Automatic Number Plate Recognition, also known as ANPR, describes the process of digitally finding or reading the number plates of vehicles. This technology is becoming increasingly popular and accurate, particularly as intelligent vehicle management is becoming a vital part of many industries.
The ability to identify vehicles through their license plates can be a highly beneficial tool for a broad spectrum of applications. For government bodies through to small businesses and everything in between, there are countless uses for the technology and many different advantages.
While the concept of ANPR is relatively simple to understand, the technology behind Hikvision ANPR is sophisticated and constantly being improved.
The process of ANPR can be broken down into four main steps:
- Reading the image (Locating Potential Plates)
- Locating the plate (False Plate Filtering)
- Locating and Partitioning the plate characters
- Reading the characters
Reading the Image (Locating Potential Plates)
Hikvision’s plate locating algorithm does not rely on the colour of the plate or the style/formatting of the characters on it like we might expect. Instead, it works primarily by looking for the contrast between the colour of the license plate itself, and the colour of the characters on it. A range of other simple variables and factors like the size of a potential license plate may be considered by the algorithm to initially determine if something might be a license plate.
The Hikvision algorithm will examine an image or video feed and highlight anything that matches these conditions. So even if an object only looks like it could be a license plate, it may still be picked up by this initial search.
In the example image shown below, the wheel of a vehicle has been highlighted by Hikvision’s ANPR because of the contrasting black patches on the white rims. A license plate has also been highlighted for the same reason.
Locating the Plate (False Plate Filtering)
Now that the first search turned up a list of possible license plates, the algorithm can filter out the incorrect results. To do this, the Hikvision ANPR algorithm more closely analyses the ‘textural’ details of the results. These details might be attributes like the rough shape of a result, how it changes from frame to frame or any number of other characteristics. Often, these details will each have a certain weighting, which essentially describes how important it is.
For each result, the ANPR algorithm will determine how likely it is to actually be a plate based on its details/attributes, and the weightings of each. If it’s below a certain threshold, the result is disregarded and assumed to not be a license plate.
Locating and Partitioning the Plate Characters
Having found the correct plate, the algorithm fine-tunes its selection in preparation for reading the characters. To do this effectively, it passes on the work to another simple algorithm designed by Hikvision specifically to detect characters. The result of this is shown in the image below.
From here, the process of distinguishing or ‘partitioning’ each individual character is the most complex part of ANPR. In simple terms, character partitioning is used to determine where a character starts and ends. Once a license plate is divided into its individual characters, these are fed to Hikvision’s character reading algorithm.
Uncontrollable factors like digital noise, camera perspective, lighting conditions, etc, can all greatly impact the readability of characters. As such, partitioning is an essential step for feeding the algorithm information it can understand and accurately read.
Hikvision uses three different methods to accurately partition characters: A preliminary rough partitioning, a fine partitioning, and an approach to read bi-layer plates. Bi-layer plates are license plates that have two rows of characters instead of the one row we’re familiar with.
Rough partitioning consists of two techniques: Vertical Projection, and Connected Component Labelling (CLL). In short, vertical projection is a technique that looks at the spaces between characters and uses those to partition each character. CLL uses the principle that the lines of a character are connected to each other, and that lines of separate characters are disconnected from each other.
As an example, ‘W’ has a single connected line with no breaks in it, whereas ‘V V’ has two disconnected/separate lines, with a break between them. The lines in ‘W’ are connected, so it must be a single character. But, the lines in ‘V V’ are disconnected, so they must be two separate characters.
Fine partitioning takes the rough partitioning a step further and deals with issues that rough partitioning can’t address. One of the techniques that Hikvision uses can be described as a kind of advanced “guess and check” method. To reduce cases where rough partitioning fails, the algorithm compares the results of rough partitioning against known examples of characters. A failed partition can be corrected depending on what examples are the best matches.
- Bi-layer plate processing is a key technique that allows Hikvision ANPR to be used globally, as not all plates have characters on the same row. A relatively simple process enables the algorithm to read plate information across more than one row.
In reality, there are quite a few more complex processes that take place in this stage (particularly with fine partitioning). Hikvision can only give away so much information without completely revealing how their algorithm works. The nuanced mechanics of this stage is what makes Hikvision’s ANPR unique and stand out from other generic ANPR algorithms.
