RFI QandQ - On the Move Capture.pdf

Request for Information for On The Move Biometric Facial Comparison Technology for CBP

DEPARTMENT OF HOMELAND SECURITY

U.S. Customs and Border Protection

Request for Information Question and Answer

On The Move Biometric Facial Comparison Technology

Issuance Date: August 27, 2024

1.0 QUESTION AND ANSWER

# Question Response 1 Could you please clarify what is the process for making an evaluation? Will this be done through this RFI or will there be a selection after a trial, potentially?

This RFI does not constitute a promise to issue an RFP or RFQ. If an RFP is issued, it is anticipated that it may be released in FY2025.

2 Could you please clarify what is the timeline for RFP release once the RFI submissions have been delivered and reviewed?

This RFI does not constitute a promise to issue an RFP or RFQ. If an RFP is issued, it is anticipated that it may be released in FY2025.

3 Is the agency seeking to capture data on travelers entering the checkpoint solely through vehicular traffic, foot traffic, or both? Does this solution apply to land, air, and sea-based ingress and egress of border traffic?

The focus of the RFI is on the pedestrian air ingress environment.

4 Does the solution have to be deployed as of today in a productized environment?

The solution should be at a Technology Readiness Level (TRL) of 8 (system complete and qualified) or 9 (actual system proven in operational environment) at the time of response submission.

5 What is the pilot timeline - when will it start and how long will it run for?

We anticipate technical demonstrations to begin in FY2025.

6 What is the expected number of locations that will be involved in the project or initial deployment?

CBP is looking to industry to provide technical solutions that can be deployed at one or more locations.

7 Is it the government’s intention to use existing cameras in place today across CBP entry points, or should the proposed solution assume fresh install and include a recommendation for camera specifications?

The proposed solution should assume fresh camera install and include a recommendation for camera specifications.

8 Will CBP consider software-only solutions with hardware provided separately by 3rd party vendors?

CBP is seeking solutions that propose and integrate both hardware and software.

9 How does CBP view the use of metadata (aka, non-identifying) for use in monitoring the performance of infrastructure, model performance, drift, and other possible indications of AI model operation? Would we be able to

CBP will allow the collection of metadata to monitor performance.

collect this? Transport it to a secure location to monitor model and infrastructure performance across all deployed installations?

10 Does CBP have a position or interest to include audit capabilities? Either CBP personnel-based audits at random? Data driven metadata-triggered audits? 3rd Party Service company audits to ensure model accuracy? The challenge will always be ensuring that inference conditions are performing as expected, and that new performance and/or edge case condition capabilities are added to improve model accuracy of the life cycle of this platform.

CBP is interested in auditing capabilities.

11 Does CPB anticipate use of an existing terminal at each CBP officer station, that our solution we will send a message on, or will we provide another box, tablet, or terminal screen at the officer station for the results/feedback of the system on the traveler?

CBP is open to a solution that provides a tablet, terminal, or other hardware solution for the Officer to view system feedback with traveler results.

12 Are there any requirements for the solution to integrate the TVS matching response with CBP vetting systems for officers to make the admissibility decision?

This is not a requirement, but CBP would be interested in solutions that could provide this service.

13 May you provide more details about the corridor the travelers walk through? Will it allow multiple travelers side-by-side? Will there be any separation between travelers? Will it prevent smaller travelers (e.g. children) to be completely obscured by taller travelers?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

14 What are the specific performance requirements for the system in terms of accuracy, speed, and throughput?

The solution will have an operational service-level agreement (SLA) that CBP is currently in the process of defining.

15 What are the onsite power assumptions? Power will be provided by CBP/airport (specific sites yet to be determined) and will be available for the technology's use, consistent with power requirements at the given airport.

16 Should the solution include a tablet device with custom software for each gate that: a) Runs locally and interfaces with TVS for matching? b) Displays a green light for a match and a red light for a non-match? c) Operates without requiring login? d) Shows recent matches and upcoming persons in the queue?

CBP is open to a solution with a tablet or other hardware device with custom software that interfaces with TVS to receive matching and vetting results. The government is looking to industry responders to provide proposed solutions and technology options for displaying responses by CBP.

