i.MX RT106F face recognition distance and camera angle of view

取消
显示结果 
显示  仅  | 搜索替代 
您的意思是: 
已解决

i.MX RT106F face recognition distance and camera angle of view

跳至解决方案
2,153 次查看
Ryo_Aoki
Contributor III

Hi community.

Our customer is interested in i.MXRT106F's face recognition solution.

The face recognition distance is listed as up to 1.4m in SLN-VIZNAS-IOT-UG(3.2.15 Configure Detection Resolution) .

What angle of view are you using for this condition?

We believe that this distance will change for wide-angle and telephoto cameras.
Please tell me the angle of view of the camera that satisfies the condition of 0.3m to 1.4m.

Best regards.

标签 (1)
0 项奖励
回复
1 解答
2,081 次查看
jeremyzhou
NXP Employee
NXP Employee

Sorry for replying late.
There's the below feedback from the solution team.

One thing I should caution about those distance figures, is that 1.4m is the maximum distance in the ideal case (i.e., face is right in front of the camera). The maximum angle supported by the application will largely depend on the customer’s needs and can be customized by the camera module vendor/maker.

In general, we also recommend that for "Smart Lock" applications (read: battery-based applications), the distance should be restricted to < 1m because the liveness algorithm uses the IR & RGB cameras for the liveness detection algorithm, and this is highly dependent on lighting conditions and the use of power-hungry illuminator LEDs. Therefore, detecting a face at a further range will require driving these LEDs at a higher power state which can reduce battery life for battery-based applications.

Finally, regardless of the conditions in which the image is capture (e.g., face’s position in the image, resolution of the camera, distance from the camera, etc.), we need to be able to ensure that the "face box" cropped from the image is larger than 100*100 pixels, otherwise, the face detection algorithm will not accept it as a "qualified" face.


Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

在原帖中查看解决方案

0 项奖励
回复
6 回复数
2,137 次查看
jeremyzhou
NXP Employee
NXP Employee

Hi @Ryo_Aoki ,

Thank you for your interest in NXP Semiconductor products and for the opportunity to serve you.
According to your reply, you want to know how the detection distance is influenced by wide-angle, telephoto cameras, light angle and intensity, etc, is it right?
Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

2,130 次查看
Ryo_Aoki
Contributor III

Hi, jeremyzhou.

Thanks for your reply.
That's about right.
Is there a table showing how the detection distance changes depending on the angle of view of the camera?

In addition, it would be great if there was information about how large a face must appear in an image to be recognized.

Best regards.

0 项奖励
回复
2,126 次查看
jeremyzhou
NXP Employee
NXP Employee

Hi @Ryo_Aoki ,

Thanks for your reply.
I'll contact the solution team about your inquiries soon later, and inform you ASAP if I get some feedback from them.
Actually, I'm not very clear with your question, so can you clarify it again?
Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

2,120 次查看
Ryo_Aoki
Contributor III

Hi, jeremyzhou.

Thanks for your reply.
I look forward to your feedback.
And sorry if my question is hard to understand.

1.
Simply put, I would like to know the angle of view of the camera used in SLN-VIZNAS-IOT.

Our customers care about the performance of the camera paired with the imx106F.
For example, with a telephoto camera, I think the distance at which face recognition is possible will be far. (10m to 20m, etc.)
And with a wide-angle camera, we believe that the distance at which facial recognition can be performed will be closer. (10cm to 50cm etc.)

SLN-VIZNAS-IOT-UG describes the face recognition distance from .3m to 1.4m.
Our customers want to know what angle of view a camera should be used to achieve this level of face recognition distance.


2.
I think that the size of the face in the image will ultimately determine whether or not it can be recognized.

For example, if the face is only 10x10 pixels in the VGA image, I think that face recognition is not possible.
Looking at SLN-VIZNAS-IOT-UG,
It is expected that face recognition can be performed if the face is about 200x200 pixels in the VGA image.
I would like to know what percentage of the image the face must occupy for face recognition.

Best regards.

0 项奖励
回复
2,082 次查看
jeremyzhou
NXP Employee
NXP Employee

Sorry for replying late.
There's the below feedback from the solution team.

One thing I should caution about those distance figures, is that 1.4m is the maximum distance in the ideal case (i.e., face is right in front of the camera). The maximum angle supported by the application will largely depend on the customer’s needs and can be customized by the camera module vendor/maker.

In general, we also recommend that for "Smart Lock" applications (read: battery-based applications), the distance should be restricted to < 1m because the liveness algorithm uses the IR & RGB cameras for the liveness detection algorithm, and this is highly dependent on lighting conditions and the use of power-hungry illuminator LEDs. Therefore, detecting a face at a further range will require driving these LEDs at a higher power state which can reduce battery life for battery-based applications.

Finally, regardless of the conditions in which the image is capture (e.g., face’s position in the image, resolution of the camera, distance from the camera, etc.), we need to be able to ensure that the "face box" cropped from the image is larger than 100*100 pixels, otherwise, the face detection algorithm will not accept it as a "qualified" face.


Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

0 项奖励
回复
2,068 次查看
Ryo_Aoki
Contributor III

Hi, jeremyzhou.

Thanks for your reply. I will discuss with the customer.

 

Best regards.