I’ve been asked about license plate recognition many times and how can one capture readable plates day & night. The answer is actually counter-intuitive to what most people think.

I’ve tested a few solutions from a free open source solution to costly commercial solutions. Some people want the ability to just be able to read a license plate in a video and some want to be able to record license plate numbers as text along with a picture of the car so I broke it down both ways. License Plate Identification I’ll start with the first as it’s the simplest.

May 29, 2005. High performance Italian License Plate Reader or Automatic Number Plate reader with very low error rate close or zero error. Our IMPS(TM) is a high performance. CARMEN Parking Plate Recognition Software. CARMEN Parking is a special version of CARMEN Automatic Number Plate Recognition technology. And prices on CARMEN Parking. - CARMEN ANPR/LPR software. CARMEN ANPR software ARH Inc. - CARMEN ANPR software engine.

Most cameras are capable of viewing license plates during the day as long as the plate is large enough to read the numbers. Axis recommends that numbers/letters on the plate be at least 15 pixels tall. You can always buy a higher resolution camera and narrower field of view (telephoto) lens to reach that goal. The problem comes at night. Most people have a nice wide view of an area and when they look at a license plate at night, it’s usually not visible because the headlights or brake lights overwhelm the plate number or the reflectivity of the plate makes seeing numbers impossible, like looking at a white empty rectangle. The reason this happens is because the camera is likely setup to look at a wide area.

At night this area is dark. When a car approaches, a very small percentage of the image are the headlights and the camera uses an average exposure to figure out how to expose the image. This causes headlights to overwhelm the camera which can’t react in time or may never react due to the vast field of darkness. Typically all you see for the front of the car is two bright lights with very little evidence of a plate number. Most people’s school of thought is to have as many megapixels as possible to be able to see the plate and still cover a large area to see the activity, but that’s the biggest mistake people make, using a multi-megapixel camera to cover a large area but then fail at night to see plates. To effectively see plates at night you need to have a telephoto lens set in the smallest possible area around the plate.

At most, you should try and capture only the front or rear of the car at about the width of a lane, or about 8′ wide and the narrower, the better. When the headlights come, they will make up the majority of the field of view and the camera will adjust the exposure quicker. Of course some cameras respond quicker to light changes than others so camera choice is very important. Another misconception is they can overcome the headlights using features like wide dynamic range (WDR) and backlight compensation (BLC) and it helps a little, but not enough and not fast enough to process as a car drives.

Sometimes these features hurt because they may introduce additional image noise and ghosting of fast moving objects. Also, WDR and BLC on some cameras is better than others. The other misconception people have is they think that the IR illuminator in the camera is working against them by reflecting the license plate as an empty white rectangle. That’s not true, the more light you have, the faster the camera will be able to react to headlights because the exposure will already be close to what it needs to be.

Also, setting the exposure darker than normal helps as the priority is capturing the reflective plates. The other problem I’ve noticed is people put a camera perpendicular to the street and expect to capture plates. The ideal location is one where the car is coming towards or away from the camera, as perpendicular to the plate as possible. This is for two reason, plates are hard for anyone to read if you are reading them at a steep angle, the second reason is plate coming towards or away for rear plates do not have the side to side motion that creates motion blur, so you can get away with slower shutter speeds. What you need to view plates day or night is; • a camera dedicated for this purpose • a telephoto lens set for a field of view not more than 6-8′ wide as the car and plate passes • enough lighting to, a) reduce noise, b) allow a faster shutter speed, c) to overcome bright head or brake lights • a camera with extremely good low light sensitivity • a second camera to get an overview of the area so you know what the car is and what it’s doing • resolution over 1MP (720P) is not needed. If you feel you need more, your field of view is too large Equipment Camera I spent a lot of time at trade shows talking to different LPR software vendors about what’s the best camera for the job. I was expecting each vendor to have their favorites but I was wrong, they all said the same thing, the (do not confuse this with the Q1604-E).

I can give you all the technical reasons and feature acronyms why this camera works for LPR, but it just does. This is about a camera (street price).

