Another thread in this forum has some talk about the effect of shutter speed on motion blur. I thought I would show my calculations and tests that helped me to determine what limits to place on my camera's shutter speed.
First, I want the camera to be useful for ID of a person walking by. My fast walking pace is less than 15 minutes per mile which is 70 inches per second. I feel that in order to identify a face the blur should be less than 1/2 inch. This indicates that the shutter speed should be 1/140 second or less. Please note that this is independent of the camera resolution or distance to the person. It only matters what the transverse speed is and what motion blur is acceptable.
I then tested this by walking at a fast pace by my front door camera. I wanted some sort of quantitative measure of the motion blur so I held a printed copy of the 1951 USAF resolution test chart. Each group of bars is twice the size of the proceeding group and within a group each element increases in size by 12%. I tried holding the chart so that some bars are parallel to my motion and some are perpendicular.
I tested at 1/30, 1/60, 1/120, and 1/2000 of a second. I walked at a fast pace along the sidewalk transverse to my front door and then turned and walked head-on towards the camera.
All of the pictures are full resolution crops out of a 12 MP Dahua IPC-HDBW81230E-Z dome camera and include both the test chart and a face. If the camera was perfect it would be able to resolve 0.06 inches, based on the pixel spacing in the image. The head-on images come within 50% of ideal at the higher shutter speeds. The resollution of the horizontal bars comes close to the resolution of the head-on motion at the same shutter speed.
BUT the resolution of the vertical bars is terrible (as expected) except at the highest shutter speed. Based on these tests I have set all of my (non LPR) cameras to shutter priority and use 0-5 mS during the day and 0-10 mS at night. Allowing a slower shutter speed at night is a balance between motion blur and noise.
Head-on, 1/30 sec (ignore Post-it that says 1/60)
Transverse, 1/30 second
Head-on, 1/60 second
Transverse, 1/60 second
Head-on, 1/120 second
Transverse, 1/120 second
Head-on, 1/2000 second
Transverse, 1/2000 second
For my LPR cameras, as similar calculation uses a car traveling at 30 mph or 528 inches per second. The camera is at an angle of around 30 degrees. This reduces the transverse motion by a factor of two. During the day I use shutter priority and 0-0.5 mS to keep the motion blur under 1/8 inch. At night I use 1/1000 second for a motion blur of 1/4 inch. No photos, but this is what I arrived at while empirically setting up the LPRs.
First, I want the camera to be useful for ID of a person walking by. My fast walking pace is less than 15 minutes per mile which is 70 inches per second. I feel that in order to identify a face the blur should be less than 1/2 inch. This indicates that the shutter speed should be 1/140 second or less. Please note that this is independent of the camera resolution or distance to the person. It only matters what the transverse speed is and what motion blur is acceptable.
I then tested this by walking at a fast pace by my front door camera. I wanted some sort of quantitative measure of the motion blur so I held a printed copy of the 1951 USAF resolution test chart. Each group of bars is twice the size of the proceeding group and within a group each element increases in size by 12%. I tried holding the chart so that some bars are parallel to my motion and some are perpendicular.
I tested at 1/30, 1/60, 1/120, and 1/2000 of a second. I walked at a fast pace along the sidewalk transverse to my front door and then turned and walked head-on towards the camera.
All of the pictures are full resolution crops out of a 12 MP Dahua IPC-HDBW81230E-Z dome camera and include both the test chart and a face. If the camera was perfect it would be able to resolve 0.06 inches, based on the pixel spacing in the image. The head-on images come within 50% of ideal at the higher shutter speeds. The resollution of the horizontal bars comes close to the resolution of the head-on motion at the same shutter speed.
BUT the resolution of the vertical bars is terrible (as expected) except at the highest shutter speed. Based on these tests I have set all of my (non LPR) cameras to shutter priority and use 0-5 mS during the day and 0-10 mS at night. Allowing a slower shutter speed at night is a balance between motion blur and noise.
| Resolution (Inches) | Lateral Motion - (v, h) bars | Head-on Motion |
| 1/2000 S | (0.13, 0.13) | 0.09 |
| 1/120 S | (0.48, 0.12) | 0.09 |
| 1/60 S | (>0.6, 0.15) | 0.12 |
| 1/30 S | (>0.6, 0.21) | 0.17 |
Head-on, 1/30 sec (ignore Post-it that says 1/60)
Transverse, 1/30 second
Head-on, 1/60 second
Transverse, 1/60 second
Head-on, 1/120 second
Transverse, 1/120 second
Head-on, 1/2000 second
Transverse, 1/2000 second
For my LPR cameras, as similar calculation uses a car traveling at 30 mph or 528 inches per second. The camera is at an angle of around 30 degrees. This reduces the transverse motion by a factor of two. During the day I use shutter priority and 0-0.5 mS to keep the motion blur under 1/8 inch. At night I use 1/1000 second for a motion blur of 1/4 inch. No photos, but this is what I arrived at while empirically setting up the LPRs.
