The Research Experience for Teachers Program /

Activity Title: “Face recognition using Kinect”

Procedure

Background:In this lab you will work with Kinect to obtain the (x, y) coordinates of different features on their face and will save them in a file for further analyses. Also, you will use ImageJ to determine a relationship among the distances between facial features. The Face Tracking Visualization program provides the (x, y) coordinates of facial features. In total, 88 points on the face will be registered by the Kinect as shown in Figure (1). The program will take pictures of the user’s face for further studies.

Preparation:

  • If the Kinect for Windows SDK is not installed, download and run the “KinectSDK-v1.7-Setup.exe” and perform the default Installation. This will install the drivers and libraries for the Kinect camera.
  • Set up the Kinect camera in a location that provides sufficient space for capturing face images. For best results, the Kinect front should be at the same level as the user’s noise. There should not be anyone else in the frame. Make sure the user’s eyes are level .
  • Make sure your computer has the appropriate software.

Lab Activity:

Student Handout

Pre-assessment:

  • How could we use ImageJ to determine the lengths and angles of facial features (eyebrow length, width of face, angle from eye to nose and mouth, etc.)?
  • How closely would you expect the lengths and angles of the facial features for each group member to appear?
  • What differences in the above measurements would you expect to see between your group members?
  • How do you think ratios could apply to the human face?

Procedure

Every member of the team will stand in front of the Kinect and will perform these steps individually:

  • Open the Face Tracking Visualization program that is located on the computer’s Desktop.
  • Keep your face about one meter (3 feet) away and directly facing the Kinect to obtain the best image. As Figure 2 shows, the capturing window will appear once the program is correctly running and the face is registered by the Kinect.
  • To capture an image, open and close your mouth once when you are in the most suitable position directly facing the camera.

Figure 1-capturiung windows

  • As soon as you close your mouth, the picture of your face will be saved in the My Picture Folder on the C:\ Drive of your PC. Once a picture is taken, you need to go to the Libraries\Pictures directory and rename the image file to your name.
  • To view the collected (x, y) coordinates of the 88 points, go to the Face Tracking Visualization folder on the Desktop and then to the Single Face Folder to find the Point.txt file and rename to your name as well. This file contains the data on the 88 coordinates.
  • Repeat this for all members in your team and every time rename the image file and the data file with the team member’s name.
  • After obtaining all the images and data files, open each of the text files, and the corresponding images in ImageJ for analyses as shown in Figure 2.

Figure 2-Analyze coordinates

  • As part of the analysis, you will try to find critical features that are unique for each member.
  • You are required to find some distances, angles or ratios that can assist you recognize faces based on these unique features.
  • You need to have discussions with your teammates to determine which features are best for recognizing faces.
  • After developing an approach, then start analyzing the images of the team members.
  • You will use ImageJ to process the images and to obtain the parameters that you determined are best for recognition of different faces.

Figure 3 shows an example of some of the important facial features

Figure 3-facial feature [2]

Data/Observations

What approach will your group utilize?

What distances and angles are you going to measure?

What ratios will you calculate?

Group Members / Distance 1: Width of face / Distance 2: Length between pupils / Ratio
Group Members / Distance 3: / Distance 4: / Ratio
Group Members / Distance 5: / Distance 6: / Ratio
Group Members / Angle 1: / Angle2: / Angle 3: / Angle 4:

Conclusions:

In what way were your group member’s faces similar? Different?

What comparison between group members created the largest variability?

Do you think you could use your method(s) to identify a wanted criminal?

End of Student Handout

Reference Material:
[1] Microsoft Kinect SDK website (

[2] Biometric technology research (

[3]

Assessment

Pre-Assessment: Use the pre-assessment questionnaire to determine a student’s attitude and familiarity with the topics in this lesson.

Results/Conclusions: Students should develop two methods to find unique features for facial recognition: one using length and ratio, one using angle measures. (An extension standard could be to develop a method using area.)