The included lens is throw away because it’s only 2.8-8mm, too wide for most applications. The reason most other cameras like bullets and domes do not work is because they come with lenses that are too wide for reading plates further than 15′ away and at that distance, it’s likely the angle may be too steep. So you need a camera with a lens that can get that 6-8′ field of view say at 25-40′ or more. There may be other brands of cameras that work equally well, but you’ll need a box style camera so you can use it with the long lens. If anyone has positive experience with other cameras, please report back. Lens Next you need a lens that can you can set to the exact field of view that works for you. I would suggest a good quality varifocal that is at least 15-50mm for up to about 60′ away.

We use this Axis/Fujinon We also reached the 50mm limit of the lens to work with one solution, so don’t skimp on the lens. If it’s further than 50-60′, consider the next step up in lens that goes up to 80mm. Housing Next you’ll need a outdoor housing because box cameras by their nature are not weatherproof. The Q1604-E even though it comes with an outdoor housing is not large enough to hold a good quality varifocal telephoto lens.

Axis makes a very nice enclosure, the but can be pricey at just under $300 but is very well made and easy to adjust in the field. Lights Get as much light on the subject as you can afford. We use the Raytec RM100 with a 30 degree light spread and we have had good results. The equivalent from Axis is the and we chose the RayTec only because it was more readily available at the time. Frankly the Q1604 works so well with minimal lighting that I didn’t even realize one day I left it in day mode and was not IR sensitive but still captured my plate.

The images had noise but the plates were still readable. The better the lights, the better the results. Just to show you what a rear plate capture looks like at night with the wrong type of camera and inadequate lighting, in this case, a Mobotix D14 Dome using their 1.3MP B&W night sensor and a 26W fluorescent light. If the car had not applied it’s brakes, we may have had a chance at capturing the plate. Ironically, the camera in front of the car actually can make out a plate because there’s a lot more light on that end, but it’s not easy to read and not suitable for LPR. License Plate Capture While I strive to have cameras that can view a license plate number, sometimes only seeing a plate number is not enough. I want to be able to capture the plate number in an actionable or searchable form.

You can use this to go back and see when a car has driven by and maybe use that information to see if that correlates to a pattern of crimes. This information can be invaluable, especially in a community where a burglars keep coming back. The other use is to keep lists and then set an event based on a match. For example, say there’s someone that’s known to cause trouble, you can have the camera alert you or local law enforcement to their presence. Or you can use it in a parking garage or a gated area to trigger a gate to open if it there’s a plate match. LPR is being used more and more for many reason. In Texas, the tollways bills you based on LPR.

This summer in California, many of the toll roads that use FastTrak are going to be using LPR as an option over the existing transponder system. So whatever the reason you want or need LPR, I’m going share my experience as to what works and what doesn’t. Hopefully this will save you some time experimenting. First you need the proper equipment as I outlined above.

Cheating yourself out of the proper equipment is just going to lead to frustration. We tried with a Mobotix dome and really had very poor results as shown above. Why Mobotix, because they boasted on their site you have so much resolution you can read plates in a gas station that’s several lanes wide. Yes, I fell for it, but for me, even with a narrow field of view and their dedicated night sensor it does not work well.

If the car stops and the camera has some time to adjust you can make out a plate, but making out a plate means I spend 10 minutes on Photoshop enhancing the image to get the plate number. So imagine if a $1,600 camera does not work well for this with a dedicated night sensor, you have to ask yourself, is buying based on specs or perceived value of a high price going to just be a long and risky science experiment. Next you have to realize that the more pixels you have, the faster the frame rate, the harder that LPR software has to work. So believe it or not, most recommend 4 frames per second and VGA resolution. So yes, that cool 1MP camera with 30fps has to be set to a small fraction of it’s capabilities or it’s not going to work. I would imagine if you have a project for a toll road where cars are going 60-100mph, then 4 fps will not work, but I’m able to cover a street with cars going under 40mph without a problem. Actually it’s worse if they go too slow as you’ll get the same plate capture over and over again.

Also, what makes for a pretty picture may distract from plate recognition. Most recommend setting sharpness down as sharpening tends to make noise more pronounced. Also make contrast higher. Also, make the image darker than normal as the plates are so reflective that it helps at night. SOFTWARE The concept of translating an image to text, commonly known as OCR is not a new concept.

It’s been around for years. But advancements in shape recognition and faster character recognition has made the technology more suitable for license plates.

