MARCO METODOLÓGICO

5.1 Research Summary

In this research an attempt was made to design and build an autonomous security robot that conducts its operation based on face recognition and password authentication. The main robot operation involves many sub-tasks:

 Motion detection.  Face detection.  Face recognition.  Navigation.  Obstacle avoidance.  Authentication.

Each of these processes is based on previously researched work conducted by fellow researchers around the world. Here these processes were tested and implemented based on a practical evaluation of the parameters involved. This practical evaluation was done in order to fine tune the applied parameters and refine and module’s operation and improve the outcome of the methods. In some cases new approaches or novel methods were used to optimize the operation of these modules to increase the efficiency of the robot operation.

The research objectives mentioned in section 1.2 have been met as follows:

 The robot has been equipped with 360 degree vision using three wide angle cameras.

 The robot has been equipped with a fourth camera featuring Full-HD video recording and image capturing capability. This camera was used to capture images of detected subject.

 The robot successfully tracked and approached subjects as mentioned in section 4.5.1.4 and shown in videos provided in the accompanying disk.

 Biometric subject verification was successfully performed using face- recognition as mentioned in section 4.3.2.5

 Subject authentication was also successfully carried out as shown in Figure 3.33.

5.2 Contributions

The contributions of this work are divided into the following areas:

 Using three wide angle cameras to simultaneously conduct motion detection of the area surrounding the robot resulting in the following benefits eliminating the need for catadioptric (fisheye) cameras.

 The use of a Full-HD wide angle camera for face detection using the Viola Jones method with skin-content verification. Face detection systems normally use normal cameras in their operation; however, since the robot used in this project conducts surveillance operations, the same wide-angle cameras used for surveillance have been employed for face detection purposes.

 Simultaneous localization-of and navigation-towards a dynamic target based on face or body location while avoiding static and dynamic obstacles in a non-structured environment by an autonomous mobile robot.

 The use of multiple and diverse face recognition algorithms for recognizing a person’s face in live video on a mobile robot. The Fisher faces method, Eigen faces method and Local Binary Pattern Histogram method were used together to recognizer a person’s face and determine his/her familiarity.

 Robot design: A different design of a security robot in which two dc motors are used to drive four big wheels in a differential manner to increase stability while maintaining high speed capability. Also, the robot has variable height (150 – 160 cm) with the observation and monitoring apparatus located at the top of the robot while the main robot weight is located at the base to improve stability during operation.

5.2.1 Cost Comparison

One of the objectives of this work was to build a cost effective security robot. To better highlight the contribution of this work, a comparison has been made between the security robot presented in this work as well as commercially available security robots in cost as well as specification.

Cost Comparison: (The prices are in US$)

Cost ASR 1.0 Knightscope’s K5 Gamma2Robotics’ Vigilant

Initial 1,000 - 45,000

3-Years Contract - 164250* 66,000

*K5’s cost breakdown:

1 hour = US$6.25; 24 hours = US$150.0 1 Year = 54,750.0; 3 Year = 164,250.0

5.2.2 Efficiency Comparison

The robot presented in this work boasts better efficiency than previous robots research robots. Table 2.1 presents a general comparison of this robot in with previous autonomous security robots used in research.

As can be seen in the table, the robot presented in this research is the only one that performs all of the following inclusively:

 Better vision: 360 degrees vs. 90 degrees for most other robots.

 Face detection: Many of the other robots do not perform face detection.

 Face recognition: Many of the other robots do not perform face recognition.

 Authentication: Most of the other robots do not perform password authentication.

All the above in addition to the speed of the robot (1.9 meters/second) deduces a more efficient security operation when compared to previous autonomous security robots.

5.3 Limitations and Future Work

Although the current work has been successful in conducting security operations in the indoor environment of the robot, several possibilities for future work exist:

 Using a PTZ camera to capture the eye image of a subject and use it to extract the image of eye iris. This iris image can be used to further determine the identity of the person. This would result in a more efficient security operation and may limit or eliminate having to request a password entry by unrecognized subjects. This would greatly automate the robot’s operation and limit the number of times the robot has to approach a subject to request him/her to verify his/her identity with a password entry or otherwise.

 Adding a capturing capability to enable the robot to capture intruders who fail to verify their identity and refuse to leave the site. This could be a mechanical structure attached to the robot body and activated in predetermined scenarios and situations.