Microsoft Patent | Coarse Relocalization Using Signal Fingerprints
Publication Number: 10600252
Publication Date: 20200324
A first display device and method are provided for accelerating the coarse relocalization process of the first display device by generating a session-specific identifier and sending it to a second display device, which transmits the identifier to a cloud service, which returns at least a portion of a map data set corresponding to the identifier. The returned map data set is then stitched into the local map data of the first display device to create an integrated map, which is used to render one or a plurality of holograms. The first display device may comprise a processor, a memory operatively coupled to the processor, and a fingerprint program stored in the memory and executed by the processor.
6-DoF tracking, also known as six degrees of freedom tracking, is a method by which a device (e.g. mixed-reality head-mounted device (HMD), robot, smartphone, etc.) uses sensors (e.g. cameras, inertial measurement units, etc.) to localize itself, or determine its position relative to its surrounding physical environment. When a device is turned on in a new environment, the device initially performs a coarse relocalization step, in which the rough vicinity of the current location of the device is determined, before the device performs a more fine-grained localization step to determine a more precise position of itself relative to its surrounding physical environment. For example, a mixed-reality HMD or smartphone can use this positional understanding to place holograms or digital content so as to appear to be world-locked to a position in the physical world, and a robot can use this positional understanding to navigate itself relative to its surroundings.
Of increasing value is the ability of the device to quickly orient itself or determine its own position relative to a three-dimensional coordinate space of the physical environment by efficiently loading the appropriate map data sets to accelerate the coarse relocalization step before proceeding to the more fine-grained localization step. A conventional method of self-orientation is to load all the map data sets of the physical environment into memory before the device starts to orient itself. However, this conventional method may be associated with practical disadvantages, including large memory requirements and long load times.
To address these issues, a first display device and method are provided to streamline and accelerate the coarse relocalization process of the first display device. The first display device may comprise a processor, a memory operatively coupled to the processor, and a client fingerprint program stored in the memory and executed by the processor.
The client fingerprint program may be configured to receive a session-specific identifier from a second device, the session-specific identifier corresponding to a map data set; transmit the session-specific identifier to a remote system; receive from the remote system the map data set corresponding to the session-specific identifier; and stitch the map data set into a local map data of the first display device to create an integrated map.
This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows a computing device in the form of a head-mounted display device, according to an example embodiment of the present disclosure.
FIG. 2 is a schematic view of an example predictive corrective algorithm for simultaneous location and mapping by the computing device of FIG. 1 within a physical environment.
FIG. 3 is a time series view of data captured by various sensors of the computing device of FIG. 1 and processed to generate a series of key frames linked by a pose graph.
FIG. 4A is a schematic view of a computing device for generating a current fingerprint, comparing the current fingerprint against fingerprint data, and sending a request for a matching map data set according to a first embodiment of the present disclosure.
FIG. 4B is a schematic view of a computing device depicting one example of the signals and signal sources shown in FIG. 4A.
FIG. 4C is a detailed schematic view of one example of the fingerprints, fingerprint data, and corresponding map data sets shown in FIG. 4A.
FIG. 5 is a schematic view of a computing device for generating a current fingerprint, comparing the current fingerprint against fingerprint data, and sending a request for a matching map data set according to a second embodiment of the present disclosure.
FIG. 6 shows users wearing head-mounted display devices of FIG. 1 in a dining room according to an example of the present disclosure.
FIG. 7 is a schematic view of one embodiment of map data including a corresponding map data set according to the second embodiment of the present disclosure.
FIGS. 8A, 8B and 8C are flow charts of a method of generating a current fingerprint and retrieving corresponding map data sets according to three examples of the present disclosure.
FIG. 9 shows a computing system according to an embodiment of the present disclosure.
FIG. 10 is a schematic view of a computing device for obtaining a session-specific identifier, and sending a request for a matching map data set using the session-specific identifier according to a third embodiment of the present disclosure.