OpenALPR I ran across a free open source project called OpenALPR (). It’s the code you need to read a plate number from an image. I’m not C++ developer so I worked with the author to create a sample Windows program that when passed an image would read the plate and this is more for testing the tech rather being practical.

BTW, this is not limited to Windows and can be compiled for Linux. It does amazingly well. So I wrote a Windows script around it to extract an image at a time from the camera and pass it to this program and that worked pretty well expect for one problem, it was too slow at a little over a frame per second.

The reason is the startup time for the program to start was too slow. But if someone can take this to the next level and pass the images from a camera in a loop inside his code, it would be a pretty good solution. If some developers want to help and make contributions to this open source project, it would allow for continuous improvements that can make this a great solution for those on limited budgets and competition to the commercial products that would have to step up their game. The following is from the sample program I ran on Windows. You give it an image and it returns plate numbers. This helps in developing your own solution.

The image used is not even VGA resolution, but as you can see, the image is little wider than the car, assuming the car is 6′ wide, the width is about 7-8′ wide and the plate is very visible good size for plate recognition and having a high resolution image would not help – C: Users Carl Desktop openalpr>alpr -r “c: Users Carl Desktop openalpr runtime_d ata” sample.jpg plate0: 10 results — Processing Time = 0ms. – MIDLYPH confidence: 90.2613 template_match: 0 – HIDLYPH confidence: 82.9942 template_match: 0 – NIDLYPH confidence: 82.3028 template_match: 0 – MIDLYPB confidence: 82.1119 template_match: 0 – MI0LYPH confidence: 81.8572 template_match: 0 – MIDLYPM confidence: 81.4123 template_match: 0 – MIOLYPH confidence: 81.3235 template_match: 0 – MIQLYPH confidence: 81.2702 template_match: 0 – MIBLYPH confidence: 80.856 template_match: 0 – MIDLYPR confidence: 80.402 template_match: 0 ipConfigure – embedded LPR They are at every trade show in this business and boast the only in-camera app that can do LPR. You basically run a program that installs the app on an Axis Q1604 and all the functionality of capturing plates alongside the image of the car are done without the need for any software on a PC which is a very cool idea. Installation and setup were reasonably straight forward.

You can create a list of plates and you can create an event that triggers on a plate match. After a lot of time and effort, we did the get the product to work, but with caveats. Their recommendation was to have the plate be about 120-150 pixels wide. Since it’s limited to 640×480 resolution and a standard plate is 12″ long, the plate would have to be 1/5th to 1/4th of the image and the widest area to cover is about 4-5′. This can be problematic for standard lane widths in California that are 12′ wide.

When you can only cover a 4-5′ area, and plates are not always in the middle of the car (for example, our Jeep Wrangler has it’s plate on the far left edge of the car), it difficult to cover an entire lane. Their minimum was 100 pixels wide, and that’s approximately 6′ and barely acceptable for a 12′ wide lane, the accuracy dropped dramatically to where it was getting plates wrong over half the time in real world testing. You can angle the camera in a way that the car sweeps across it to cover more horizontal area, but this affect readability too. Here’s some screen shots from the product.

Remember, this is not on a PC but actually running as an app inside the camera. When you connect to the camera’s web interface, you are not greeted by the traditional Axis initial screen but the ipConfigure initial screen that has 4 tabs, Live, Search, Watchlist and Settings. The live view from the camera is on the left and there’s a few second latency from when the car passes by and you see it in ipConfigure Live. The right side of the display contains the most recent plate recognitions and comes about 5-7 seconds later. This includes the date/time, the text it recognized, the area of the image used for character recognition and a thumbnail of the car.

You can see that after several attempts, it almost got it correct. The reason for the wild character recognition is that it attempted to read the writing on the back of the van and in one plate read it interpreted 55 as 55, then as RR, it read the 2 as a Z. All products tended to wander and provide plate captures from random shadows and letters on window stickers, signs on trucks and such.

Clicking on the thumbnail brings up the image it captured. This plate does have a thick frame around it which does lower accuracy in all products tested.

The Watchlist tab is where you enter your list of license plates. They only have provisions for one list while their PC product can have multiple lists. We didn’t find this to be a problem.

