Face Recognition

Post Reply
Tech Support
Posts: 179
Joined: Fri Mar 29, 2019 11:44 pm

Face Recognition

Post by Tech Support » Thu Feb 13, 2020 11:23 pm

NVR Face Recognition
Configuration Guide

1.Introduction
Face recognition and comparison process: The IPC uploads the captured face images to the NVR. The NVR stores and compares the
images and generates the corresponding alarms for matched images. In addition, the NVR supports alarm control linkage to output
diversified alarms.

2Preparations
1.A Model B NVR with the version later than D023SP10 has been obtained.
2.A hard disk has been installed in the first hard disk slot of the NVR.
3.A smart bar has been obtained.
4.A camera has been obtained and will be used with the Face Detection Function
5.Other commissioning tools such as display and computer have been obtained. The computer has been installed with the Microsoft Excel
program.

Note: Because the face recognition commissioning will be implemented on the computer, the commissioning introduced in this document is mainly accomplished through the computer's access to the NVR's web UI.

3.Commissioning
This section describes the commissioning process.

Step 1: Insert the smart bar and check the status.
step1.png
step1.png (28.51 KiB) Viewed 2732 times
Enable the NVR and insert the smart bar. Choose Man-machine System Maintenance>Smart Bar Information. Check the smart bar status
until it is changed to Normal.

Step 2: Confirm the initial use prompt.
STEP2.png
STEP2.png (14.71 KiB) Viewed 2732 times
You need to format the hard disk if you upgrade the NVR from a version earlier than D002 to D023SP10. On the man-machine UI, right-
click to select Face Recognition from the shortcut menu. The system prompts to format the hard disk and restart the system. Click Yes.

Till now, the man-machine operation is completed.

Step 3: Export the template.
STEP3.png
STEP3.png (24.73 KiB) Viewed 2732 times
In the IE browser, enter the IP address of the NVR, log in to the system, enter the Smart page, and click Export Template.

Step 4: Edit the image and template.

4.1 The following figure shows the exported template in .CSV format.
STEP4.png
STEP4.png (20.94 KiB) Viewed 2732 times
Template.csv

4.2 Open the template using the Microsoft Excel. Edit the template according to the image requirements.
STEP4.2.png
STEP4.2.png (31.5 KiB) Viewed 2732 times
4.3 Editing the Template Parameters
STEP4.3.png
STEP4.3.png (32.75 KiB) Viewed 2732 times

1. Except for the gender and image path, other items can empty.
2. You can customize the template name, for example, East Gate Control Namelist.CSV.
Dongmen Control Namelist.CSV

4.4 Editing Images
STEP 4.4.png
STEP 4.4.png (15.2 KiB) Viewed 2732 times
Photo requirements: The photo size is less than 512 KB, and the effective face ratio should reference to the certificate photo. The resolution is within the range of 288*352 to 600*600. The recommended value is 288*352. The photo name can use Chinese characters, English letters, or numbers.

Step 5: Import the template.

5.1 File Normalization
STEP 5.1.png
STEP 5.1.png (70.57 KiB) Viewed 2732 times
Save photos and CSV templates in the same folder.

5.2 Adding a Namelist Library
STEP 5.2.png
STEP 5.2.png (41.43 KiB) Viewed 2732 times
Enter the Smart page, select the namelist library type, and click Add. Name it and select the check mark.


5.3 Importing the Namelist

1. Select the namelist library. After it is highlighted blue, click Mass Import, select the template as prompted, and click Import.
STEP5.3.png
STEP5.3.png (28.3 KiB) Viewed 2732 times
2. The system prompts that the operation is successful. Then, the import is completed, as shown in the figure below.
STEP 5.3B.png
STEP 5.3B.png (23.6 KiB) Viewed 2732 times
5.4 Importing Photos
STEP 5.4.png
STEP 5.4.png (111.19 KiB) Viewed 2732 times
After the template is imported, select the namelist library to view the images in the namelist library, as shown in the figure below.

Step 6: Connect the camera and enable face photo capture.\

6.1 Connecting the Camera
Connect the IPC to the NVR using Uniview protocol.

6.2 Configuring Parameters
STEP6.2.png
STEP6.2.png (52.65 KiB) Viewed 2732 times
On the NVR web UI, choose Configuration>Smart Detection>Face Recognition. Configure the following parameters.
STEP 6.2 B.png
STEP 6.2 B.png (215.59 KiB) Viewed 2732 times
Step 7: Configure the alarm control arrangement.

7.1 Enter the NVR web UI, choose Smart>Alarm Control Arrangement, as shown in the figure below.
STEP 7.1.png
STEP 7.1.png (42.54 KiB) Viewed 2732 times
7.2 Select the control namelist (black or white), and click New.
STEP 7.2.png
STEP 7.2.png (98.29 KiB) Viewed 2732 times

Step 8: Start the face recognition.

After the face recognition configuration is completed, right-click on the man-machine UI to select face recognition, and select a camera
channel on the left side of the UI. Then, you can see the live video image at the upper part of the screen, small face images in the
upper right part, and alarm comparison information in the lower part.



4.FAQ

1.Question: Why can't I see the captured images but only live video after commissioning according to the above steps?

Answer: Log in to the IPC web UI, choose Configuration>System>Server>Smart Server. Change the TMS server address to the NVR IP
address. Then, the IPC will send captured face images to the NVR. This setting is automatically changed when a camera in the new
version accesses to the NVR. However, in some earlier versions, this setting may not be automatically changed by the IPC.

2. Question: How many black and white namelist photos can be imported to the face library?

Answer: A maximum of five libraries can be created and these libraries can store up to 2000 face photos.

3. Question: Where are images captured by the camera stored?

Answer: Captured images are saved in the NVR hard disk. You can query the images by choosing NVR Man-machine Smart Retrieval>Face
Retrieval.

4.Question: How many channels can be arranged for a namelist library?

Answer: One namelist can be arranged on multiple cameras, and one camera supports multiple namelists.

5.Question: Generally, the identity card number is relatively long, and the format in the Excel is displayed as 3.12+17. How can it be
displayed normally?

Answer: Because the input number is displayed as scientific notation by default. To solve this problem, you need to edit the template,
select the document number column, right-click to select Set Cell Format, and set the format to Text.

6.Question: Why does the system prompt me to format the hard disk?

Answer: For the NVR that upgrades from a version earlier than D022 to the face recognition version, you need to format the hard disk.
For the NVR that upgrades from a version later than D022, you do not need to format the hard disk.

7. Question: Why does a template with information correctly filled fail to be imported?
Answer: Photo import also has strict requirements. If a photo does not contain any face, the system cannot analyze the face and the
import fails. Therefore, ensure that a photo contains a face.

8. What do control namelist and alarm threshold mean?

Answer: The control namelist must be selected. Currently, totally five back and white namelists are adopted. After a control namelist is
selected, you can set multiple control channels for the namelist library. This is a one-to-multiple relationship. For example, a criminal
namelist is created on the cell NVR, cameras installed on four gates (east, west, south, and north) can use this namelist library for
control. If no namelist library is configured on the white list, you cannot select the namelist library. The alarm threshold means that the
NVR compares the captured face images with the images in the blacklist database. After the comparison, if it is found that the captured
image is 90% similar to an image in the namelist library, the system generates an alarm. If no captured image is 90% similar to an image
in the namelist library, the system does not generate any alarm. The processing principle of the whitelist is on the contrary. Therefore,
when the alarm threshold is set too low, it is easy to cause false alarms; when the alarm threshold is set too high, it is easy to miss
alarms. The recommended value is 90.

Post Reply