Baylab

35mm photography lens probably is a good solution for ancient 2/3'' or 1/2" 1.3 mega pxiel imager, it won't work with current 1/2.5" or 1/2.3" 5MP or 10MP sensor, 1/2.5" 5MP sensor has the same pixel density as a 180MP 35mm full frame sensor ( which haven't been on the market yet, maybe 5- 10 years later?). 20MP should be the limitation for most 35mm photography lens. if you're lucky to get a Canon 5D MKII (which has a 20MP full frame sensor), you need to invest a lots of money on Canon's expensive "L" series lens as well ( you can't expect really sharp image just from their low end zoom lens) . Imager alignment is really trick stuff, in order to get best result, you might need a 6 DOF (degree of freedom, XYZ translation and rotation, yes , even the rotation around Z axis does matter, since most real world lenses are not symmetrical). So for the camera with wide angle lens, the best one probably is the one come with a fixed lens ( the lens is fixed to the imager frame by glue or something else), of course, as long as the manufacturer made thoroughly alignment before sending them to customer. (Why? because the sensor itself is not absolutely parallel to it mounting surface, the variation of this parallelism is larger enough to cause some trouble, and for the lens itself, the optical axis of lens is not coincident with the axis of lens barrel either, the variation could be as larger as half degree )

And as a comparison, here are the test results of a 4.8mm 1/2.5" lens which was develop for a panoramic (with a view angle of 48 deg by 65 degree, so it 6 or 8 such lenses can cover 360degree) camera. ROI in the centre of image. (Cropped ,300% rescaled) ROI at left side of image. (Cropped ,300% rescaled) ROI at right side of image. (Cropped ,300% rescaled) So when you choose a lens for some critical application, you need to do some test yourself, the rated megapixel value ( megapixel, 1.5megapixel, 3 megapixel or 5/10 megapixel) or even MTF ( 100lp/mm 120lp/mm or even 150lp/mm) are not guarantee of sharp image for video surveillance project. And, another issue, some lenses were optimised for visible light, if you try to use them for IR illumination, the image will become softer (even after refocusing).

Cropped and rescaled(300%) ROI in the central of image. Cropped and rescaled(300%)ROI at left side of image. Cropped and rescaled(300%)ROI at right side of image.

FA lens has been the only feasible choice for high definition video camera, not matter for industrial quality control application or general purpose video surveillance application, simply because the common CCTV lens can't provide enough resolving ability(resolution of CCTV lens is around 40-60lp/mm, while the resolution of megapixel FA lens could be as high as 120lp/mm at central area and 80-100lp/mm at the edge area ). However, FA lens , not matter rated as megapixel or 3 mega pixel or even 5/10 megapixel, maybe not an perfect solution for the application which need sharp and vivid colour image, some of them may suffer excessive colour aberration. For most industrial inspection project, such colour aberration won't be a real problem, because most of such kind of application use single colour LED illumintion (red, green, or even IR, actually more than half of industrial cameras are B/W model), so the only drawback of such lens is just a slight drop of resolving ability in the edge area. but for the application which need broad band white light illumination, for example general purpose video surveillance or ANPR, the colour aberration will casue serious problem. Here are some test results of a 5mm 1/2" F1.4 megapixel lens (to my opinion, it probably is the only affordable wide angle FA lens which can generate decent image). I attached this lens to a camera with 1/2.5" CMOS imager and reduced the aperture to F2.0 ( just roughly, since there is no any "clicking" mechanism inside its aperture adjusting ring), then turned the camera left and right to capture some picture with different ROI. ROI located at the centre of image, the ROI area is quite sharp and there is no any noticeable chromatic aberration ROI located at the left side of image, the ROI area suffer significant chromatic aberration. ROI located at the right side of image, the ROI area suffer chromatic aberration too, just a little bit better than left side.

first thing, ask for a detailed specification from your customer (they may have some unreasonable expectations), if they don't have, and don't know what they really need, draw a spec for them. An In car ANPR system could be monster which can cover 4-5 lanes and cost more than 70K euro (really fancy one may use a 16M full frame 35mm CCD sensor and works with a powerful tailored flash) or a cheapie which cost less than 1K and works just a little bit better than a dummy camera. Software, you might need to find some 3rd party ANPR software, most them should come with a database and support some data management function, for example, print/issue ticket... One important issue is the image/video interface of the software, some of them only support 768x576 (actually only 768x288 pixel is used ), some of them may support larger format, that means you can choose some high definition camera. Illumination is another challenge, I don't think there is any built-in illuminatior has a range more than 15 m . you might need a separate illumintor . For the camera, as a gold standard, to get a decent recognition rate, the width of number plate should be more than 120 pixel, that means the picture of 768x288pix can only cover the head area of vehicle, it would be very difficult if not possible to capture the face of driver ( in some country, the driver face is a must for a bookable ticket). Triggering device, most of 3rd party ANPR software should come with video triggering function, however it won't be as reliable as radar or laser and wont' provide reliable speed information. and laser/radar =more cost. Visible light or IR, Colour or B/W are two important issues as well. Colour picture with the illumination of visible light is attractive, however, it may cause some road safety problem and may be banned by authority. Colour at daytime and B/W at night with IR seems a good compromise, but you will have to choose a camera with switchable filter and a REAL IR corrected lens. Mounting adaptor( for camea illuminator and in-car computer), such metal parts may cost every more than the camera ( an analogue camera), since you might have to order the customized one from your local supplier.