There is also a search tab and after the most recent firmware update, I was able to search by plate number using wildcards and worked as expected and was very responsive. The Setting tab has the actual camera settings which puts you in the app settings, but from here you have access to all the camera settings. You setup events in the traditional way but in this case, the events is triggered by the Application. The rule setup is similar to their standard Axis interface, but here it takes trigger from the application when there’s a “Watchlist Hit”.

In our case, we tested it with alarm output. There’s about a 5-10 second delay between the plate entering the area and the alarm out signal being triggered. They are continually improving this as initially it was twice that long. As of this writing, the retail cost for the ipConfigure Embedded LPR App is $1,499. The concept is great, the suitability depends on your needs and expectations.

If you have an Axis Q1604, you can download a trial version to test it out. More information is available on their website. Milestone XProtect LPR The next product tested is Milestone XProtect LPR. Not to long ago, they reduced their requirement for their LPR product and now was able to run on the XProtect Express Edition making it more affordable. This requires that the camera first be setup on a minimum of ($99/camera), for my testing I used their Professional version. Then you purchase & install the LPR components which at the time of this writing costs $300 for the NVR base per server + one LPR camera license at $1,295.

Milestone recommends a PC separate from XProtect for this, not sure why but it worked fine on the same PC as the NVR software. Worked closely with their support to make this happen and they very responsive. Not going to kid you, this is complex software to understand and setup. But once setup it delivers on the promise.

We where able to open the field of view to cover a lane and still get a very high percentage of accurate plate recognitions. Free Download Audio Driver For Windows 7 Ultimate 64 Bit. Sure, Q may read as a O if the person has a license plate frame that covers part of the letter, but other than anomalies like this, it works as advertised. What amazes me the most was that it was able to capture a plate on a car changing lanes at an angle. Because the vehicles can be stopped or moving very slowly we do get a lot of duplicate captures but with more tweaking of the many parameters this can be likely be improved. Here’s some screenshots.

If you have used Milestone XProtect before, this feature adds an LPR tab to the Smart Client. Selecting this tab lets you search for plates. I searched for my Smart Car with Smart Client and found 3 times I entered.

This is when I entered at night, still able to read the plate accurately with my brake lights on. To setup everything up involves first adding the camera to Milestone XProtect. I cleverly called it LPR Cam. You then go to the LPR section to create license plate lists and configure and tune LPR. There’s a few things I did here that helped. One problem I ran into was thinking if you set the camera to VGA and 4 fps, it will translate here, nope. You have to set those options from within the Milestone camera setup.

I strongly recommend no more than VGA resolution. 4 fps works well for us, even cars going 30-40 mph. Also, turn down camera sharpening and kicking up the contrast made a difference. If your goal is to have a list of plate numbers to act or report on, this is where you add them, under License Plate Lists in the menu above. I just created a simple list called “Entry List” and I searched for my car. You need to create one list for each action you want to take. For example, you may want a list to text you on a match and separate list to turn an alarm on based on a match and another list to open a garage door on a match.

You select which event you want to act on if there’s a match at the bottom where it says “Events triggered by list match” Then under LPR Server, you’ll see the server name and camera. Double clicking on this brings this screen up with 4 tabs.

The first tab I left at default. This is the second tab, character height, very important to set this right. In our case, the smallest plates represented about 3.8% so I set it to 3.5% and the largest plates were about 8-9% so I set this to 10%. Using the defaults, I got nothing, so make sure you work on this. You can select an image to help with the settings.

Reading direction I left at defaults and Processing is where you tweak how accurate you want to accept plate numbers and the frame rate. 4 may not seem like much, but it depends on vehicle speed. If they stop, you will get lots of LPR records, 4 per second, so take that into consideration. At the bottom of these tabs is a test button. This brings up the screen below and you just wait and as cars go by, it displays the plate number, the dictionary used, the character height and the confidence factor. This will help you set the parameters in the tab. Make sure under countries, you select the correct dictionary to use. One dictionary is included in the base product and additional dictionaries can be purchased separately.

The default did not work well for us and support had me change it to X_USA. If you want to set an event, after hours of working with support, I figured it out on my own. I guess it’s not that common. To trigger alarm out on the camera, one would think you can set a Hardware Output Event.

The answer is half right. Doing so will not show up as an selectable event for that plate list. You have to setup a Manual Event which I left at the default name of “Manual Event 1” and the output was to trigger the Gate Open event which is the Hardware Output Event. Support was convinced it should work directly, but it would not so I had to experiment. Hope this helps someone out there.

