ZN-Smart Eye — Features functions and presenting the

ZN-Smart. Eye® - Features, functions and presenting the system to clients ZN Vision Technologies AG Universitätsstraße 160 44801 Bochum Germany Phone +49 (0)234 97 87 -0 Fax +49 (0)234 97 87 -77 E-mail: info@zn-ag. com Internet: www. zn-ag. com

Agenda § Product concept, structure and features § Options and versions § Presenting the system to clients and potential buyers § Exemplary configuration, example offer § FAQs, dealing with objections 2

Product concept, structure and features 3

Identification solution ZN-Smart. Eye® – Facial and person recognition in real-time Description Options: § 1: n identification in non-cooperative scenarios § Overall solutions comprising hard- AND software system § Real-time identification of individuals in video sequences § Facial recognition in a crowd and comparison with the system's database § Recognition process in less than 0. 3 seconds § Alarm function advises of the presence of a known person § Database deleted automatically after a given or prescribed period of time. § Reference system and customer-specific category filter (e. g. VIP, employee) as well as storing additional data are possible (e. g. name, address) § Protocol function § Input into the database as a direct video shot, digital photo or digitally scanned portrait § Supports common SQL databases 4

Identification solution ZN-Smart. Eye® – Facial and person recognition in real-time Clients: Performance: Advantages: § § Airports, train stations § § § Efficient crime prevention and investigation § Biometrics is the only technology understandable for humans, system administrators can "comprehend" the pictures of faces Permanent real-time surveillance outperforms human beings when it comes to observing complex scenarios (fatigue, stress) Knowledge of the presence of wanted persons is continuously available, thus enabling direct action of security staff Protection of privacy: distinction between criminals and passers-by is possible § § § Stadiums, major events Casinos, night clubs Supermarkets, banks Enhancement of security without additional personnel Combination of surveillance and protection of privacy 5

The core functions of ZN-Smart. Eye® A video camera records a scene at the selected location (e. g. the entrance of a building or the access to a highsecurity area). Finding the face: The face is sought in the camera's video image. Modelling the face: The face's characteristic features are registered and transformed into a mathematical model. Facial recognition: The facial model of a currently registered person is compared to the facial models of all stored persons. Database Output: A positive result of the comparison (facial recognition) can be transmitted as an optical and acoustic signal. If required, this signal may be transferred to a central point. 6

The procedures Finding the face Where and what is the face in the video image? The face is the area defined by eyebrows, mouth and outer eye area (cf. rectangle in the face)! See page 5 -30 for the configuration of the face size recognizable for the system. The size of the rectangle on the face (not its position) must be between the small and the large white rectangle shown above. Modelling the face Facial recognition Model Database Registers the features characterizing a face and transforms them into a mathematical model. The facial model of a currently registered person is compared to the facial models of all stored persons. The face is characterized by a person's total facial area (the graph has to be positioned on the complete face)! The facial model stored for a person in a database determines the quality of facial recognition. Quality and consistency of the facial models in the database are paramount for the result of the facial recognition. Thus one person should never be stored in the database several times. 7

Options and versions 8

ZN-Smart. Eye® Compact – the all-in-one solution System module Components Compact Station System administration, face capturing & recognition, representation of results, alarm Compact Station. Box Compact Station. Licence Terminal (screen, keyboard, mouse) Camera Manual Digital Camera** Net Price in Euro See accessory list € 25, 000. 00 See accessory list Printer** See accessory list * Information about larger systems available on request ** These system modules are optional 9

Components 1. When unpacking the system you find the following components: 1) The ZN-Smart. Eye® engine 2) The ZN-Smart. Eye® camera 3) An 8 -48 mm auto iris vario objective 4) 10 m video cable RG 59 including adapter (in total: chinch plug on BNC plug) 5) 10 m RS-232 D-sub 9 male/female closed signal for camera control 6) 15“ TFT monitor 7) Mouse 8) Keyboard 9) Installation manual 10)User's manual 11)2 Power cables 2. What you need in addition: 1) A tripod (3 -legged) with an extension of 2 -3 m, 2) Extension cable for power supply 3) Multicontact plug 10

Components 11

Preconditions for successfully employing ZN-Smart. Eye® 12

What are the preconditions for successfully employing ZNSmart. Eye®? 1. Include the employee representatives and the data protection commissioners into the project phase 1. Evaluate whether and in which way data protection is anchored in the company's code of practice. 2. Explain the way the system deals with data for specific and non-specific purposes. 3. Take away the fears of employee surveillance. I. Expose the system's purpose and benefit II. Expose that a. employees could be entered and significant recognitions of the system could be suppressed (on the desktop as well as in the protocol), b. no employees are entered into the database and that data are recorded just like data for non-specific purposes. c. the final decision is left to the human being, no automation due to the system 13

What are the preconditions for successfully employing ZNSmart. Eye®? 2. Inform the client about what to expect of the product, never try to fulfil false expectations. 1. Never admit negative comparisons! 2. Don't assign yes/no decisions to the system! 3. Integrate ZN-Smart. Eye® into the client's work routine. 4. Choose the correct locations and/or correct recording positions in order to get frontal views of the face. 5. Use pictures taken by the camera for person recognition. 6. Plan personnel training from the very beginning. 14