720x480 NTSC or 720x576 PAL = D1 resolution i also say D1 is 720x480 NTSC / 720x576. i call it true D1. but 1000s of dvr manufactures also say 704x480 NTSC / 704x576 is also D1 but i know it as 4cif. this is why we see so many images of D1 dvrs with what people call interlacing artifacts on moving objects. its a 4cif. 4cif in its own right is good but needs another component to function each pixel must have data. data known as bits (bitrate).While a 4CIF image can produce a better quality image then a CIF image because the more pixels used to represent an image, the closer the result can resemble the original; it does not necessarily have to be the case. 4CIF does not mention the amount or quality of the data to be displayed in the pixels. These are “empty” pixels. Pixels need to be filled with data. Data is referred to in terms of bits. Accordingly, a CIF image filled with more bits then a 4CIF image can produce better visual results. If the "bits" means the colour depth of digitizer, most of modern chips(ancient model may use 8bit) use 9bit or 10bit resolution. However, even the "raw" data is 9bits or 10 bits, the interface between digitzer and image/video compressor is just 8 bit. the firmware will rescale 10/9bit data to 8 bit one. so 10/9bit models may have marginal superiority over 8bit model (most likely on dynamic range), such difference may only visible for uncompressed (or lossless compressed) video. It is true the 4CIF won't provide 4 time more detail than 1CIF. the reasons are: 1) for moving target, in order to remove the interlacing artifact, one field is dropped for every from, so the effective pixels in each 4CIF picture is only 2CIF. 2)for the horizontal resolution, say, 570TVL, or 700TVL, doesn't means it can resolve 570 or 700 lines with 100% fidelity, actually, for most camera manufacturer such figure just means the contrast at 570/700TVL could as low as 10% . so the detail in 4CIF may be even less than 2CIF. I would say, a 4CIF picture may carry 50% more detail than CIF picture. Both 4CIF and CIF video will lost some detail during compression process, and the loss in CIF video may be more significant.

It depend on the definition of "blur", if you want to catch a readable number plate, the displacement of the vehicle IN THE IMAGE ( not along its moving direction, except the optical axis of your camera is perpendicular to moving direction) should be smaller than 10mm. Roughly, 30MPH equal to 13.4m/s, if the angle between camera and vehicle moving direction is 20 degree. the speed of vehicle image is around 4.8m/s, so in order to get a readable image of number plate, a shutter speed of 1/480 sec. is the maximum ( only for the angle of 20deg, if you use different one, you might need to do more calculation).

This post remind me about 20 years ago, When I still in a college (locates in the north of China), when you try to send a message by telegraph, they would charge you by the number of words...

There is some standard size of filter for example: M37, M49, M62. However, not every lens support screw-on filter, some low end lenses just come without any thread for filter. but, as long as the lens come with a thread, it should be a standard size(it makes no sense to make a non standard thread for any lens manufacturer) . for the exact size, if it isn't specified in their manual, you can try to measure it yourself.

I don't think there is any thing wrong. look at the rod against the wall, it is sharp enough, so the lens and focusing should be all right. Of course, there is excessive noise and false colour in the image ( for a good illumination like this), it could be caused by high AGC in camera setting. If you mean the "double image" of the truck, it was caused by the interlacing nature of CCD, it is impossible to overcome it without losing another field in one frame.

For any ordinary optical glass, the spectrum responses at visible and near IR band are quite similar, that why some blocking coating is needed to get good colour reproduction. You can expect marginal improvement when you choose the megapixel rated M12 lens. such improvement might be unnoticeable from naked eye. A real visible / IR lens should be optimized for FOCUSING （not just the spectrum response), that means, when you switch from visible light to IR illumination , there should be no noticeable focusing shift.

There is some kind of "Day & Night" filter, which can block some Infra red except a window at 850nm (with a window width of 30-50nm). Most of low end day and night camera use this kind of filter. With such a filter, you don't need any switching mechanism, however, since live vegetation has extraordinary high reflectivity to infra red radiation, you can't get good colour reproduction in under bright sun light for tree and grass. Therefore, if you really need high quality image/video for day and night, you might have to use a mechanical filter switching module.

try to search with keyword "wiki elphel", you might find some open source stuff. and you can find the source code from their sourceforge website as well

it seems HDCCTV will cost more in cabling: it is possible to transfer severally channel of analog video signal through single co-axial cable, it is also possible to transfer the video stream from several HD IP camera through single CAT5 cable. Is it possible to transfer more than one channel HDCCTV video through single co-axial cable under existing frame? if not, it will be quite costly to expand the system in the future.

Let's back to the camera itself In order to get good result from an megapixel IP camera (actually, all most any kind of camera), decent optics is absolutely necessary. The optics with this ACTi camera suffers at least: 1) Severe astigmatism at corner and edge area, A decent lens should be astigmatism free, and the resolution at at the corner area should not be worse than 50% of central area. for this lens. 2) Not fully corrected for IR illumination. Sure, it is quite difficult to find optics which can produce sharp image in the broad band from visible to IR. You may need put more element inside the lens barrel, and some elements have to be made from expensive low dispersion glass. Therefore it is nearly impossible to find the "affordable" well corrected optics to match this low cost IP megapixel camera.