It might not seem that obvious to an average car user, but license plate recognition (LPR) technology is now more prevalent than ever before.
It might not seem that obvious to an average car user, but license plate recognition (LPR) technology is now more prevalent than ever before. With the development of technology, it has become an economical and robust solution to several requirements like access control and policing activities.
However, many are still unaware of how LPR (or Automated LPR) actually functions. Bob Mesnik, President of the US-bases systems integrator Kintronics points out, in a recent article, that the key elements that go into an ALPR system are an analytic software that converts images to computer characters and a good IP camera. The analytic conversion software is similar to optical character recognition (OCR) software. “ALPR is available as a standalone LPR system or included as an option with some IP camera video management systems,” Mesnik said.
Parts of the ALPR system
The first component of the license plate recognition solution is the IP camera itself. It is important that the camera that is used for this should be able to capture excellent quality images, as the work of analytics software are heavily dependent on this. The resolution should be at the right level regardless of day or night.
Speaking of day/night operation, it is critical to note that most IP cameras will work well during the day time, but it is at the night time that they have a challenge.
“If there isn’t standard lighting available, IR illumination can be added,” Mesnik said “This can be a separate light or provided by an IP camera that has a built-in IR illuminator. Many license plates have reflective paint that enhances the number when illuminated by IR. The camera should also support wide dynamic range (WDR) to handle the extreme lighting that may be present.”
Then there are special requirements for specific purposes. For instance, if the camera needs to capture license plates from oncoming vehicles, there should be a provision to block out the headlight glare.
The speed of the vehicle is also a decisive factor when it comes to achieving high-quality images. The faster the movement, the more difficult it is to capture and process the plates. This is where the frame-rate and shutter speed of the camera comes into play. To function efficiently, the IP camera should have a fast frame rate and short shutter speed. According to Mesnik, in many cases, a shutter speed of less than 1/1000 of a second is required to prevent blurring.
The placement of the camera is also important. The installers should make sure that it is not placed in a position to views the license plate at an extreme angle. In most cases, the camera captures one lane at a time, with the size of the lane being determined by the camera resolution. “The wider the field of view, the higher, the higher resolution required,” Mesnik said.
Camera resolution
The role of resolution is quite obvious. The higher the resolution, the better the chances of clearly capturing the characters on the license plate. To make things easier, many LPR systems will specify the resolution rates for their systems to work properly, usually by pixels across the license plate, pixels across a character or pixels per foot.
“People can recognize a blurry character better than a computer,” Mesnik said. “The computer requires more resolution than the human eye. How much resolution is required? For this analysis, it doesn’t matter how large or small the character is. The right resolution is determined by the number of pixels across a character. The camera’s field of view can be adjusted so that the pixels across the characters are correct.”
The required camera resolution can be defined once the resolution across the license plate is known.
“People can recognize a blurry character better than a computer,” Mesnik said. “The computer requires more resolution than the human eye. How much resolution is required? For this analysis, it doesn’t matter how large or small the character is. The right resolution is determined by the number of pixels across a character. The camera’s field of view can be adjusted so that the pixels across the characters are correct.”
Similarly the type of lens that should be used can also be calculated. The lens will decided the distance from the camera, and installers have to use trigonometry to calculate the angle of the lens needed. IP camera lenses are specified by their “mm” or angle, so selecting the lens is not difficult. For instance, a camera with a 20-degree lens can be placed about 46 ft. away from the point where the camera reads the license plate.
LPR Algorithms
The final component of a license plate recognition solution involves the software that’s used for character recognition. The main technology that’s used here is optical character recognition. OCR engines use mainly two types of algorithms.
“The first technique matches the images,” Mesnik said. “It stores all the pixels and matches the pattern of the image against stored images. The second method extracts features such as lines, curves, and intersects to create a match. In all cases, the license plate image is converted to a coded computer character. This information can be compared to a database or just recorded for later use.”