Once you do this, you can go to the plate list and select Manual Event 1 as the event if there’s a plate match. The Hardware Output Event tells the camera to trigger alarm out for a specified amount of time which I set to a few seconds. So how well does this perform? There’s inherent delays in the software, the camera and the gate controllers, but from seat of the pants measurements, from the time you pull up to the gate until the gates start opening is about 2-3 seconds. Using an access control card takes about 1 second. This does use more CPU than just Milestone XProtect. We are running it on a 5 year old low end computer we had available and with LPR running, it can go up to 40-50% with one camera, but a modern day i5 or better would have no trouble with this.

Conclusion LPR means different things to different people but having the right equipment setup the right way makes a big difference. I hope the time I spent in trial & error and what works and what doesn’t helps save you time and makes you successful. I think the open source community is in early stages and with the help of a community can put together an affordable solution someday. Having scoured the trade show floors for the right solution I found commercial solutions are not cheap.

The least expensive I found was about $800 + the NVR license but camera companies advised against it. There are some solutions that were over $5,000 to start for just one camera. Talking to different people at trade shows, Milestone seemed to one of the recommended products. Look forward to hearing your responses as to what you tried, what works, what doesn’t. Arch Evil Riddim Zippyshare.

Just don’t ask me if a certain camera would work for this because it’s all trial and error. Avoid bullets or domes because you will likely not be able to get narrow enough field of view to make this work. Where to Buy To get all the pieces to the puzzle from an authorized Axis and Milestone partner than can help you with the process, try. They have been very helpful in collaborating through the process. That’s an expensive IR illuminator 🙂 Any cost effective solutions? I’m having motion blur issues at night (and even in daylight on a cloudy day).

Are you using this camera with a 25 mm lens for plate captures? I have mine mounted under the eaves of my garage and therefore is shooting across at about a 45 degree angle out into the street. I also have Smart IR turned off — I’m not sure what that buys me by turning that on as running “tests” with it on vs off didn’t really yield much of a difference to the naked eye (at least for me). I run a plate cam on an overgrown street with low ambient light conditions. I was having similar issues until I increased my telephoto so the plate would take no less than 15% of the image frame width. Bump your frame rate up. Your shutter speed looks good and with a larger plate image you’ll capture more light for cleaner captures.

My contrast is at 90%. You’ll barely see a vehicle fly by until you do a motion search. The plate will be there when you do a frame by frame exam. Your mileage may vary but it’s been working better than I expected here. I’ve recently deployed a SPECO CLPR67B4B to capture license plate numbers off vehicles leaving our neighborhood due to a rash of burglaries in our area.

I read your excellent article with much interest and want to express my deep appreciation to you for taking the time to craft such a thorough article as it has helped me overcome some issues I encountered during deployment. Nevertheless, I’m still getting less than acceptable nighttime performance.

Our neighborhood street has no street lights and is in a semi-rural area with low ambient light. During the day, the camera has no problem imaging the plates. At night it’s a mixed bag which I’ve tried to address by adding a Blackdiamond UFLED Illuminator (with less success than expected). I’ve contact my reseller who informs me that I have to “tune” the camera by parking a license plate 15 feet from the camera. At 15′ the plate to camera angle is in excess the camera specs of 8° to 35° (15′ sets the camera at ~45° to the plate.) The camera specs claim the unit will capture plates at 90’+. To get a proper bead on moving plates and set the camera at an acceptable angular Read more ». The world of analog HD is in so much turmoil, not sure if that’s a safe bet.

There’s SDI, HDCVI, HDTVI and they are not compatible with each other and limited to HD at a time UHD (4K) cameras are coming out, always one step behind IP cameras. That is great you already have fiber from the gates, that’s half the battle. For example, we just installed 2 Hikvision Covert IP cameras in our community inside the entry dialer system, I doubt such a camera exists in the HD analog world, so you limit yourself in choices of products. Also powering the cameras with PoE vs. How it’s done in Analog is more cost effect with IP and easier to troubleshoot at the camera level and coax is more sensitive to interference, especially around gates with strong gate motors, lots of underground AC wiring, AC lighting. So why are they pushing for HD analog, because they are installers that only know analog, have been installing analog for years and are not smart enough to learn some networking skills. Find another vendor.