FIGS. 11A, 11B, 11C, and 11D are flow charts of methods of obtaining a session-specific identifier and retrieving corresponding map data sets according to four examples of the present disclosure.
FIG. 1 illustrates a head mounted display device 10 embodiment of the computing device 10, according to one embodiment of the present disclosure, which has been conceived to address the issues discussed above. As shown, the computing device 10 includes processor 12, volatile storage device 14, non-volatile storage device 16, cameras 18, display 20, active depth camera 21. The processor 12 is configured to execute software programs stored in non-volatile storage device 16 using portions of volatile storage device 14 to perform the various functions recited herein. In one example, the processor 12, volatile storage device 14, and non-volatile storage device 16 may be included in a system-on-chip configuration included in the head mounted display device 10. It will be appreciated that the computing device 10 may also take the form of other types of mobile computing devices, such as, for example, a smartphone device, a tablet device, a laptop, a machine vision processing unit for an autonomous vehicle, robot, drone, or other types of autonomous devices, etc. In the systems described herein, devices in the form of computing device 10 may be utilized as a display device as illustrated in FIGS. 2-8C, discussed below.
Display 20 is configured to be at least partially see-through, and includes right and left display regions 120A, 120B which are configured to display different images to each eye of the user. By controlling the images displayed on these right and left display regions 120A, 120B, a hologram 50 may be displayed in a manner so as to appear to the eyes of the user to be positioned at a distance from the user within the physical environment 9. As used herein, a hologram is an image formed by displaying left and right images on respective left and right near-eye displays that appears due to stereoscopic effects to be positioned at a distance from the user. Typically, holograms are anchored to the map of the physical environment by virtual anchors 64, which are placed within the map according to their coordinates and typically positioned on visual features recognizable through image processing of captured images of the environment. These anchors are world-locked, and the holograms are configured to be displayed in a location that is computed relative to the anchor. Typically the holograms are defined so as to be displayed within a threshold distance such as 3 meters from their corresponding anchor, although other threshold distances may be used.
In the configuration illustrated in FIG. 1, a plurality of cameras 18 are provided on the computing device 10 and are configured to capture images of the surrounding physical environment of the computing device 10. In one embodiment, four cameras 18 are provided, although the precise number of cameras 18 may vary. The raw images from the cameras 18 may be stitched together with perspective correction to form a 360 degree view of the physical environment, in some configurations. Typically, cameras 18 are visible light cameras. Images from two or more of the cameras 18 may be compared to provide an estimate of depth, using passive stereo depth estimation techniques.
In addition to visible light cameras 18, a depth camera 21 may be provided that uses an active non-visible light illuminator 23 and non-visible light sensor 22 to emit light in a phased or gated manner and estimate depth using time of flight techniques, or to emit light in structured patterns and estimate depth using structured light techniques.
Computing device 10 also typically includes a six degree of freedom inertial motion unit 19 that includes accelerometers, gyroscopes, and possibly magnometers configured to measure the position of the computing device in six degrees of freedom, namely x, y, z, pitch, roll and yaw.
Data captured by the visible light cameras 18, the depth camera 21, and the inertial motion unit 19 can be used to perform simultaneous location and mapping (SLAM) within the physical environment 9, to thereby produce a map of the physical environment including a mesh of reconstructed surfaces, and to locate the computing device 10 within the map of the physical environment 9. The location of the computing device 10 is computed in six degrees of freedom, which is important to displaying world-locked holograms 50 on the at least partially see through display 20. Without an accurate identification of the position and orientation of the computing device 10, holograms 50 that are displayed on the display 20 may appear to slightly move or vibrate relative to the physical environment, when they should remain in place, in a world-locked position. This data is also useful in relocating the computing device 10 when it is turned on, a process which involves ascertaining its position within the map of the physical environment, and loading in appropriate data from non-volatile memory to volatile memory to display holograms 50 located within the physical environment.
