Attorney Docket No.: 025714-010100US
`Client Reference No.: A2908
`
`PROVISIONAL
`
`PATENT APPLICATION
`
`REMOTE POSITIONING
`
`Inventor:
`
`Peter France, a citizen of New Zealand, residing at
`18 Highcrest Heights, Westmorland
`Christchurch, 8025 New Zealand
`
`Assignee:
`
`Trimble Navigation Limited
`935 Stewart Drive
`Sunnyvale, CA 94085
`
`Entity:
`
`Large
`
`KILPATRICK TOWNSEND & STOCKTON LLP
`Two Embarcadero Center, Eighth Floor
`San Francisco, California 94111-3834
`Tel: 650-326-2400
`
`

`

`PATENT
`
`Attorney Docket No.: 025714-010100US
`Client Reference No.: A2908
`
`REMOTE POSITIONING
`
`BACKGROUND
`
`[0001] Mapping and geospatial information system (GIS) field workers need to collect
`
`positions of assets. Global navigation satellite systems (e.g., GPS) are the tool of choice, but
`
`sometimes remote positioning is desired. For example, GPS may not be accurate enough near
`
`tall buildings or undertrees or overhangs, or the field worker may want to avoid having to
`
`physically visit assets in the middle of a road or across a stream.
`
`[0002] A simple and efficient workflow is desired where only sub-meter or better accuracy is
`
`10
`
`required (that is, cm or mm accuracyis not required). Many commonscenariosonly require a
`
`short range of 2-15m. For example, if the feature is undera tree or overhang, a good sky-view
`
`can usually be obtained only a few meters away.
`
`[0003]
`
`Laser rangefinders are commonly used, most commonlyasa single distance reading
`
`combined with a magnetic compassbearing and inclinometer. Laser rangefindersare relatively
`
`15
`
`difficult to aim, however, as they have high-poweroptical sights. Sometimesit is difficult to tell
`
`if you are measuring a distance to a pole or a distance to an object that is 20m behindthe pole.
`
`Other drawbacksare that the rangefinder has to be charged, carried and used, with the attendant
`
`bulk, weight, cost, and power consumption.
`
`[0004] Other techniques(e.g., bearing-bearing intersection) do not require a laser, but they do
`
`20
`
`require a two-shot workflow.
`
`[0005]
`
`Total stations are sometimes used. While more accurate, the cost, time, and complexity
`
`are unattractive.
`
`SUMMARY
`
`25
`
`[0006]
`
`This technique defies the conventional wisdom that one-shot remote positioning must
`
`include a distance measurement. The method gives usable accuracy over short ranges as long as
`
`a few conditions hold true. You could typically expect sub-meter results to a range of 7m, and
`
`sub-2m results to a range of 15m.
`
`

`

`[0007]
`
`The operator has a GPS handheld with an aiming device (usually an integrated camera),
`
`magnetic compass,andtilt sensors (but no laser rangefinder).
`
`[0008] Only oneshotis required, and the user interface may be simple and intuitive —
`
`“Location by pointing.”
`
`DETAILED DESCRIPTION
`
`[0009]
`
`FIG. 1 is a simplified diagram illustrating one-shot remote positioning in accordance
`
`with an embodimentof the present invention. An operator 102 determinesa distance d to a point
`
`106 using a measurement device 104. The distance d is determined by pointing the measurement
`
`10
`
`device 104 at the point 106 using an aiming device and determininga tilt of the measurement
`
`device 104 using tilt sensors. The measurement device 104 is held at a known height h above the
`
`ground, and the distance d is computed using the equation:
`
`
`
`d= h
`tan 0
`
`(1)
`
`[0010]
`
`The measurement device 104 may also be used to determine a position of the point 106.
`
`15
`
`In this embodiment, a position measurementdevice (e.g., GPS) may be used to determine a
`
`position of the measurement device 104, and a bearing measurement device(e.g., electronic
`
`compass) may be used to determine a bearing of the measurement device 104. The position
`
`measurement device and the bearing measurement device may be integrated with the
`
`measurement device 104. The position of the point 106 may be determined usingthe position of
`
`20
`
`the measurement device 104, the bearing of the measurement device 104, and the distance
`
`determined using equation (1).
`
`[0011]
`
`The workflow can be simple. First, select the “Location by pointing” function in the
`
`user interface of the measurement device 104. A live camera view may be shownon a screen of
`
`the measurement device 104. Next, tap on the screen to identify the target. The position of the
`
`25
`
`remote point is instantly calculated using equation (1). An error estimate mayalso be calculated.
`
`[0012]
`
`The user interface may show the newly calculated position as a graphic overlaid on the
`
`live camera view as shownin FIG. 2. This enables instant intuitive feedback, especially if you
`
`walk a few meters to the side and check that the calculated position lines up with the real object.
`
`