Choosing the location – according to preferences 1. Internal, constant illumination, persons well isolated, no hidden persons Example 1: location within a building, no or little daylight, downwards escalator – department store, train stations, airports Example 2: cash desk scenario or metal detector scenario – discotheques, banks, airports, department stores, train stations - no further requirements with regard to illumination for the system, - very good rates of facial detection and recognition can be achieved 2. Internal, constant illumination, persons well isolated – although persons may be hidden Example 1: location within a building, no or very little daylight, upwards escalator – department store, train stations, airports Example 2: persons in an arcade/on a floor in a building - no further requirements with regard to illumination for the system, - good to very good rates of facial detection and recognition can be achieved 15

Choosing the location – according to preferences 3. Internal, considerable daylight, no direct illumination of the faces, no light shining towards the camera, persons well isolated Example 1: entrance halls, camera installed in the background, door not directly visible in the picture employ ZN‘s camera regulation. good to very good rates of facial detection and recognition can be achieved 4. Light shining towards the camera during daytime, dark background at night, system required to work in both scenarios Example 1: entrance surveillance at train stations, airports, casinos Required measures: I. Contact manufacturer to plan the installation a concept is required II. Draw up an illumination concept – plan expenditures 16

Choosing the location – according to preferences 5. Outdoor application 1. Examples: airport premises, station platforms, public places Required measures I. Contact manufacturer to plan the installation a concept is required II. Draw up an illumination concept for the camera – project 17

Which data are required for recognition? 1. Recommendation: It is advisable to use pictures of cameras used for the recognition. - This guarantees good to very good facial recognition results 2. Register the faces frontally in order to achieve very good recognition rates 3. Point out the importance of database maintenance to client content of client training. 4. If no camera pictures are available for recognition, data can also be imported (by means of scanner or printer) - however, this will affect the recognition capacity 5. Only pictures and personal ID data are required, all additional personal data are optional (also of importance with regard to data protection) 6. Recommendation: When adding a person to the database also register additional data (category, male/female etc. ) providing further search criteria apart from the face. 18

Presenting the system to the client 19

The first step: Definition and determination of the goal Determine the goal! It is paramount to hold a conversation determining the requirements and the intended working method / surroundings under which the facial recognition shall be employed. The following question has to be answered unambiguously: Is ZN-Smart. Eye® the best solution for the task or would ZN-Face® be more suitable? 1. Notes with regard to ZN-Face®: I. A person has to be recognized with a security of 100% – not possible with ZNSmart. Eye®! II. A means of identification or an e-ticketing system already exists. III. 100% security only in case of cooperative persons console procedures are the best solution. 2. Notes with regard to ZN-Smart. Eye®: I. 100% security not required, user wants to be advised if a person of interest enters the building II. Surveillance to be carried out discreetly and inconspicuously. III. Surveillance to be integrated into existing video surveillance. 20

Structure of the work screen Video Module Event Module (empty, no signal) (live image) Status Module Protocol Datenbank Module Control Module 21

Video Module and Status Module § Display of controlled area § Camera status ( on / off) § Name and position of camera § Date and time 22

Control Module Pause Start / Stop Help Database Configuration Off Protocol 23

Advantages of the Control Module § Standard buttons lead to intuitive operation § Independence of language due to symbols § Easy operation thanks to large buttons § Immediate operation, no special knowledge required 24

Event Module Display of faces found Most similar faces, in declining order 26 14 8 34 13 12 Previous events 25

Advantages of the Event Module § Automatic display of the faces found § Selection of the most similar persons from the database 26 14 8 34 13 12 § History of the latest comparisons § Display of a similarity value § Alarm in case of significant similarity 26

Protocol Module Display of events Indications for the event filter with regard to camera Start event filtering personal ID Cancel event filtering 27 time

Advantages of the Protocol Module § Archive containing previous events § Quick access § Possibility of filtering according to time and / or person § Adding persons and / or images to the person database § Quick sifting of video recordings 28

Database Reference image Personal data Visitor categories 29

Database information § Additional pictures § Address § Documents (ID card and driving licence) § Notes 30

Configuration Module § User administration § Administration of visitor categories § Administration of actions § Database administration § Adjusting an optional printer § Camera administration § System status § Working procedures and licensing info 31

How to install the system 1. You will need approximately half an hour to install the system. 2. Choose a location for the camera where the faces are illuminated evenly, preferably from above or from the front. Make sure that the faces appearing in the picture are - neither overexposed, - nor dark, - nor show hard shadows 3. Use a tripod to raise the camera to a height of 2 – 3 m 4. Add a few pictures of yourself to the person database. 5. Ask persons who will be present at the presentation for their approval to add them to the database. 32

How to enrol persons Add Person/Image Dialogue Event Module A B 2 Adjust graph BASIS FOR GOOD RECOGNITION RESULTS 3 Add face to person database Graph positions on faces unsuitable for person recognition Graph too small 1 click left mouse button Graph too large Graph positioned correctly on face A Graph too large Correct graph C Event Protocol View Person Database View 33 Sequence of views Sequence of steps