The IMU 19 measures the position and orientation of the computing device 10 in six degrees of freedom, and also measures the accelerations and rotational velocities. These values can be recorded as a pose graph to aid in tracking the display device 10. Accordingly, even when there are few visual cues to enable visual tracking, in poorly lighted areas or texture-less environments for example, accelerometers and gyroscopes can still enable spatial tracking by the display device 10 in the absence of visual tracking. Other components in the display device 10 may include and are not limited to speakers, microphones, gravity sensors, Wi-Fi sensors, temperature sensors, touch sensors, biometric sensors, other image sensors, eye-gaze detection systems, energy-storage components (e.g. battery), a communication facility, etc.
FIG. 2 illustrates a general overview of one possible SLAM algorithm for simultaneous location and mapping by computing device 10. According to the SLAM algorithm, a predictive corrective model 28 is typically applied, which may, for example, be a Kalman filter. In each frame (i.e., timestep), a predicted pose 29 in a next frame is computed based on odeometry data 119A from IMU 19 by an integration engine 27, and then a correction to the predicted pose 29 is computed by the predictive corrective model (e.g., Kalman filter) 28 based on estimated and observed positions of features in the physical environment 9 sensed by sensors such as cameras 18 and depth camera 21, and finally the predicted pose 29 is updated and an updated pose 35 is fed back to the integration engine 27, for computing the predicted pose 29 at the subsequent time step. Additional sensors such as LIDAR, microphone arrays, etc. may also be used to obtain observations of features in the environment which may be used to compute the correction and updated pose by the predictive corrective model 28.
Feature descriptors 111A that describe features such as edges, corners, and other patterns that are detectable through image processing techniques are prestored in a feature library 11 in non-volatile storage device 16. In real time, images 118A and depth images 121A are respectively captured by cameras 18 and depth camera 21, and processed by a feature matching engine 13 executed by processor 12 to detect whether features matching the prestored feature descriptors 111A are present in the captured images 118A, 121A by looking for regions in the captured images that match the feature descriptors 111A. Some of the detected features may have anchors 64 associated with them, since anchors are often positioned on such visually recognizable features in an environment. For each detected feature, the location (e.g., coordinate area) and type of the feature are stored as observation data 17 associated with each frame. It will be appreciated that dozens or hundreds of such features may be recognized in an image, and the collection of these observations 17 of features may be referred to informally as a pointcloud of detected features in the image. Further, for at least selected detected features in the image, a patch 15 from the image is taken surrounding the selected detected features and stored in memory for later recall. This patch 15 is typically a two-dimensional array of pixels or voxels from the region of the captured image, and can be used in future localization steps when the computing device 10 captures images of the selected detected features from another angle, by performing perspective correction on the patch to determine whether (and where) the selected detected features in the perspective corrected patch are present in the subsequent image. The physical manifestations of these selected detected features in the physical environment are also known as anchors. The features matching the feature descriptors 111A, observations 17, and patches 15 for each frame are collectively referred to as feature matching data 113A. The feature matching data 113A typically does not include the depth image 121A or RGB image data 118A. The feature matching data 113A may be stored in non-volatile or volatile memory for certain of the frames, referred to as keyframes, as discussed below. Together, the pose graph 80, feature matching data 113A, surface reconstruction data 82, and keyframes 84 linked by pose graph 80 may collectively be referred to as map data 86. As the computing device 10 moves throughout the physical environment 9, it maps the environment and stores its aggregated knowledge of the environment as map data 86. As will be discussed below, sharing of a portion of this aggregated map data with another device, either directly or through intermediary devices such as a server, can enable other devices to more quickly and accurately localize themselves within the physical environment, saving time and processing power for the other devices.