17 Is the requirement to: a) Collect multiple 2D images of each traveler over a period (potentially 1-20 minutes)? b) Temporarily store these

The requirements include:

a) Collecting one or multiple 2D images over a period of time for a pedestrian traveler to walk through the ingress lanes (distance images? c) Select and send only the best quality image to TVS for matching?

varies by location, up to a few hundred feet).

b) The solution should not store images, including on local devices. Images may be stored temporarily for the purposes of selecting and sending images, receiving matching results, and temporarily notifying the Officer of the result.

c) If the proposed solution collects multiple images of a traveler, the requirement is to select and send only the best quality image to TBS for matching.

18 Is there a preferred method for communicated response to CBP for proper traveler segmentation?

CBP is looking to industry responders to provide proposed solutions and technology options for communicating a response to support traveler segmentation.

19 What are the onsite internet uptime assumptions? Uptime is expected to be near or equal to 100%.

20 Can you please specify any software or network requirements for functioning on a CBP network?

The CBP network is multi-faceted and requires multiple steps to access a non-guest network. Generally, access will be limited to devices with specifically provisioned certificates.

21 Could you please describe the average capturing area, including background, lighting conditions, open versus closed-air, queue requirements, queue size, length of capture corridor, etc.?

Typical Federal Inspection Service areas are indoor, closed air, with primarily artificial overhead lighting of varying quality. The solution should be adaptable for varying queue environments.

22 Could you please specify if there is an expectation in terms of the dimensions of the control space you wish to exclude the system from capturing and processing through the solution?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

23 What is the expected number of travelers to be processed in a single frame? What is the maximum number of travelers to be processed in a single frame?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

24 In order to support the ADA and PRM passengers who may be in a wheelchair, could you please specify if CBP will whitelist the Wheelchair pushers to avoid creating alerts when they are supporting a passenger?

Based on CBP CONOPS, whitelisting will not be required for this solution. However, CBP would be interested in a solution that could offer this service.

25 Could you please specify if there is any restriction on the age or height of the children that will be allowed to use the solution? Can the

There are no restrictions on height or age of children using the solution. Solution should be children walk side-by-side with accompanying adults, or be carried in hands? What kind of visual notification to the CBP Officer is expected when an adult with multiple children walks through?

able to effectively capture travelers with varying heights of all ages.

26 Can CBP provide more detail as to their requirement for directional guidance that should be provided to the traveler based on the response received from CBP systems? What messages are expected to be displayed? Do occlusions need to be detected and reported? Furthermore, could you please clarify if the messages are expected to be shown to passengers which were approved, rejected or in need of additional verification?

Visual directions should be simple and easy to understand by the majority of travelers and should be customizable for varying environments. Occlusions need to be detected and guidance provided. Various interactive instructions should be provided to communicate to the traveler whether to proceed, wait, take off a hat or glasses, or a enter a specific queue.

27 Does CBP accept the use of an IP camera or is a USB camera preferred?

CBP is interested in all solutions.

28 Regarding the requirement, "Does not store images, including on local devices," can CBP confirm it is referring to local disk or other permanent storage devices and not RAM?

CBP is referring to local disk or other permanent storage devices

29 If the data is volatile and destroyed in real-time, can this solution communicate with CBP cloud compute resources? Would CBP consider creating a local gallery at the airport with preauthorized passenger tokens and matching locally? If not, could you please confirm if all the matching has to be completed in the DHS / CBP cloud environment?

All matching will be completed by CBP.

30 Does the data need to be centralized or is edge deployment sufficient? Are you expecting each facility that implements the solution to have its own servers/edge devices in a non-managed environment?

Depends on the solution --- there could be solutions that allow for applications to be housed within the CBP network.

31 Are data/images required to be on premise in the US (physically at rest)?

Yes.

32 What is the performance that CBP would guarantee for the network and matching process once the captured image leaves the vendor environment and is sent to the CBP network for processing and matching?

The matching process will be solely managed by CBP.

33 Is it expected to create a buffer zone, to confirm the passengers that have moved through the corridor and are processed and are in the USA?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

34 Once the passengers have crossed the threshold would be a requirement to update the CBP TVS records or other system proving that travelers have entered the country?

No.

35 Regarding the requirement, “Ensures privacy regulations of travelers and non-travelers," what privacy regulations is CBP referring to?

