It includes length, width, speed and inferred classification. The video analysis software was on an Intel i5 core PC, or better. Additionally, it marked this information with a date, timestamp and stored it in separate files to be easily accessible for further processing.

The collected data was required to be accessed and analysed regularly. Therefore, the results of the video analysis were stored with derived parameters so that they could be easily analysed and verified, both online and offline. The system was also trainable and could be modified to suit various classification schemes and camera angles. We installed a 30 fps IP camera with HVGA resolution. 

As the product would require a large amount of storage, we established a standalone and a server-based system to be derived from the solution based on the need for deployment. By employing all of these solutions that our experienced team of engineers planned and implemented, we were able to meet several of our client’s specifications for the new product. We were also able to achieve the following –

  • Market leadership through the first indigenously designed counter classifier
  • New market for automated counting and classification
  • Maintain product relevance through enhanced technology
  • Provide our client with a complete technical ownership of the solution

Conclusion

The project required our team to work in coordination and employ a wide range of technologies that functioned on accuracy. It includes image analysis as well as video analysis done with openCV. With the help of our hardware services to develop the video analysis software, in addition to our software services that finalized the product, we were successfully able to build a solution that met all our client’s requirements.

  • Image Analysis
  • Video Analysis 
  • Vehicle Counter
  • Vehicle Classifier
  • openCV
  • 30 fps IP camera with HVGA resolution
  • Intel i5 core (or better) PC with on board proprietary video analysis software
  • Python
  • openCV