How to enrol persons 1. Graphs have to be positioned correctly. 2. It is advisable to store as many pictures in various positions and under different lighting conditions as possible. 3. NEVER enter one person into the database several times! 34

Which features does a human being evaluate as opposed to the machine? Feature Human Poses/positions of head without frontal view of face X Other physical features, gestures, facial expressions, height X Surroundings where the person was met for the first time X Scenario/conditions under which the person was met for the first time X Voice/language/vocabulary X Frontal view of face Machine X Consequences: The number of data used by human beings to recognize a person is considerably higher. The single parameter "frontal view of the face" can be evaluated much better by a machine. X Advantages of the human being: Recognizes a person known and familiar to him even in scenarios where no frontal view of the face is available. Disadvantage of the human being: Depends on all the stated parameters and on his experiences no objectivity, knowledge tied to one person, no recognition without experience (e. g. Chinese people) 35

FAQs and dealing with objections 36

Which questions will the client ask? 1. How reliably does the system work? - Recognition rate of 10 to 99 %, depending on the location, illumination and the available and/or comparable pictures - Reaction: That is not sufficient. Such an application makes no sense. I need the system to make a yes/no decision definition of the goal. 2. What are your system's FAR and FRR rates? - Up to date, no official measurements have been carried on this issue. As far as I know, the FAR is around 5% and the FRR around 15%. 3. Can the system also be used to filter out persons not contained in the database? - No, that is not possible. (negative comparison) - Why not? Ø Statistic model, a differentiation between the last "hits" and the rest of the world is not possible. According to my current knowledge there is no system able to carry out such comparisons. 37

Which questions will the client ask? 4. How many cameras can be connected to the system? - One camera and a terminal can be connected to ZN-Smart. Eye® Compact. 5. How many persons can be stored as a maximum? - As a standard, up to 10, 000 persons can be stored in ZN-Smart. Eye® Compact. 6. In which way does a higher number of persons affect the search speed? - The search speed is hardly affected by a total number of 10, 000 persons. 7. How long does it take the system to find a person? - Finding the face – i. e. deciding whether a face is in the picture or not – takes only fractions of a second. - Facial recognition takes 1 -2 seconds for one face. 8. Is the system able to detect more than one face in the picture? - Yes, the system simultaneously registers all the faces detected in the picture and searches for them in the database sequentially. 38

Which questions will the client ask? 9. Why does it take so long to recognize a third person in the picture? - This is because following the simultaneous registration of the faces in a picture, the facial recognition is carried out sequentially, i. e. when there are 5 persons in the picture, ZN-Smart. Eye® needs approximately 10 seconds for displaying the last result. 10. Is the system able to scan large groups of people? - Yes, this was proven by real-life and tolerance tests. 11. Why does the system not always display the recognized person on the first position? - ZN-Smart. Eye® makes suggestions to the viewer. ZN-Smart. Eye® is not able to evaluate a presented face with the highest significance in every facial pose and under all lighting conditions. Based on the suggestions, however, the viewer is in a position to arrange the corresponding reactions. 39

Pilot Installations 1. 2. 3. Every client will inquire about a pilot installation; references will not be enough, but he will want to try the product himself. This pilot phase is to be avoided as it requires plenty of personnel and expenditures. Thus the following procedure is advisable: I. Convince the client by visiting a reference client II. Should this not be sufficient: Make an offer for the test. a. The test must not be free of charge. b. The test always has to be accompanied by staff: No test without supervision! c. The test should only be carried out with standard components - minimize the risk! d. The test has to have a certain goal and it has to be defined clearly that the client has to buy the product if this goals are achieved. e. The results are put down in writing. III. If the first offer is refused, make a second offer which is even more expensive for the client. Force the client to return to Offer 1. IV. If the client is only interested in a free test, refuse it. Tests should be only available for clients with a real interest. 40

Data protection 1. 2. Data protection is an extremely sensitive topic which is always an issue when it comes to biometric data. 1. Biometric data are personal data. 2. New possibilities for the users are offered not by the biometric data themselves but by centralizing such data (storage). However, this also holds certain risks. I. In Germany the generation of a centralized ID database containing biometric data is prohibited by law. 3. Best method of data protection: no storage at all 4. Second best method of data protection: decentralized data storage and surveillance of access to data Personal data may only be stored if 1. they are stored for a certain purpose (definition of purpose according to BDSG [Federal German Data Protection Act]), 2. such storing is carried out with the consent of the persons concerned. 41

Data protection 3. 4. 5. However, the BDSG is only an "omnibus law". The following order applies to legal examination of possible violations of data protection: I. Specific laws of the trade: e. g. regulations for accident prevention, company code of practice, police laws, gambling regulations etc. II. Data Protection Acts of the Länder III. Federal Data Protection Act IV. EU guidelines Important: Employee representatives as well as data protection commissioners should be involved in a project as soon as possible Further information: cf. White. Paper on Data Protection 42