CBP must be confident that traveler data is adequately controlled and protected.

Technology solutions abide by security requirements as defined by CBP.

36 Are there any requirements for the device to notify officers for any missed facial capture in the assigned dedicated lanes?

Yes. The system must alert Officers to failed photo capture in adequate time for the Officer to redirect the traveler.

37 Are power outage conditions of issue we need to account for, and if so, what are the requirements?

No.

38 Will there be multiple zones at an airport for example that could be used for capture of the facial images?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

39 Please define speed as movement of traveler or speed of ability to capture quickly. Is this expected to operate as a continuous facial capture system for the capture area or be activated as needed?

The system will be in a continuous state of operation to capture travelers at varying walking speeds during continuous airport operations.

40 Is there liveness criteria? Should capture prevent also 3D masks or just 2D photographs? Is the expectation to use multiple cameras placed strategically around queues and near CBP gates to capture only real, live 2D facial images of travelers?

There is no liveness verification requirement.

41 Is the system allowed to temporarily retain the images or tokenized copies of the images during the timeframe individuals are passing through the "border crossing" area to identify which traveler's image were already sent to the system for matching? Will temporarily retaining the images be allowed to enable determining if a traveler leaves and later reenters the lane, or repositions themselves within a given or different line be allowed?

Yes.

42 What happens if the traveler leaves the line after being photographed and submitted to CBP and prior to CBP admissibility decision then later re-enters the line; will capturing a second photograph and submitting the photograph to CBP be required?

Yes. The solution should prevent duplicate photo submissions to CBP within a pre-defined short timeframe, however, not prevent a duplicate photo after a customizable timeframe has expired.

43 Where is the CBP officer located in relation to the corridor travelers are walking through? Is it expected that the CBP officer will be mobile or stationed behind a desk?

The government is looking to industry responders to provide proposed solutions and technology options to address the objectives and challenges described in the RFI.

44 In what format will data be transmitted and received to and from CBP systems? What is the average processing time?

This depends on the CBP system being interfaced with --- generally, APIs are REST.

45 Requirement 13 states, "Provide a user interface that clearly communicates traveler expectations to achieve a TVS quality photograph capture at walking speed." Is this to mean that the traveler needs to acknowledge and allow a TVS photograph capture, or is it just information for the traveler to read while walking? Does the information need to be visible and audible and is there a multi-language requirement?

Instructions provided should be simple enough to indicate to travelers where to look, to remove items that can obstruct camera views, and how to proceed. Solutions need to be 508 compliant.

46 Does TVS provide any additional information beyond a simple match/non-match result? For example, are there different types of "green" or "red" statuses based on various factors? Meaning, will our system simply provide RED/GREEN and image, or is there other feedback "states" that we will need to report to the CBP officer?

CBP would provide a singular response, if any.

47 Is the functionality for camera remote adjustment, directional, and zoom control based on the idea of "fine tuning" camera position post installation? Or is it to accommodate changes to the border crossing area? Lanes, Booths, etc.? Or is it for officers/personnel to scan for specific persons/activities in real-time?

The intent of the technical question is for fine tuning camera functionality post installation.

48 Are there any specific cybersecurity or data protection requirements beyond those mentioned in the RFI?

Compliant with cybersecurity requirements per the DHS Information Security Program for DHS sensitive systems and systems that process sensitive information for DHS.

49 Will graphics, diagrams, and tables be counted as part of the 15-page limit for responses? Is it possible to provide Appendices which may provide additional information not central to the main response but provides context to the overall solution.

No.

50 Please confirm whether the video demonstration is required as part of our submission package due on September 12th, or if the government will follow up with select vendors after evaluating initial RFI responses.

Video demonstration is not required as part of the RFI response.

51 How can the camera control subject head tilt, roll or pan when the subjects are only semi cooperative? Is the camera system supposed to reject images that are outside of this spec?

Images that do not meet specification should be rejected.

52 With regard to the requirements for the Photo Capture specifications, as the quality will depend on the customer behavior, if these are strictly adhered to, it may result in increased rejection rates. Are ICAO standards acceptable for the

Images are required to meet TVS standards.

solution, for example, in terms of size of the photo and number of pixels between the eyes?