What I did to find vendors is I called the regional offices for a major camera company like Axis Read more ». Not sure about rolling shutter, but the camera works fine as an LPR camera. Have not tried the Messoa IP cameras but really any modern camera should be able to be used for LPR given a telephoto enough lens so you can get a close enough shot of the plate. Here’s two shots of a plate 45′ away from a Hikvision ds-2cd2032-i custom fitted with a 25mm lens This has the proper character height and works well both day and night at high shutter speeds and the IR illuminator carries that far and this is at VGA resolution, the camera is capable of 3MP. This night shot was taken at 1/200th/sec shutter speed, good enough for highway speed captures –. Hello I have this camera with the 25mm lens installed. I set my exposure to 1/200 with the gain to 75 for night time plate captures.

When a car is parked outside my home at night, the plate shows up perfectly. However, when a car drives by anywhere from 20 – 35 MPH, the image is blurred and video playback appears to show a tracing of the vehicle as it drives past. I uploaded an image to show a side by side of a car that drives by next to the parked car: In addition to the screen shots showing the configs, I also have set: WDR is on level 6 Digital Noise Reduction at 100 I’ve tried other Exposure settings with no luck in getting that clear shot of the plates. Any advice on what I need to do in order to get that clear shot when vehicles drive by at night would be greatly appreciated. I made the recommended changes and there is improvement (see link — last set of photos with updated timestamps).

I set the bitrate to 12Mbps, turned off WDR and set exposure to 1/150. I now have a clear shot of the license plate (at night) when parked, but I’m having a heck of a time getting that clear of a shot when a car is in motion at 25 – 30 MPH with 4 fps and an exposure of 1/150. I’ve experimented with other frame rates and the plate remains blurred. This is proving to be difficult to achieve for night plate captures 🙂.

I’ve had some time to make some adjustments but I can’t seem to get it. I updated the above link with screenshots showing my current config along with some sample captures taken from Live View and Playback. I know you’re time is valuable and would appreciate any guidance. I’m not sure what I’m doing wrong, and if getting a clear capture of a plate at night is doable with my particular setup (as I am shooting semi-across instead of head on). On average, cars drive by my house anywhere from 25 to 35 MPH. Thanks for your time. Hi, I am currently using a Axis P-1346 camera with a 9-20mm Fujinon lens.

My camera settings are set as shown below:- Brightness 40 Sharpness 30 Contrast 80 White Balance Fixed Outdoor 1 Wide Dynamic Range DISABLED Dynamic Contrast. I am having a problem with reflective car plates. During the day, particularly in the afternoon, I will have problems recognizing some of the License Plate due to the reflective car plates. I manage to reduce the reflective car plates by disabling the dynamic contrast as shown above. However, once the sun is starting to set, the car plate will be too dark to be seen due to the intense backlight which forces me to enable the dynamic contrast. So here is the summary of my problem:- Turn off Dynamic Contrast during the afternoon to reduce reflective car plate Turn on Dynamic Contrast during the evening to brighten up the scene.

Is there anyway I could some how set the camera so that I could solve both problems using only 1 type of setting? Hope you could understand what I am trying to ask here. Sorry if it sounded confusing, I really can’t express my question nicely.

Hi Thanks for the very informative article. I’m looking at putting in a camera to monitor an open space with a roadway running along one side. I want to view the wider area in decent resolution but also capture license plates as well. From your article it looks like it would be massively processor intensive to run LPR on a whole HD image but I’ve seen HD cameras split into 2 feeds: one viewing a wide area and one viewing a specific target (eg. Could you split the feed from a high definition camera, sectioning off a portion of a wider image to run LPR? I guess this would incur an extra camera licence for my Milestone software, but would be cheaper than running a dedicated camera for LPR. Is your camera recording 4fps constantly or are you using motion detect?

How big is your capture box? I am looking at setting something up for number plate capture only on a street that is 9m (30 feet) wide and has no constant points where vehicles are forced to pass through. Camera position will be about 3m (10 feet) from the gutter.

I was hoping to have one camera do both directions. Traffic speed in the street is generally below 60km/h (35mph) though we do get the odd Fangio. Is one camera and a decent illuminator like the RM100 you use a realistic option or should I be considering overlapping lanes?