FIG. 3 shows an example schematic representation of data collected by the cameras 18, depth camera 21, IMU 19, and GPS receiver 25 in the sensor suite of the computing device 10, and processed by the processor 12 of the computing device 10. The data points in FIG. 3 are shown arranged along a time axis, where later times are shown to the right of earlier times. Typically, data is collected periodically from each sensor, at a sampling rate. The sampling rate of the various sensors is depicted to be quantized at the same sampling rate in the figure for ease of illustration (although in practice the sampling rates may differ), and each timestep at which data is sampled from the sensors is referred to as a frame 546. Thus, in the first frame 546A, second frame 546B, and each frame thereafter, image data 118A is collected by cameras 18 (four cameras in this example), depth camera data 121A is collected using depth camera 21, odeometry data 119A is collected by IMU 19, and GPS data 125A is collected by GPS receiver 25.
The processor 12 may use simultaneous localization and mapping (SLAM) techniques, discussed above, based on sensor suite inputs include the image data 118A, depth image data 121A, odeometry data 119A, and GPS data 125A to generate pose graph 80, feature matching data 113A, and surface reconstruction data 82. The pose graph 80 is a directed graph with nodes that are a series of updated poses 35 detected over time. A pose is typically a unit vector with an origin at a predetermined location (x, y, and z) and extending in a predetermined orientation (pitch, yaw, and roll) in the physical space, and is calculated as described in relation to FIG. 2. The origin corresponds to a predetermined fixed location on the computing device, such as equidistant between the left and right displays, and the unit vector extends outward in at a fixed attitude from the display, in a gaze direction of the user. The processor 12 typically generates the pose graph 80 in each frame, but alternatively may generate the pose graph 80 less often, such as every other frame, etc. The surface reconstruction data 82 typically includes a mesh of detected surfaces in the physical environment that have been detected using depth image data 121A and/or stereoscopic analysis of the image data 118A. As discussed above, the feature data 113A typically includes one or more patch, observation, and/or feature descriptor associated with each pose of the pose graph 80.
The processor 12 may generate and store in memory key frame data which includes a plurality of key frames 84. Each key frame 84 includes one pose of the pose graph 80, and thus the key frames 84 are linked by the pose graph 80. Each key frame 84 further includes the feature matching data 113A, which includes one or more (and typically multiple) observations 17, features matching feature descriptors 111A, and associated patch 15 for that frame. The key frame data may further include metadata, which may for example include GPS data 125A, odeometry data 119A, hardware data (e.g., camera lens type), ambient temperature, etc. applicable for the frame. The key frames 84 may be generated at a periodic interval within the series of successive frames, such as every other frame, or every 10.sup.th frame, etc. Alternatively, key frames 84 may be generated at a predetermined spatial interval as the computing device 10 moves through the physical environment 9, such as every 1 or 2 meters.
FIG. 4A is a schematic illustration of a display device generating a current signal fingerprint (hereinafter referred to as “current fingerprint”) and interacting with a server computing device according to the first embodiment of the present disclosure. As explained in more detail below, the server computing device 200 may be used to store signal fingerprint data 54 (hereinafter referred to as “fingerprint data”) and map data sets 58. Computing device 200 may take the form of a server, networking computer, gaming console, mobile communication device, desktop computer, laptop computer, tablet computer, set-top box (e.g. cable television box, satellite television box), or any other type of suitable computing device. In some examples, computing device 200 may comprise an embedded system within a larger electronic or mechanical device or system. In other examples, the computing device 200 may be identical to the display device 30. Additional details regarding the components and computing aspects of the computing device 200 are described in more detail below with respect to FIG. 9.