53 Will frontal facial scan suffice? CBP is interested in all solutions.

54 Provided that the latest NIST publications split the ANSI/NIST ITL 1-2011 Type 10 standards in to Type 11 for Face and type 12 for SMT, could you please clarify if the supplier should focus on Type 11 only, or if it is required to comply with Type 12 as well?

Type 12 is not required for this RFI.

55 Spec. 2 states, "Minimum resolution is 480 x 760pixels". To confirm, this is 480-pixel width and 760-pixel height, correct?

Correct.

56 Spec. 3 states, "Minimum distance of 80 pixels between the eyes". Is this distance the interpupillary distance? In other words, the pixel distance between the center of pupils of each eye?

Correct.

57 Spec.4 states, "Eye roll of no more than 30 degrees". Could the definition of eye roll please be provided? This term is not in ANSI/NIST-ITL 1-2011.

This requirement can be omitted from the RFI response.

58 Spec 5 explains tilt, roll, and pan limits as a percentage. Could the definitions of tilt, roll, and pan and percentage please be provided? ANSI/NIST-ITL 1-2011 Update 2013 in Section E.7, Figure 27 only defines roll, yaw and pitch of the subject's heads. These parameters are defined in units of degrees, so it is unclear to what percentage would apply. Is percentage supposed to instead be degrees?

Tilt is equivalent to pitch, roll is same as roll, and pan is equivalent to yaw.

59 Spec. 6a (i) Exposure defines average 8-bit RGB for an "area". What does the term area refer to? Is it just the face area with any hair, hats, etc., masked out? Are we looking for all subject's skin to have an average value of 105-125?

The Uniform Illumination section provides general guidance to clients to help ensure facial image capture is a uniformly lit photo.

60 Spec. 6a (i) states, "Exposure states a standard deviation of +/-10." How can standard deviation be +/- because standard deviation (stdev) is a positive value? It is impossible to guarantee a stdev if we have no control over illumination or shadowing created by other travelers that will result in shadowing on a subject's face.

The Uniform Illumination section provides general guidance to help ensure facial image capture is a uniformly lit photo.

61 Spec. 6a (i) states "The maximum difference between the means of any two of the RGB components shall not exceed 10". Does this mean AbsoluteValue[ Mean(R) - Mean(B) ] <10 and same for other two channel component pairings? Looking at images in our employee face database taking with a standard consumer

The Uniform Illumination section provides general guidance to help ensure facial image capture is a uniformly lit photo.

automatic camera, Mean(R) > Mean(G) > Mean(B) across ethnicities. The differences in means far exceed the 10 gray levels spec given. Do you require that the gray levels of the face image be equalized across color channels? This would compromise any presentation attack algorithm you may desire to implement in the future. It would also alter the ability of a CBP officer to look at the color equalized image and compare live to the subject standing before them if a visual comparison is required for the envisioned use case.

62 Spec 6a (ii) states, "...RGB code value to fall in range of 0.5 to 1.5 neutral density." Please explain how neutral density relates to the RGB values of an image pixel.

The Uniform Illumination section provides general guidance to help ensure facial image capture is a uniformly lit photo.

63 Spec. 6b (i) states, "... no saturation on the face." If subject is wearing glasses, hat with reflective areas, or is bald, in order to have appropriate RGB values that Spec 6a(i) requires, there may be saturated areas on glasses or on bald head depending upon lighting. Is this acceptable?

The Uniform Illumination section provides general guidance to help ensure facial image capture is a uniformly lit photo.

64 Spec. 8 defines SNR > 53dB. Is SNR defined same as for biometric specs for finger or iris? For Finger, FBI requires >42dB for Appendix F and about 37dB for PIV and for iris ISO_IEC_39794-6 standard requires 36dB. 53dB seems rather high, please clarify.

This is a general guideline to try to achieve greater than or equal to 53dB SNR.

65 What is the field of view (FOV) and distances the camera must operate over?

Camera must choose a FOV setting as to not distort the face and the distance of the camera operation must allow the capture of at least 80 px between the subject's eyes.

66 Are there any optical resolution requirements (e.g., MTF or CTF @ ?? lp/mm at the face) for the face camera system?

Not at this time.