I decided to write this thread, as there is more and more rumor about Smart IR function on the forum, and it seems to me that most users have a false vision of how it should work and what’s its main porpoise. Unfortunately even such services as IPVM suggest in some of its reviews, that smartIR is supposed to react, as a person or any other object appears in the frame and lower its IR radiance, to a level where an object is properly lit and not overexposed. In reality there is a totally different mechanism responsible for it, not smartIR.
To demonstrate it, I used 2 camera models. The 5442TM-AS with updated firmware and HDW5831R-ZE. The first camera has an AUTO option in IR Light section. The second one has a Smart IR option. Both options work exactly the same.
There are two factors that make SmartIR/Auto function useless for a situation, when a person appears in camera. First is its adaptation speed and second is its brightness span. The IR light brightness in Web Interface is available as a scale from 1%-100%. When converted to exposition values (EV) the lowest brightness (1%) is approximately 2EV lower than brightest level (100%). In other words, at 1% value the IR LED shines 4 times weaker than at 100% value.
And now comes the Inverse-Square Law of Light. Let’s assume a person looks at your camera from an X distance. The IR light (at 100%) lights his/her face. If the person comes 2x closer to the camera, so the distance will be 0.5X, the amount of light on his/her face will quadruplet so the face will be 4x brighter. So to equalize the increase of face brightness, the IR LED has to be dimmed 4 times. Again, if the person halves this distance to 0.25X, its face will be 8x brighter than in first position.
In other words, if a person appears in frame, in a distance of 2m (IR at 100%) and will come as close as 1m (IR at 1%), the IR LED will cope with it. Any further or closer to the camera, the IR LED won’t be able to adjust its brightness level. The same goes for distances between 2m-4m, 3m-6m, 4m-8m and so on.
At this point you might think it’s ok. Sounds good. But it’s not, because in most cases the camera will not react to a person appearing in a frame at further distances. This is because of light metering system, which is averaging the whole frame. It means that it meters the entire scene and adjusts exposition/IR light of the camera to the whole image (I’m not sure Dahua uses exactly this method, as there is no technical specs on this, but it seems it’s actually averaging the whole image rather than certain area). I don’t want to get deeper into the metering methods and how it exactly works, but if there is mostly darkness in frame, and a person appears in a distance, the camera will not adjust its exposure or IR light. Only after person’s face/body fills most of the frame, the camera will start to react. Look at the examples below.
The second factor limiting SmartIR usage is the rate of adaptation to a scene. Look again at the examples above and notice how slow it is. In real life, when a person appears in the image, it lasts from a fraction of a second to a few seconds. Here, both AUTO and SmartIR needs 5 to 10 seconds to adopt.
So what’s the solution? Simply use camera exposure system to do the job. DO NOT LOCK EXPOSURE for night time at 1/30s or so. Also, do not lock gain. If you lock both or any of these, the camera will not be able to adopt to existing conditions. If you want your cam to work at 1/25s exposure, set the range between 0-40ms. If something extremely bright appears in frame the cam will shorten the exposure. It will adopt in a fraction of a second and the available brightness span will range from 1/25s to 1/100000s, which is almost 12EV. If you want to keep GAIN at lower levels, do not lock it at one value. Set a range between 0 and your desired value.
The metering problem also can be solved. Use HLC function. The camera will be more “sensitive” to overexposured areas (even small/distant ones) and will instantly adopt exposure to this area. The higher HLC value you set, the more sensitive the camera will be.
SmartIR function is only usefull to adopt IR light to a static scene the camera is viewing. It is there, so that you don’t have to do it manually. It will not work with fast moving objects like cars, people or animals.
To demonstrate it, I used 2 camera models. The 5442TM-AS with updated firmware and HDW5831R-ZE. The first camera has an AUTO option in IR Light section. The second one has a Smart IR option. Both options work exactly the same.
There are two factors that make SmartIR/Auto function useless for a situation, when a person appears in camera. First is its adaptation speed and second is its brightness span. The IR light brightness in Web Interface is available as a scale from 1%-100%. When converted to exposition values (EV) the lowest brightness (1%) is approximately 2EV lower than brightest level (100%). In other words, at 1% value the IR LED shines 4 times weaker than at 100% value.
And now comes the Inverse-Square Law of Light. Let’s assume a person looks at your camera from an X distance. The IR light (at 100%) lights his/her face. If the person comes 2x closer to the camera, so the distance will be 0.5X, the amount of light on his/her face will quadruplet so the face will be 4x brighter. So to equalize the increase of face brightness, the IR LED has to be dimmed 4 times. Again, if the person halves this distance to 0.25X, its face will be 8x brighter than in first position.
In other words, if a person appears in frame, in a distance of 2m (IR at 100%) and will come as close as 1m (IR at 1%), the IR LED will cope with it. Any further or closer to the camera, the IR LED won’t be able to adjust its brightness level. The same goes for distances between 2m-4m, 3m-6m, 4m-8m and so on.
At this point you might think it’s ok. Sounds good. But it’s not, because in most cases the camera will not react to a person appearing in a frame at further distances. This is because of light metering system, which is averaging the whole frame. It means that it meters the entire scene and adjusts exposition/IR light of the camera to the whole image (I’m not sure Dahua uses exactly this method, as there is no technical specs on this, but it seems it’s actually averaging the whole image rather than certain area). I don’t want to get deeper into the metering methods and how it exactly works, but if there is mostly darkness in frame, and a person appears in a distance, the camera will not adjust its exposure or IR light. Only after person’s face/body fills most of the frame, the camera will start to react. Look at the examples below.
The second factor limiting SmartIR usage is the rate of adaptation to a scene. Look again at the examples above and notice how slow it is. In real life, when a person appears in the image, it lasts from a fraction of a second to a few seconds. Here, both AUTO and SmartIR needs 5 to 10 seconds to adopt.
So what’s the solution? Simply use camera exposure system to do the job. DO NOT LOCK EXPOSURE for night time at 1/30s or so. Also, do not lock gain. If you lock both or any of these, the camera will not be able to adopt to existing conditions. If you want your cam to work at 1/25s exposure, set the range between 0-40ms. If something extremely bright appears in frame the cam will shorten the exposure. It will adopt in a fraction of a second and the available brightness span will range from 1/25s to 1/100000s, which is almost 12EV. If you want to keep GAIN at lower levels, do not lock it at one value. Set a range between 0 and your desired value.
The metering problem also can be solved. Use HLC function. The camera will be more “sensitive” to overexposured areas (even small/distant ones) and will instantly adopt exposure to this area. The higher HLC value you set, the more sensitive the camera will be.
SmartIR function is only usefull to adopt IR light to a static scene the camera is viewing. It is there, so that you don’t have to do it manually. It will not work with fast moving objects like cars, people or animals.
Last edited:
