METHOD FOR ASSISTED ROAD EXTRAPOLATION FROM IMAGERY

Abstract

A method for assisted map extrapolation computing establishing a database of geographic images and establishing a database of geospatial datasets extracted from the database of geographic images The method provides an unskilled user interface hosted remotely or self contained within a computing device loaded with imagery and analysis software wherein an unskilled user may select a geographic area The method superimposes any existing data from the geospatial datasets onto the geographic images for a geographic area and displays it to the unskilled user A set of algorithm parameter controls are provided wherein the individual may teach image analysis software how to pick transportation network features such as roads or other objects such as building footprints from the imagery The geospatial datasets are then updated with the features identified by the unskilled user assisted software

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates generally to a system and method for extrapolating information from geographic images for the generation of digital maps. More particularly, this invention relates to a system and method for utilizing unskilled user assistance to improve semi automated map extrapolation.

2. Related Art

Mapping and navigational devices require complex geospatial map data for operation. With the constantly changing landscape of modern society, it has become increasingly difficult to maintain accurate and up to date map data. The digital maps are obtained by various methods, including high resolution imagery taken from space, as well as ortho-rectified images taken from land-based vehicle. One popular methodology utilizes high resolution multi-spectral images that is available for virtually all of the earth surface.

There exists a wide variety of systems and methods for automatically or semi-automatically extracting road vectors or spatial objects from this imagery. However, the imagery is typically of varying quality and exposure. Furthermore, the road or other features in the imagery may be obscured by trees, clouds, and other sources. Consequently the quality of automatically extracted objects from imagery is generally poor. Better quality is derived by a semi-automated process where a skilled technician provides seed information that assists a process in distinguishing roads or other objects on the imagery.

A relatively new approach to improving the map data has been the use of crowd sourcing. Present methodologies utilize passive position sensors such as GPS sensors in a Personal Navigation Device (PND). As a user operates the PND, the device passively records GPS traces along the travel path. This passive data is uploaded to the map developers for use in improving their databases. In many areas, however, the utilization of PND or GPS equipped smart phones is limited and therefore passive approaches to crowd sourcing are not available. In addition, even where such data is available it is only possible to extract the geometry and not detailed attributions such as road names, speed limit.

Detailed imagery of the entire world is readily available, and roads or other transportation corridors can be readily identified on the images by humans, but in many cases are difficult for a computer to identify. As mentioned above, a skilled operator can teach a computer how to pick roads from imagery. However, with an appropriately designed interface, an unskilled operator is able to perform this function. What is needed is a method and system for positively engaging end users of navigation systems or digital map displays to perform the function of teaching a computer to extract roads and other objects from imagery that was here-to-fore performed by skilled technicians. Furthermore, such a method would allow for accurate modifications of existing geometry and would provide for the addition of detailed attributions that would otherwise be unavailable. It is therefore an object of the present invention to provide a system and method for having end users operate semi-automated road extraction software from high resolution imagery.

SUMMARY OF THE INVENTION AND ADVANTAGES

These and other objects and advantages are achieved according to one aspect of the invention by a method for assisted map extrapolation comprising establishing a database of geographic images and establishing a database of geospatial datasets extracted from the database of geographic images. The method provides an unskilled user interface hosted remotely or self contained within a computing device loaded with imagery and analysis software wherein an unskilled user may select a geographic area. The method superimposes any existing data from the geospatial datasets onto the geographic images for a geographic area and displays it to the unskilled user. A set of algorithm parameter controls are provided wherein the unskilled user may teach image analysis software how to pick transportation network features such as roads or other objects such as building footprints from the imagery. The geospatial datasets are then updated with the features identified by the unskilled user assisted software.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other features and advantages of the present invention will become more readily appreciated when considered in connection with the following detailed description and appended drawings, wherein:

FIG. 1 is a schematic view of a method and system for assisted map extrapolation 10 in accordance with the present invention;

FIG. 2 is an illustration of an unskilled user selecting a portion of a geospatial object for use as seed points in accordance with the system described in FIG. 1;

FIG. 3 is a detailed view of a first pass geography extrapolation superimposed on a geographic image in accordance with the present invention;

FIG. 4 is a detailed view a unskilled user moving an algorithm parameter control and the results of a multi-pass geography extrapolation superimposed on the geographical image shown in FIG. 3;

FIG. 5 is a detailed view of an unskilled user accessing a manual cleanup module to erase a geospatial object such as a road;

FIG. 6 is a detailed view of an unskilled user accessing a manual cleanup module to move a geospatial object such as a road; and

FIG. 7 is a detailed view of an unskilled user accessing a manual cleanup module to add a geospatial object such as a road;

FIG. 8 is a detailed illustration of an attribution input module in accordance with the present invention;

FIG. 9 is an illustration of an alternate embodiment of the present invention, the embodiment utilized selected point parings; and

FIG. 10 is a schematic view of the alternate method and system for assisted map extrapolation 10 as illustrated in FIG. 9 and in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

Referring to the Figures, wherein like numerals indicate like or corresponding parts throughout the several views, this invention pertains to digital transportation network navigation systems and to the digital maps, databases and devices used by navigation systems. By “digital transportation network”, it is meant to include digital mapping systems for various established transportation networks, including various roadway surfaces for motorized and non-motorized vehicles, walking, biking, skiing and hiking trails, and other established routes along which users of navigations devices travel for business or pleasure. For purposes of this invention, the term “roadway” is intended to be used in a most general way and to be inclusive of all paved and unpaved roads, paths, trails and the like for use by any type of vehicle or a pedestrian.