`

`[0013]
`
`Ifthe accuracy of the measurementis not sufficient, the results may be fine-tuned by
`
`taking a new measurement from a position a few meters to the side. This supplies the data
`
`required to calculate a bearing-bearing intersection, giving a more accurate position. The user
`
`interface could be as easy as dragging the graphic icon to the correct location on the screen.
`
`Alternative workflows for target selection
`
`[0014]
`
`It may notbe easy to tap a (moving) target on the screen while a live camera view is
`
`shown, especially if your handsare not steady or the target is not close.
`
`[0015] Another method showscross-hairs in the middle of the live camera view. The operator
`
`maypress a button on the measurement device 104 whenthe target is in the cross-hairs.
`
`10
`
`[0016]
`
`Possibly the easiest and most accurate targeting methodis to press a button to freeze the
`
`camera view and record the sensors measurements. Then you can tap the target moreeasily, asit
`
`is not moving aroundthe screen.
`
`[0017] Digital zoom can optionally be used during all of the above methods. Zoom increases
`
`targeting precision. Zoom may makeit harder to aim in live camera mode; therefore zoom may
`
`15
`
`workbest along with freeze mode.
`
`[0018] Accuracy generally depends on:
`
`(1) how close the target point is to groundlevel; (2)
`
`how flat the ground is between the operator and the target; and (3) how accurately the height of
`
`the device is known.
`
`[0019]
`
`Fortunately, these conditions are met in a large proportion of scenarios: (1) The height
`
`20
`
`at ground level is usually the elevation of commonfeatures suchastrees, poles, buildings,
`
`hydrants etc.; (2) The groundis flat in many locations, especially over short ranges (10m orso).
`
`Evenin hilly areas, it is quite possible to stand across the slope (rather than up or down the
`
`slope); (3) The user normally sets the antenna height, even for handheld operation. Experiments
`
`have shownthat users don’t normally vary the device height by more than +/-4cm in normaluse.
`
`25
`
`[0020] With the assumption that the target lies on a planar surface at groundlevel, target pixel
`
`identification plus GPS andorientation sensors can be used to describe a ray from the camera to
`
`the target. The remote position is simply the intersection of the ray and the groundsurface as
`
`shown in FIG. 1.
`
`[0021]
`
`The main error sourcesare:
`
`

`

`Magnetic compass error
`Target elevation variation
`Device height variation
`Tilt error
`
`GPS error
`
`Lensdistortion
`
`[0022]
`
`The following analysis assumes that the measurement device 104 uses low-cost
`
`commercially available sensors, such as those available in some Trimble® GeoExplorer® series
`
`devices.
`
`10
`
`[0023]
`
`In these devices, magnetic compass error may nominally be about 1 degree in good
`
`conditions, but could be up to tens of degrees near very severe magnetic disturbances. A five
`
`degree upperlimit is reasonable in sensible working conditions. The effect of compasserrorvs.
`
`distance to the target is plotted in FIG. 3. Note that compasserror also affects conventional
`
`techniques, such as laser-plus-compassand bearing-bearingintersection.
`
`15
`
`[0024]
`
`Ifthe target is not at ground level (defined as the flat plane at the operator’s feet), then
`
`somepositional error will accumulate. The magnitude of this error is proportional to the distance
`
`to the target as shown in FIG. 4.
`
`[0025]
`
`It can be assumedthat the operator has entered a correct antenna height. The software
`
`can issue a warningif the value is near zero. Experiments have shownthat operators typically
`
`20
`
`hold the device within +/-4cm of their nominal antenna height. The resultant effect on position
`
`error is proportionalto the distance to the target as shownin FIG. 4.
`
`[0026]
`
`Tilt error is typically small, generally about 0.1 degree. The effect of tilt error increases
`
`with distance to the target, as the target nears the horizon. This is shownin FIG. 5.
`
`[0027] With Trimble® GeoExplorer® series devices, GPS erroris typically <10cm in open
`
`25
`
`environments. Even when using a GPS that has only sub-meter accuracy, often the accuracy is
`
`better than 50cm in good conditions. Note that the GPS error is the sameforall the other
`
`remote-positioning methods such as laser rangefinding and bearing-bearing intersection.
`
`[0028]
`
`Lensdistortion can introduce error whenthe target is not in the centre of the image.
`
`This error can be compensated by using a lens distortion calibration, either for the model ofthe
`
`30
`
`lens used, or for the specific lens in each device. Most operatorsare likely to intuitively aim the
`
`