Reading the Characters
All of the above steps and the advanced processes within them feed into this final stage. With character partitioning now complete, Hikvision’s character recognition algorithm can finally get to work on reading the plate. Based on an AI machine learning algorithm, Hikvision ANPR can read a range of different languages and has advanced features not seen in other ANPR systems. The image below described the general steps the algorithm takes.
Hikvision ANPR Applications
As mention earlier, Hikvision ANPR technology has a broad spectrum of applications. With uses ranging from monitoring the flow of heavy machinery in industrial sites through to parking enforcement for small businesses. Hikvision ANPR is a truly versatile piece of tech.
Hikvision ANPR cameras are a popular choice for law enforcement and other miscellaneous government bodies. Within Australia, NSW was the first to trial ANPR technology back in 2005. Their ability to autonomously read license plates can be invaluable for sniffing out unregistered or stolen vehicles, disqualified or suspended drivers, vehicles of interest, etc.
In addition to law enforcement, ANPR cameras can be used for traffic control in car parks, highways, streets, and other such thoroughfares. In such applications, these surveillance cameras can deliver valuable and accurate data. This includes the usage of particular roads, areas of high and low congestion, electronic toll management and collection, and the frequency, location and cause of road works.
ANPR can also be used for live entrance and exit management and blacklist or VIP list notifications. This feature can be particularly useful for casinos, airports, hotels, or practically any other entity that might want to identify specific vehicles.
Other organisations that can benefit greatly from Hikvision’s ANPR technology include:
- Maritime ports
- Military facilities
- Government facilities
- Transportation centres
- Industrial and mining sites
At iCam Security Services, we are Diamond Partners with Hikvision Australia. We stock a range of Hikvision ANPR cameras and an ANPR CCTV kit. If you're looking to buy Hikvision with ANPR functionality, these are the products for you.
Hikvision DS-2CD7A26G0/P-IZ DeepInView DarkFighter ANPR 2MP Varifocal Bullet Network Camera
The DS-2CD7A26G0/P-IZ is a Hikvision bullet camera that features DarkFighter technology, making it suitable for ultra low light environments, and is part of the Hikvision DeepInView range. This ANPR camera has a varifocal lens available in the 2.8mm to 12mm or 8mm to 32mm lens size.
Featuring IP67 and IK10 ratings, this security camera is perfect for outdoor environments where harsh weather conditions may be experienced. The DS-2CD7A26G0/P-IZ is an extremely flexible CCTV camera, offering a variety of functions suitable for many different scenarios.
Hikvision DS-2CD4A26FWD-IZS/P ANPR DarkFighter 2MP Varifocal Bullet Network Camera
Similar to the DS-2CD7A26G0/P-IZ, this ANPR camera is a Hikvision bullet camera featuring Hikvision DarkFighter technology, with 2.8mm to 12mm and 8mm to 32mm lens sizes available.
The DS-2CD4A26FWD-IZS/P is suitable for every ANPR need, with a capture rate greater than 99%, and a recognition rate of over 98%. This Hikvision camera offers 2 megapixel resolution and image enhancement technologies. These technologies include 120dB Wide Dynamic Range (WDR), 3D Digital Noise Reduction (3D DNR), defog, auto-iris, and more.
Hikvision DS-2CD4526FWD-IZP 2MP DarkFighter ANPR Varifocal Dome Camera
The DS-2CD4526FWD-IZP is a Hikvision dome camera with a 2.8-12mm lens size and an IR range of up to 40 metres. This Hikvision ANPR camera can recognise license plates from a distance up to 7 metres, and the ANPR algorithm has been optimised for Australian number plates.
The DS-2CD4526FWD-IZP is protected from weather, impact and vandalism with an IP67 and IK10 rating. This camera provides exceptional footage at night with Hikvision DarkFighter technology and can capture license plates in ultra-low light scenarios at speeds of up to 80km/h.
Hikvision 2MP & 6MP 16CH Ultimate ANPR IP CCTV Kit (with 3TB HDD)
This ANPR kit includes:
- 1 x 16 Channel NVR
- 2 x 6MP Panoramic Fisheye Camera
- 1 x 2MP 2.8-12mm ANPR Bullet Camera
- 4 x 6MP 2.8mm Turret Cameras
This selection of high-quality equipment makes this the ultimate Hikvision CCTV system with ANPR functionality.