Referring now to FIG. 1, which is a schematic illustration for a system of assisted map extrapolation 10 in accordance with the present invention. The system utilizes a database of geographic images 12, at the same scale and using the same map projection so as features on the image are coincident with features displayed from the transportation database. These images 12 may be accumulated through a variety of sources including the use of high-resolution multi-spectral images of the earth surface. The system 10 further includes a geo-spatial map dataset database 14. In at least one embodiment, the initial geo-spatial map data-set database 14 is comprised of geospatial object data 15 automatically extrapolated from the geographic images database 12 using imaging processing algorithms. The geospatial object data 15 defines a plurality of geospatial objects 17 such as roads, buildings, bodies of water, etc.

The system 10 contemplates the use of a unskilled user interface 16 wherein an unskilled user 11 may interface with the system for assisted map extrapolation 10. The term unskilled user 11 is intended to encompass any individual without map extrapolation experience. In one embodiment, it is contemplated that the unskilled user 11 may be a member of the general public. In other embodiments, the unskilled user 11 is one of a select group of individuals with proper motivation but no formal map training. It is contemplated that the unskilled user interface 16 may be web based or contained within a stand-alone computer system loaded with imagery.

The unskilled user interface 16 includes a first interface element 18 to prompt the unskilled user to select a geographic area 20. In one embodiment, it is conceived that the unskilled user 11 will utilize the unskilled user interface 16 to identify regions he/she is familiar with in order to review the veracity of the geo-spatial map dataset database 14 as it relates to the geographic images 12. It is further conceived that the geographic area 20 could be automatically selected to a region near the unskilled user's location. Once a geographic area 20 has been identified, the system 10 utilizes an extrapolation algorithm module 22 to perform a first pass geography extrapolation 24 wherein the module 22 extrapolates geospatial object data 15 from the geographic image 12 of the selected geographic area 20. It is contemplated that the first pass geography extrapolation 24 is performed utilizing initial seed points 26 provided to the extrapolation algorithm module 22.

It is contemplated that the seed points 26 can be obtained in a variety of fashions. In one embodiment, the unskilled user 11 is prompted to select portions 28 of the geographic image 12 to identify a particular geospatial object 17, such as a portion of a road or building. (FIG. 2). The visual parameters 30 of the selected portion 28 are utilized by the extrapolation algorithm module 22 to begin the first pass geography extrapolation 24. In another approach, existing data from the geo-spatial map dataset database 14 may be utilized as the initial seed points 26. In still another embodiment, where the existing geo-spatial map dataset database 14 information is fairly complete, it is contemplated that the existing geospatial object data 15 for the selected geographic area 20 may be utilized without utilizing a first pass geography extrapolation 24.

The interface 16 is then configured to display the first pass geography extrapolation 24 superimposed on the geography image 12 of the selected geography area 20 (FIG. 3). The interface 16 further provides the unskilled user 11 with at least one algorithm parameter control 32 which controls at least one parameter 34 utilized by the extrapolation algorithm module 22 when analyzing a geographic image 12. An individual control 32 may alter a single parameter 34 or a plurality of parameters. It is contemplated that the control 32 may adjust the aggressiveness of the extrapolation algorithm module 22. The term aggressiveness is known in the art to refer to the ability of an algorithm to associate portions of an image with an object. For lenticular objects such as roads the algorithm will generally utilize a weighted graph constructed from the line segments identified as roads and the gaps between lines segments where the road is obscured by trees and other impediments. The algorithm will attempt to fill in these gaps. Various parameters can be set to determine how aggressively the algorithm will fill in the gaps. Depending on how these are set, the algorithm may be too aggressive and determine certain linticular features are roads when they are not and/or link road segments together when they should not be. Alternatively, they could not be aggressive enough and miss many roads.