The server computing device 200 may include a server fingerprint program 214 that may be stored in mass storage 218 of the computing device 200. The server fingerprint program 214 may be loaded into memory 220 and executed by a processor 260 of the server computing device 200 to perform one or more of the methods and processes for sending fingerprint data 54 to the display device 30 responsive to receiving a fingerprint request 52 from the display device 30, and retrieving and sending one or a plurality of corresponding map data sets to a display device 30 responsive to receiving a map data request 48 from the display device 30 as described in more detail below. The server computing device 200 may be configured with a wireless transceiver 230 that wirelessly communicates with the display device 30 to receive fingerprint requests 52 and map data requests 48 from the display device 30 and transmits fingerprint data 54 and matching map data sets 66 to the display device 30. It will be appreciated that matching map data sets 66 are one or a plurality of map data sets corresponding to one or a plurality of closest matching fingerprints. The type of map data applied in the present disclosure is not particularly limited, and will be understood to be any set of data that correlates points in the three-dimensional coordinate space in the physical environment to information that help orient and locate the display device in the three-dimensional space. One possible embodiment of this map data is described in more detail below with respect to FIGS. 6 and 7, and includes keyframes 60, pose graphs 62, and anchors 64 in the form of anchor data. The format of the anchor data is not particularly limited, and may be encoded as unique string identifiers, which identify the anchor, and coordinates, which identify the position of the anchor within the map.
The server computing device 200 may be communicatively coupled to one or more other devices via a wired connection or a wireless connection to a network. In some examples, the network may take the form of a local area network (LAN), wide area network (WAN), wired network, wireless network, personal area network, or a combination thereof, and may include the Internet. In the example of FIG. 4A, server computing device 200 is communicatively coupled to a display device 30 via one or more networks. In other examples the server computing device 200 may be operatively connected with additional devices. The display 20 of the display device 30 may display one or more holograms at a virtual place-located anchor 64 from the vantage point of the display device 30. In some examples, the virtual location of the virtual place-located anchor 64 may be world-locked to a virtual position that is fixed in a three-dimensional coordinate space overlaid upon the real world three-dimensional environment. In other examples, the virtual location of the virtual place-located anchor 64 may be world-locked to a virtual position relative to an object in a real world three-dimensional environment.
The one or more processors of the display device 30 execute a client fingerprint program 38. The display device 30 stores into local memory a local map data 36. The local map data 36 may include the recorded rotational and translational motions of the display device 30 tracked by the visual sensors and/or inertial measurement sensors 18 in the display device 30. The display device 30 may be communicatively coupled to one or more other devices via a wired connection or a wireless connection to a network, but the communicative coupling is not particularly limited, and may encompass any form of data stream, including Wi-Fi, e-mail, external data storage devices, cloud storage devices, etc. In some examples, the network may take the form of a local area network (LAN), wide area network (WAN), wired network, wireless network, personal area network, or a combination thereof, and may include the Internet.
The sensor 18 in the display device receives signals 33 from one or a plurality of signal sources 32. The signal sources 32 may include at least one of a radiofrequency source, an electromagnetic source, a light source, a sound source, and a heat source. The signals 33 may include at least one of electromagnetic signals, gravity, magnetic fields, and temperature. The electromagnetic signals may include at least one of FM signals, GPS signals, cellular signals, and Wi-Fi signals. Referring to FIG. 4B, the signal sources may comprise one or a plurality of wireless access points 132A-C that broadcast beacon frames 133A-C. Each beacon frame 133C comprises an Ethernet header 135C in which a MAC address 137C is included. The body of the beacon frame 133C may comprise an SSID (not illustrated). The beacon frame 133C provides information on a type, make, or model of the originating wireless access point 132C.
Returning to FIG. 4A, responsive to the sensors 18 receiving the signals 33A-C from the signal sources 32A-C, the display device 30 records these detected signals 33A-C and packages them into a current fingerprint that is associated with a current time. The generation of the current fingerprint may be continuously performed in real time, or periodically performed at regular or irregular time intervals in batches. In other embodiments, the current fingerprint may be generated at one or a plurality of predetermined events, such as when the display device 30 is turned on. It will be appreciated that the format of the current fingerprint is similar to the format of the fingerprints 56 in the mass storage 218 of the server computing device, so that the data in the current fingerprint can be compared and contrasted to the fingerprint data 54 of the fingerprints 56.
