Our client is a provider of bespoke, high level systems and safety engineering solutions for a wide variety of industrial sectors. Zircon were subcontracted to take part in a feasibility study for the client and another organisation to develop software that was capable of detecting and classifying a series of specified objects.
Zircon was responsible for taking the clients requirements and determining whether it was possible to develop an effective and successful detection system that could not only identify and classify each object, whilst discarding unimportant details surrounding the target object, but track it over multiple frames. As with all CCTV projects it was desired that the program could function accurately in a variety of different conditions and with a variety of image formats.
In order to identify the best possible solution a wide variety of object detection techniques were investigated and tested using a library of reference images. Following these investigations it was decided that a filtering process was required to reduce the high numbers of false positive results each technique generated. As before in order to determine the most effective process a variety of filtering techniques were examined and tested using the same library of reference images.
Upon the conclusion of all of these investigations it was determined that following the implementation of a filtering process it was indeed possible to identify and track an object over a series of frames, however the effectiveness of the process was directly affected by the quality of the reference images used in the library and the quality of the images being searched.
Even though the concept had been proved possible a number of potential future enhancements with regards to improving the speed, detection and classification and library were provided by the engineers involved in the project.