US-12620486-B2 - System and method for high performance, vendor-agnostic inference appliance

Abstract

Urgent screening in high-throughput, secure environments such as emergency rooms, or security, typically involves multiple devices. Devices that are used for screening are diverse, and typically sourced from multiple vendors. Currently, devices are typically stand-alone. Further, large devices, are costly, having high capital expenses with long commercial lifetimes, sometimes approaching a decade or more. The result of these features leads to duplication of computational resources, obsolescent computational infrastructure, and lack of interconnection between elements. Aspects of this invention include a device, system, and methods to provide vendor-agnostic interconnection between the multiple elements of a defined environment. The disclosed approach untethers AI algorithms from data generation system and increases flexibility in deployment of newer technologies and algorithms. Example systems can be updated or replaced with new hardware, as computational capabilities develop on short or emergent quality improvement cycles, and can adapt nimbly to changes in threats, regulatory requirements or market developments.

Inventors

• Thomas Anthony
• Frank M. Skidmore

Assignees

• ANALYTICAL AI INC.

Dates

Publication Date

20260505

Application Date

20221222

Claims (19)

1. 1 . A computer-implemented method for collecting and processing data from multiple devices to generate an actionable output for security screening or medical screening in secure environments, the method comprising: receiving data associated with an image from at least one of the multiple devices; providing the received data to a processor, and performing data acquisition at the processor; transferring the data, via a high-speed network, to an appliance; at the appliance, performing command, control and inferencing by one or more graphical processing units (GPUs) to generate the actionable output, wherein the inferencing comprises automated threat recognition (ATR) for the security screening, decoupled from a scanner that scans the image to generate the data associated with the image, wherein the appliance is not accessible by any device during operation; using an ATR algorithm to perform a detection and at least one of a look-back analysis and a look-forward analysis to track a condition in time or space, wherein the look-forward analysis comprises registering a person, object, or item in one camera view, and identifying the person, the object, or the item in a second camera view at a different time; and providing the actionable output, via the high-speed network, for executing an action, wherein the data is accessed from the devices of multiple software and hardware vendors, in a manner that is independent of the software and hardware vendors, and further wherein multiple ATR algorithms can be executed on the appliance.
2. 2 . The computer-implemented method of claim 1 , wherein the command, control and inference comprises security detection threat to identify security threats.
3. 3 . The computer-implemented method of claim 1 , wherein the command, control and inference comprises applying artificial intelligence to automatically detect actionable threats without input by a user.
4. 4 . The computer-implemented method of claim 1 , wherein the command, control and inference comprises identifying critical medical conditions or diagnoses.
5. 5 . The computer-implemented method of claim 1 , wherein the command, control and inference comprises applying artificial intelligence to automatically detect disease or conditions without input by a user.
6. 6 . The computer-implemented method of claim 1 , wherein the data from the multiple devices is presented and managed in a common format.
7. 7 . The computer-implemented method of claim 6 , wherein the common format comprises at least one of a United States DICOS (Digital Imaging and Communications in Security) imaging format, a DICOM (Digital Imaging and Communications in Medicine) imaging format, or a Unified File Format (UFF).
8. 8 . The computer-implemented method of claim 1 , wherein the look-back analysis comprises tracking a history of an identified object or an individual before an event common format comprises a DICOM (Digital Imaging and Communications in Medicine) imaging format.
9. 9 . A system for collecting and processing data from multiple devices to generate an actionable output, the system comprising: a scanner device configured to scan data associated with an image from at least one of the multiple devices; a processor configured to receive the scanned data and perform data acquisition; a high-speed network configured to transfer the data from the processor to an appliance; and the appliance configured to perform command, control and inference by one or more graphical processing units (GPUs) to generate the actionable output, wherein the inference is associated with automated threat recognition (ATR) for a security screening, decoupled from a scanner that scans the image to generate the data associated with the image, wherein the appliance is not accessible by any device during operation, wherein the actionable output is provided, via the high-speed network, for executing an action, wherein the appliance uses an ATR algorithm to perform a detection and at least one of a look-back analysis and a look-forward analysis to track a condition in time or space, wherein the look-forward analysis comprises registering a person, object, or item in one camera view, and identifying the person, the object, or the item in a second camera view at a different time, and wherein the data is accessed from the devices of multiple software and hardware vendors, in a manner that is independent of the software and hardware vendors, and further wherein multiple ATR algorithms can be executed on the appliance.
10. 10 . The system of claim 9 , wherein the appliance is configured to process the data at speeds configured for security screening, and presenting and managing data in a common format.
11. 11 . The system of claim 9 , wherein the high-speed network has connect capability that is embedded.
12. 12 . The system of claim 9 , wherein a graphical user interface (GUI) is located only at a remote area, only allowing remote access and monitoring of security activity.