Independently of the generation of the current fingerprint, the display device 30 also retrieves fingerprint data 54 from the server computing device 200. The server computing device 200 may send the fingerprint data 54 to the display device 30 responsive to a fingerprint request 52 sent by the display device 30. However, it will be appreciated that the server computing device 200 may alternatively or additionally initiate the sending of fingerprint data 54 to the display device 30 even when not prompted to do so by an instruction or request.
Turning to FIG. 4C, one possible example of fingerprints 56, fingerprint data 54, and map data sets 58 is depicted, in which fingerprints correspond to different, distinct areas in the physical environment, and fingerprint data comprises one or a plurality of vectors. In this example, two different fingerprints are provided for an office environment and a home environment, respectively, and each vector is a pair comprising one MAC address and one signal strength. However, it will be appreciated that SSID may also be used in place of a MAC address. The fingerprint data 54A indicates the MAC addresses and signal strengths of the beacon frame signals that would typically be detected or received by a display device in the office environment. Likewise, fingerprint data 54B indicates the MAC addresses and signal strengths of the beacon frame signals that would typically be received by a display device in the home environment. Each fingerprint is linked to its own corresponding map data set: The office fingerprint 56A is linked to the office map data set 58A, while the home fingerprint 56B is linked to the home map data set 58B. It will be appreciated that the format of the fingerprint data is not particularly limited, and may include just one environment variable (just MAC addresses, just SSIDs, or just signal strengths, for example), or a plurality of environment variables. The fingerprint data may include a mixture of different types of signals, including combinations of cellular signals and Wi-Fi signals, or infrared measurements and Wi-Fi signals, for example. There may be overlaps in fingerprint data from two or more different fingerprint–for example, the office fingerprint 56A and the home fingerprint 56B may contain a beacon frame with the same MAC address (in this example, MAC address is shared by the two fingerprints 56A and 56B). In other embodiments, one common fingerprint may be linked with a plurality of map data sets, and/or one map data set may be linked with a plurality of fingerprints. The fingerprint data 54 may comprise predicted signals that a display device 30 is predicted to detect, or actual signals that have already been detected by display devices and recorded in the server computing device 200.
Returning to FIG. 4A, the client fingerprint program 38 executed by the display device 30 compares the current fingerprint against the fingerprint data 54 retrieved from the server computing device 200, then evaluates and ranks candidate fingerprints in the fingerprint data by proximity to the current fingerprint. In this example, the candidate fingerprints are the office fingerprint 56A and the home fingerprint 56B. The comparison of the current fingerprint against the fingerprint data 54 may be performed one-dimensionally or multi-dimensionally. For example, the evaluation of the candidate fingerprints for proximity or degree of disambiguation to the current fingerprint may proceed one-dimensionally according to just signal source, or multi-dimensionally according to signal source and signal strength. Proximity may be evaluated based on statistical analysis for similarity or degree of disambiguation, and may include confidence values and distance functions with measures of distance. Examples of distance functions include cosine distance functions and L.sup.n distance functions. Examples of measures of distance include Hamming distance and Euclidean distance. Confidence algorithms ranking confidence in decreasing order and/or algorithms of degrees of disambiguation incorporating distance functions may also be applied to evaluate proximity. The current fingerprint and/or fingerprint data may undergo filtering to screen out outliers or remove otherwise unwanted data. For example, client fingerprint program 38 may be configured to filter out data from selected wireless access points, or remove MAC addresses with signal strengths below a predetermined threshold. The client fingerprint program 38 may also impose a cap, or a predetermined upper limit on the number of vectors in a current fingerprint or fingerprint data, permitting a maximum of three vectors in fingerprint data for each fingerprint, for example. User input may also be solicited in the course of the statistical analysis. For example, if multiple candidate fingerprints are within a preset threshold of similarity, with no clear determining factor, user input may be solicited to provide input to disambiguate and select between a plurality of candidate fingerprints before concluding the evaluation of the candidate fingerprints. One example of soliciting user input may be to present the user with two choices: a “high confidence” candidate fingerprint, which has a high enough confidence value to be a reasonable match as determined by a predetermined threshold, and a “similar confidence” candidate fingerprint, which is determined as a second highest match as determined by another predetermined threshold, so that user input would disambiguate between the “high confidence” and “similar confidence” candidate fingerprints.
