Support Center

Face Recognition Setup Guidelines

Last Updated: Feb 01, 2017 05:54AM PST
Article:3.01-kdb-0004 



 The Face Recognition (FR) video analytic requires adjustment to achieve the best results.

To recognize a face, that face must be captured first. Face Recognition compares a captured face to the images stored in the system. The comparison images loaded into the system should be high quality for best results.

Using a good quality digital camera is the best tool to capture faces for comparison and storage on the Gateway. Cell phone cameras are not recommended. Ensure no special lenses, such as a fisheye, are installed on the camera.

The Face Capture (FC) analytic typically requires the same adjustments as Face Recognition for best results. See the Face Capture and Face Recognition sections in Chapter 5: Controlling Cameras in the NLSS Unified Security Suite User Manual for instructions on configuring Face Capture and Face Recognition, respectively.
 



 Photograph Guidelines 



 The most important consideration for accurate face recognition is a good photo in the database.
 



  Photos must be in the JPEG format, no larger than 5 MB. 

 Photos must meet the minimum requirement of 300x300 pixels of a complete facial photo. 

 NLSS recommends that higher resolution facial photos should be used over the minimum requirements. 

 Facial photos of 900x900 pixels or better is recommended. 

 Position the camera at face level for a direct view of the person's face 

 Take more than one photo and from slightly different angles if possible. 

 If the subject wears glasses, take a pictures with them off and with them on. 

 The more reference photos the better the accuracy of the recognition system. 

 Don't use blurry or unfocused pictures. This will cause the system to produce many false positive events.
 

 

One method to ensure excellent photos is to follow the guidelines set for US Passport Photos.



  Requirements are available at:
http://travel.state.gov/passport/pptphotoreq/pptphotoreq_5333.html 

 Example photos are available at:
http://travel.state.gov/passport/pptphotoreq/photoexamples/photoexamples_5300.html 

 A free tool to crop and resize photos is available at:
http://travel.state.gov/_res/flash/cropper/FIG_cropper.html# 

  IP Camera Guidelines 



  Position the camera at face level for a direct view of the person's face 

 Verify the camera lens and cover are clean. 

 If possible, use a variable lens that can be focused on a face, rather than a wide angle lens. 

 Make sure the resolution is set for no higher than 1080p. 

 For most applications, 720p resolution is sufficient for good Face Recognition accuracy, but 1080p is better. 

 For wide area scenes, where the faces are a small part of the view, a 1080p resolution may provide better results. 

 Set frame rate to a setting no lower than 15 fps. 

 The angle of view for the face capture is no more than 15 degrees. 

 Verify the background is not washed out due to lighting or sunlight. Using a wide, dynamic range camera typically improves results across a broad range of lighting. 

 Verify that the subject face remains in the camera’s view for at least two seconds. 

 Verify that there are no shadows in the facial area 

 Images with sunglasses and/or hats do not provide good results. 

 Images with different facial hair from data base pictures do not provide good results. 

 When possible, provide procedural instructions that help ensure that the above guidelines are met. For example, require people in a security line to remove their hats and sunglasses and look at the camera as they pass through a checkpoint. 

 Use the video analytics forensic to adjust the Face Recognition parameters (sensitivity, confidence, best match) and verify the results based on different settings. See the Face Recognition and Video Forensics sections in Chapter 5: Controlling Cameras of the NLSS Unified Security Suite Users Manual for instructions. 

  General Guidelines 



  Maximum number of people per Face Recognition analytic is 5000. As the number of people in the face group decreases, the accuracy typically increases. 

 Use the video analytics forensic to adjust the Face Recognition parameters (sensitivity, confidence, best match) and verify the results based on different settings. See the Face Recognition section in Chapter 5: Controlling Cameras of the NLSS Unified Security Suite Users Manual for instructions. 

  Contact Information 

For more information, please see http://support.nlss.com or email support@nlss.com
0d564f32803c60bd6a0a0c8b3b3ea9d9@nextlevelsecurity.desk-mail.com
http://assets2.desk.com/
false
desk
Loading
seconds ago
a minute ago
minutes ago
an hour ago
hours ago
a day ago
days ago
about
false
Invalid characters found
/customer/en/portal/articles/autocomplete