13. 13 . The system of claim 9 , wherein one of the multiple software and hardware vendors can connect to one or more other ones of the multiple devices to access a network of devices.
14. 14 . The system of claim 9 , further comprising access points that are geographically remote, so as to permit allowing screening and networking capability based on geography.
15. 15 . The system of claim 9 , wherein a single one of the appliance can support multiple lanes of a security checkpoint.
16. 16 . The system of claim 9 , wherein the appliance is configured to reduce compute, power and cooling required at the processor while providing failover and redundancy, and to perform load balancing such that if one of the GPUs fails, another of the GPUs provide a required compute for proper operation of a security facility.
17. 17 . The system of claim 9 , wherein the scanner device comprises an X-Ray or computed tomography (CT) screener for baggage, a millimeter wave screen, or a terahertz screener, and the high-speed network is high speed ethernet, InfiniBand, or optical networking.
18. 18 . The system of claim 9 , wherein connections are managed through an open architecture software library.
19. 19 . The computer-implemented method of claim 1 , wherein the scanner is configured to scan data associated with an image from at least one of the multiple devices without performing inference associated with the ATR for the security screening.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS The present application claims the benefit of 35 USC § 119(a) to U.S. Patent Application No. 63/293,018, filed Dec. 22, 2021, the contents of which is incorporated herein in its entirety for all purposes. BACKGROUND 1. Field Aspects of the example implementations are directed to systems and methods for image data processing, and more specifically, to electric digital data processing associated with a high performance, vendor-agnostic inference appliance that cannot be accessed by onsite users and/or associated devices during operation. 2. Related Art Urgent screening in high-throughput, secure environments such as emergency rooms or security, involve multiple devices in the related art. Examples of related art devices that are used for screening are diverse, and are sourced from multiple vendors. In airport screening environments, for example, the screening of personal items, such as carried bags or (in an airport screening environment) air cargo, occurs simultaneously with on-person screening, using disparate technologies and devices. In some related art circumstances, multiple vendors with multiple screening software and device parameters may be positioned side by side in parallel screening lines. Related art screening devices are stand-alone. Further, large devices, such as cargo and checkpoint X-ray and CT scanners, or medical CT and MRI scanners, are costly, having high capital expenses with long commercial lifetimes, sometimes approaching a decade or more. As a result, there are related art problems and disadvantages, including but not limited to duplication of computational resources, obsolescent computational infrastructure, and lack of interconnection between elements. Operationally, another disadvantage is slower screening, reduced detection efficiency, increased labor requirements, fragility (e.g., dysfunction of a single element can interrupt service), and lack of flexibility to re-organize in the context of a new disease, security threat, or new technology (e.g., artificial intelligence). SUMMARY Aspects of the example implementations disclosed here are directed to a device, system, process, and method to provide vendor-agnostic interconnection between the multiple elements of a defined security or medical environment, including (but not limited to) an airport screening checkpoint line, a hospital imaging center, perimeter security, or passive and active screening technology or technologies for loss prevention, without permitting the user or operator to access an appliance. The present aspects untether (i.e., make inaccessible to onsite users and/or onsite operators during operation) artificial intelligence (AI) algorithms from data generation, and increase flexibility in deployment of additional (e.g., newer) technologies and algorithms, without requiring the re-certification of the OEM equipment itself, which is a slow and expensive process. Some example implementations are described herein that integrate and mediate the output of multiple different vendor software packages, to present a unified dataset and image format that can be used for analysis by humans and/or algorithms. The example implementations can be updated rapidly with new software to address new priorities or threats, adapt to new equipment, and integrate new data and algorithms as requirements develop. Example devices can be updated or replaced with new hardware, as computational capabilities develop on short or emergent quality improvement cycles. The appliance and algorithms on the appliance can be re-certified for highly regulated environments such as security and medical applications more rapidly than the entire certified system going through the recertification process. The appliance, and algorithms on the appliance, can rapidly address changes in threats, regulatory requirements, as well as market concerns and conditions. According to these aspects, a method and apparatus is provided for accessing data from multiple devices in a vendor-agnostic fashion, and presenting data in a unified format for at least the purposes of security, loss prevention, or medical screening. Additionally, a network-capable device capable of managing data flows is provided at actionable, relevant speeds for the environment in question, with capability for embedded software, including artificial intelligence software. Further, a GPU based high performance and high throughput appliance is provided for AI inferencing with failover. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 illustrates a high-performance device according to the example implementations, which can interface with one or more OEM scanner computers and transfer the scanned and reconstructed images over a low latency high speed network such as InfiniBand (IB) with Remote Direct Memory Access (RDMA) to perform the automated threat recognition (ATR) inferencing with multiple algorithms over multiple GPUS. FIG. 2 illustrates a distributed system accord