Subsequent to evaluating the candidate fingerprints for proximity to the current fingerprint, the client fingerprint program 38 makes a determination that one of the candidate fingerprints is the closest match to the current fingerprint, then generates and sends a map data request 48 for the map data set corresponding to the closest matching fingerprint. For example, if the client fingerprint program 38 makes a determination that the office fingerprint 56A is the closest matching fingerprint, the program 38 generates and sends a map data request for the office map data set 58A. The server computing device 200 receives the map data request 48, then retrieves and sends to the display device 30 the office map data set 58A as the matching map data sets 66. The display device 30 receives the office map data set 58A, stitches it into the local map data 36 to create an integrated map, and renders one or a plurality of holograms on the display 20 based on the integrated map.
FIG. 5 is a schematic illustration of a display device generating a current fingerprint and interacting with a server computing device according to a second embodiment of the present disclosure. Since the differences between the embodiments of FIGS. 4A and 5 mainly concern the structural differences in the map data sets, fingerprints, and fingerprint data, the detailed description of the display device 30 and server computing device 200 is abbreviated for the sake of brevity. It is to be noted that like parts are designated by like reference numerals throughout the detailed description and the accompanying drawings. In this embodiment, there is only one unified global map data set 158 representing one unified physical environment encompassing multiple different physical environments, rather than a plurality of different, distinct map data sets for different environments. In some examples, the multiple different physical environments in the global map data set 158 may comprise the entire world. Each fingerprint 156 is linked to its own corresponding keyframe 160 rather than to a corresponding map data set. At least two of the keyframes may be linked to a common fingerprint, especially when the two keyframes are physically close to each other. Accordingly, the client fingerprint program 38 may make a determination that a plurality of candidate fingerprints are the closest match to the current fingerprint, then generate and send a map data request 48 for the map data set or plurality of map data sets within the global map data set 158 corresponding to the identified keyframes linked to the closest matching candidate fingerprints. In some examples, the client fingerprint program 38 may identify a neighborhood surrounding the identified keyframes linked to the closest matching candidate fingerprints, which may include anchors, keyframes, and pose graphs that fall within a predetermined distance of the identified keyframes. However, it will be appreciated that the neighborhood may be arbitrarily defined to encompass any shape or size of three-dimensional space surrounding or proximate to the identified keyframes, including neighborhoods that may not necessarily include at least one of the identified keyframes. One possible embodiment of this neighborhood, in relation to the fingerprints and keyframes, is described in detail below with respect to FIG. 7. The server computing device 200 then sends the one or plurality of matching map data sets 66 corresponding to the identified keyframes to the display device 30 for stitching into the local map data 36.
With reference to FIG. 6, an example use case illustrating aspects of the present disclosure will now be presented. As schematically shown in FIG. 6, a first user 302 may be standing in a dining room 306 wearing the first display device 30, which in this example may take the form of HMD device shown in FIG. 1. However, it will be appreciated that the display devices are not particularly limited to HMD devices, and may take any form that allows users to view hologram images overlaid upon the real world three-dimensional environment, such as specialized smartphones and tablet devices, autonomous robots, etc. As noted above, first display device 30 (HMD device) and the second display device 34 may comprise an at least partially see-through display configured to visually augment the views of first user 302 and second user 304, respectively, through the display of the real world three-dimensional environment of the dining room 306. The first display device 30 may generate a virtual model of the dining room 306 using a three-dimensional coordinate space overlaid upon the real world dining room. In the example of FIG. 6, such three-dimensional coordinate space is indicated by the x, y, and z axes.