`

`cameraso the target is in the centre of the image, so it can be assumedthat this is a minimal
`
`effect.
`
`[0029]
`
`Humanerroris not accounted for in this analysis, on the groundsthat a digital zoom
`
`can be usedto precisely select the target pixel, and any aimingerror is generally swamped by
`
`larger errors. Similarly it is assumed that the camera, compass, andtilt sensors are calibrated, so
`
`any misalignment between them is knownprecisely enough to minimizeerror.
`
`[0030]
`
`The combination of these error sources is shown in FIG. 6 using typical values
`
`(extreme error values are not considered). GPSerror is not included in this example. Thetotal
`
`horizontal error is typically sub-meter to a range of 7m, and less than 2m outto a range of 15m.
`
`10
`
`[0031]
`
`Ifthe original shot is not accurate enough,the user interface can offer the option to
`
`improve the position. There are several possibilities, including:
`
`e Move a few meters to the side and take another shot. A
`bearing-bearing calculation is performed, removing the
`assumptions aboutflat ground and correct antenna height,
`therefore giving muchbetter accuracy over longer ranges.
`
`e Guide the user to walk to the target point and press a button,
`which captures a barometer reading. The barometer
`differential gives a height-change. This improvesthe
`horizontal and vertical accuracy for scenarios where the ground
`is notflat.
`
`15
`
`20
`
`[0032] Widths and heights can also be calculated. A target location can first be calculated as
`
`explained above. Then the device can be aimedat other points to calculate the width or height to
`
`the original object. The assumptionis that the width or heightis in the same plane as the focal
`
`plane.
`
`25
`
`[0033]
`
`The benefits to the user include:
`
`Simple intuitive operation.
`No laser required.
`Less weight and easier aiming than laser rangefinders.
`Higher accuracy available from a second shot a few meters away.
`
`30
`
`[0034] One disadvantages may bethat rangeis limited, and accuracy degrades with longer
`
`distances.
`
`

`

`[0035]
`
`In real hardware, the camera focal point and GPS antennaphase center are physically
`
`separated. This could be modeled to achieve moreprecise results if necessary.
`
`[0036]
`
`The compass and accelerometers generally require calibration. The camera also
`
`generally requires calibration, so that a misalignment value can be applied if necessary (because
`
`the center of the imageis notlikely to be at zero pitch and north).
`
`[0037]
`
`The camera and lens can influence accuracy. Expensive metric cameras can be used to
`
`achieve highly accurate results. But it is feasible to achieve accuracy useful to mapping and GIS
`
`customers using inexpensive consumer camera modules. A lens distortion measurement may be
`
`performed for a sample of production camera modules, and the resultant model may be used to
`
`10
`
`apply a software correction. This assumesthat there is not a large variation in lenses from
`
`camera to camera.
`
`[0038]
`
`The entire system includeserrors from the optical sensors and from the GPS
`
`measurements.
`
`[0039]
`
`The GPS positional accuracy can be sub-centimeter. MGIS products today add
`
`15
`
`dithering to 10cm,but it is feasible to use the pre-dithered solutions as input to the remote
`
`positioning calculations. The eventual position output could be dithered if accuracy is under
`
`10cm.
`
`[0040]
`
`If GPS post-processing is necessary, then some information must be stored in order to
`
`repeat the remote positioning measurement. This information consists of the yaw,pitch, roll, and
`
`20
`
`the target pixel coordinates, in addition to the GPS position/observations. The calibration data
`
`mayalso be required.
`
`