What is your opinion of Powerline Ethernet? It would be an awesome idea of my installations but concerned about whether it creates any performance issues. If it is not a problem, could you recommend a particular type? I see a few on Amazon and suspect that Trendnet might be the best. Also, would this option negate the need for an ethernet switch? I have a large home that I rent so I can’t being running a lot of wire and cutting a lot of holes. It is wired for ethernet into every room.

I suspect I could put a POE switch in the main box that is in the middle of the house that goes to everywhere. However, I wonder about the performance due to the number of connection and wire runs.

Any comments would be very much appreciated. I use a pair of Trendnet with the outlet on the front and it’s been working fine. I used to use a wall plug type PoE injector to power the camera piggybacked to the Powerline adapter, but since have gone to two cameras so I put a switch there. Having the outlet on the back of the Powerline adapter gives me a place to plug in a PoE injector or switch, saves me from getting a power strip. This kept me from having to do major drywall work on 20′ ceilings, so worth it for me. The video is smooth.

In contrast, with WiFi and 3MP cameras, the video was not smooth. As for how many cameras, it’s 200Mbps, so faster than 100Mbps from the cameras, so I think you’ll be OK for at least a half dozen 3MP cameras, more if you use lower resolution or lower frame rate. My review was more of LPR rather than LPC. You can do LPC with many cameras, the trick is to make the max shutter speed fast, like 90-250th of a second, making the scene very dark, then use a very narrow IR illuminator aimed at the plate area. The image will be so dark, you will not see anything other than the plate and head/tail light. The plate will be humanly readable, but not likely suitable for LPR.

I have not had good success with Arecont in very low light and not sure what model you are referencing. The P33 domes are awesome, but not0 good for LPR because they don’t have the low light sensitivity or the lens to create a clear enough image of a plate at night. What I would tell you is if you need more than VGA resolution, 640×480, then you are not doing it right for LPR. You need a telephoto lens to make the viewing angle narrow enough to just capture the width of the car or a little wider.

With a 12mm max on the P33, you may be able to do this up to 10-15′. It doesn’t matter if you have Read more ». Sorry, I made a typo – the Axis cam I have is the P1347, not the P3347; overkill for license plate capture, but I got it at a good price. I replaced the standard lens with a 50mm varifocal unit. My Arecont is a AV1310DN with a 25mm lens. I set them both to a max shutter speed of 1/1000th, fitted them with a filter that blocks visible light, and added a RayMax RM100 IR illuminator. Both of these cameras produce human-readable results ~95% of the time, even under less-than-deal conditions (the camera position is set back from the street about 15′ at an angle of 35 degrees or so).

However, I ran some of the pictures through the online OpenALPR demo and the success rate was very low; it only read one plate out of the 5 or so human-readable captures I tried. The problem may not be my cameras themselves but rather the placement; I’m still experimenting. 🙂 Thanks again for all the great information on this site! Hi, I’m trying to get OpenALRP working on my computer. I just need an exe to which I can pass an image and I can get the plate number. Since I am not a C++ developer, it is hard for me to resolve all the errors I get when working with cmake and make.

You said that you created “a sample Windows program that when passed an image would read the plate”. This is exactly what I need. Do you think that you can help me by enumerating in details the steps you took for this? I tried to search on the internet for the exe, but I didn’t find it 🙁 Thanks, Laura. That’s just a License Plate viewing camera.

They misuse the term LPR which means License Plate Recognition, but it’s not doing any kind of recognition, it’s relying on you to do the recognition with your eyes. The 4 LED’s are going to be worthless, claims 6-12m IR range, I change that to feet for all vendors to get a realistic value, so not likely effective for plates at any distance past say 10′. Mounted off to the side of a road, 10′ means your not likely getting a straight enough shot at the plate.

The lens has a good range, 5-50mm. You can consider that as an alternative to spending about $1,300 with Axis, maybe. The images I’ve seen from this are very dark, not sure how LPR software will distinguish with that poor a contrast as the minimum shutter speed is 1/500/sec.

Have you seen this camera work effectively as LPR, meaning LPR software can interpret the plate number? I can tell you that I can get clear legible plate numbers that anyone can read yet LPR software can’t read or gets it wrong because it’s too small. For example with all the ones I tested, if the field Read more ».