As described in more detail below, the first display device 30 and second display device 34 also may include program logic of a client fingerprint program 38 that retrieves one or a plurality of map data sets of the dining room 306. The map data sets may be structured as keyframes 60 linked by pose graphs 62, and anchors 64 that are associated with the rendering of holograms 50. In this example, a hologram 50 is projected on a table 308 using a target anchor 64A that is on a picture 310. Another neighboring anchor 64B for another hologram is located in a clock 312 that is in the vicinity of the picture 310. The first user 302 and the second user 304 are roaming about the room 306 as they operate the first display device 30 and the second display device 34, respectively, to view the hologram 50 from various angles in the room 306 from their respective vantage points. As the users roam about the room 306, the sensors 18 within the first display device 30 and the second display device 34 capture visual and/or inertial tracking data and thereby track the rotational and translational motion of the display devices through the sensor devices 18, which observe the three-dimensional rotation and translation of the sensor device 18 to be recorded as poses 62A-G and keyframes 60A-G, which are subsequently stored as local map data 36 in the first display device 30 and local map data in the second device 34. The local map data 36 may be transmitted to the server computing device 200 to be stored in mass storage 218 of the server computing device 200 and later retrieved as one or a plurality of matching map data sets 66 if the map data sets correspond to the one or the plurality of closest matching fingerprints that are indicated by the map data request 48. The poses 62A-G and keyframes 60A-G are described in more detail with respect to FIG. 7. The display devices 30 and 34 subsequently use the retrieved matching map data sets 66 to orient and locate themselves and determine more precise locations of themselves relative to the physical environment of the dining room 306.
Turning to FIG. 7, one possible embodiment of the map data applied in the present disclosure is discussed in more detail. The information for the map data may be generated by at least a sensor device in a plurality of display devices sending sensor data, including the rotational and translational motion tracked by the sensor device, to the computing device 200 in sets that are configured as keyframes 60A-G and a pose graph 80 linking poses 62A-H. Here, a display device and its keyframes and pose graphs are depicted, but other embodiments may feature two or more display devices in close proximity to each other, each with their own trajectories of key frames and pose graphs. Also contained in the map data are a plurality of virtual place-located anchors 64, including the target anchor 64A and a neighboring anchor 64B at world-locked virtual locations with known three-dimensional coordinates in the physical environment. These anchors may include visibly conspicuous features in the physical environment, such as the picture 310 and clock 312 illustrated in FIG. 6. Poses 62A-H, depicted as small arrows in the pose graph 80, are typically unit vectors that point in the direction of a fixed straight-ahead gaze out of the display of display device, as described above, and the pose graphs record the position of the poses in three-dimensional space over time. Individual keyframes 60A-G are linked to each other in pose graph 80, which links poses 62A-H. The pose graph thus includes a plurality of such poses linked to each other in a directed graph so as to track the changes in pose as the display device travels through the three-dimensional coordinate space of the physical environment. The pose graph 80 forms a linear trajectory of map data that the display device leaves behind to store as local map data and subsequently sends to the server computing device 200 for compilation and analysis as map data as the display device travels through three-dimensional coordinate space over time.
Keyframes 60A-G contain sets of information that can be used to improve the ability of the display device to ascertain its location, and thus help render holograms in stable locations. As discussed above, examples of data included in keyframes 60A-G include metadata, observations and patches, and/or image feature descriptors. Metadata may include the extrinsic data of the camera, the time when keyframe was taken, gravity data, temperature data, magnetic data, calibration data, global positioning data, etc. Observations and patches may provide information regarding detected feature points in a captured image, such as corners and high contrast color changes that help correct the estimation of the position and orientation of the display device, and accordingly help better align and position the display of a holographic image via display 20 in three-dimensional space. Image feature descriptors may be feature points, sometimes efficiently represented in a small data set, in some examples as small as 32 bytes, that are used the feature matching engine 13 described above to quickly recognize features in the real time captured images 118A and depth images 121A, to accurately estimate the position of the display device, and thus accurately render the hologram on the map of the physical environment.