Vehicle recognition

Vehicle recognition software

Vehicle recognition software to replace traditional systems

Traffic operation, pavement design, transportation planning and other key decisions of transportation agencies rely heavily on collected traffic data. Moreover, many private companies, such as retailers, petrol station networks also have a clear interest in such kind of statistics.

Traditional technologies used in traffic data collection, such as magnetic loops, pneumatic road tubes, and piezoelectric sensors have been on the market since the 1960s. But today advanced video analytics tools based on deep neural network architectures gave video surveillance systems a chance to replace traditional systems as better, more reliable and at the same time cheaper solution.

Car recognition software accepts video, captured from arbitrary viewpoints

One of the most important achievements of modern intelligence traffic surveillance system is a chance to use roadside surveillance cameras as a source. Unlike the previous generation of a video-based vehicle recognition system that needed special cameras placed strictly above the lanes on special permanent constructions, modern recognition technology is tolerant of camera placement. The artificial neural network processes images in a similar way as a human brain does. Therefore, it can process images from arbitrary viewpoints, successfully deal with low resolution or poor lighting conditions. Even a standard camera on a tripod will provide you with footage that analytics can deal with.

Next achievement – manual calibration is no more required. Traditional systems need to have dimensions of each type of vehicle accurately specified for each camera. If a vehicle in a video frame doesn’t fit saved dimensions – it will not be captured or correctly classified. This manual calibration requires specially qualified engineers and usually very costly. The new generation of car recognition software works fully automatic, which means as soon as the software receives a video stream from a highway, it starts recognizing cars. Then if you want to move such a camera from one location to another, to collect another bunch of statistical data, you do this seamlessly without calling for a special engineer’s support.

All these aspects result in clear benefits for the users in terms of system installation and maintenance costs.

Video-based vehicle recognition provides more details about detected vehicle

BitRefine offers a video-based vehicle detection and classification system that takes advantage of recent advances in deep learning techniques. The pre-trained recognition module detects motorcycles, cars, trucks, busses as well as recognizes the exact car’s model based on its unique visual features. This system’s recognition capabilities provide significantly more details about the detected vehicle than other types of state-of-the-of-the-art nonintrusive recognition systems such as radar and microwave systems.

Vehicle recognition accuracy reaches 99% in the daytime. Nighttime and poor video resolution may affect this number; however, using a hi-res sensitive camera together with good IR-illumination can keep accuracy during the nighttime above 90%.

Related Solutions

License Plate Reader Software

Car number plate detection Using optical character recognition, automatic number-plate recognition (ANPR) systems read vehicle registration plates to ...

Car counting software

The main advantage of deep learning car counting software is that it works well even if traffic is very tight. Neural networks reliably distinguish every ...

Vehicle type recognition

Vehicle classification data is critical for the effective operation of transportation systems. With respect to the image-based systems in most ...

Car model recognition

Traffic agencies, law enforcement agencies, as well as private companies have always been greatly interested in systems that could collect vehicles’ make, ...

We’ve got more video recognition to share. Let’s be friends. But you first...