In addition, although a single control 32 may be utilized, the present invention contemplates the use of multiple controls 32 each controlling a separate algorithm parameter 34. It is contemplated that these algorithm parameters 34 can include a variety of parameters including, but not limited to, object dimension, intensity, hue, saturation, and object color. Image analysis software used to identify roads or other features from the image relies on finding pixels or groups of pixels with features that are indicative of a road. These features might consist of intensity, huge, saturation, and color of the pixels that make up a road. They may also consist of lenticular groupings of the preceding features—for example, roads may be defined by a group of pixels no more than 4 pixels wide, but as long or longer than 10 pixels. All pixels may, for example, be the same or similar huge of grey. Adjacent imagery may be subtlety different from one image to the next—because of such factors as the sun angle was different when two adjacent images were recorded or the camera used had different characteristics. As the unskilled user 11 adjusts the algorithm parameter control 32, such as my moving a slider bar 36, these changes are sent to the extrapolation algorithm module 22 (FIG. 4). An unskilled user 11 does not need to know necessary what parameters 34 affect automated picking of roads from imagery, just that they do affect the accuracy of the picks. A system of controls 32 can be provided consisting of slider bars or virtual dials that adjust factors affecting how the software performs. One slide bar could be color. A unskilled user 11 would move the slide bar 36 to, for example, change the color from grey to more brown. The extrapolation algorithm module 22 would detect the change and recalculate the road picks, based on the change—then redisplay the road picks for the user. This change may have improved the road picks or made them worse. The unskilled user 11 then has the option to further tweek the color or some other parameter to see if the road picks improves or gets worse. The unskilled user 11 does not need to know what parameters 34 the controls affect, he/she simply needs to observe the effect of the change of the control 36. With practice, the use would develop skill in knowing which controls have what effect.

The extrapolation algorithm module 22 utilizes that information from the algorithm parameter control 32 to perform a second pass geography extrapolation 38 and display it superimposed on the geographic image 12. It is contemplated that the second pass geography extrapolation 38 comprises a multi-pass geography extrapolation such that the unskilled user 11 can adjust the controls 32 and visually watch the adjustments until the extrapolation most closely matches the geographic image 12. This allows an unskilled user 11 to improve the extrapolation without any knowledge of how the extrapolation methodologies actually operate. In an alternate embodiment, it is contemplated that the controls 32 comprise prompts for the unskilled user 11 identify additional portions of a particular geospatial object 17 and this is utilized by the multi-pass extrapolation in a similar fashion. The unskilled user 11 will continue to pick more points on an object, such as a road, until the extrapolation most closely matches the geographic image 12.

The system 10 further contemplates the use of a manual cleanup module 40 once the aforementioned extrapolations are satisfactory to the unskilled user 11 (FIG. 1). The manual cleanup module 40 is intended to allow the unskilled user 11 to manually adjust a particular geospatial object 17 not properly addressed by the extrapolations. It is contemplated that this manual adjustment can include, but is not limited to, deleting the particular geospatial object (FIG. 5), moving the particular geospatial object (FIG. 6), or inserting a new geospatial object (FIG. 7). This allows the unskilled user 11 to address issues not dealt with through algorithm adjustment. It is contemplated that this may arise when the view of geospatial objects 17 is impeded by other topography such as trees, shade, buildings, etc.

Finally, the system 10 further contemplates the use of an attribution input module 42 (FIG. 8). The attribution input module 42 allows the unskilled user 11 to enter non-geographic attributes 44 into the geospatial dataset data base 14. This can include information such as road names, building names, attraction information, road travel direction, speed limits, etc. This is highly valuable as this information is largely unavailable from the geographic images 12. Once the unskilled user 11 is satisfied, the adjustments made by way of the controlled extrapolation in addition to the non-geographic attributes 44 can be utilized to improve the geospatial dataset database 14.

In an additional embodiment of the present invention, the system 10 contemplates the database of geographic images 12 to comprise a database of ortho-rectified geographic images (FIGS. 9 and 10). The unskilled user 11 is still prompted to identify a geographic area 20 that is familiar. In this embodiment, however, the unskilled user interface 16 displays street information and/or the geospatial database 14 information superimposed onto the ortho-rectified geographic image 12 of the selected geographic area 20. The unskilled user 11 is prompted to select a plurality of point pairs 48, each comprising a point from the geospatial map data 14 and a point from the ortho-rectified geographic image 12 that should be in the same location but are not. In this embodiment the extrapolation algorithm module 22 skews/warps the spatial database 14 based on one of a variety of conflation techniques so that the point pairs 48 coincide and intermediate points 50 in between are skewed based on a weighting relative to the nearby point pairs 48. The technique is applied repeatedly to maximize the correspondence of the point pairs 48 and minimize the distortion of straight lines or curves in the geospatial dataset database 14. In this embodiment, the unskilled user 11 may continue to pick point pairs 48 until the visual image of the superimposed geospatial dataset database 14 over the ortho-rectified geographical image 12 is satisfactory. It is contemplated that the unskilled user 11 will be then be afforded the opportunity to access the manual cleanup module 40 and the attribution input module 42 to make further improvements.

It is contemplated that a variety of motivational techniques could be employed to encourage this public interaction. The first is simply a personal motivation. Local businesses in new developments will have an active desire to see their local areas correctly mapped such that patrons can easily find them. Rural individuals may find that correcting local maps may improve deliveries. Contests and give away promotions may be used to elicit public response. Finally, a personal credit may be given to individuals making corrections and reflected within the geo-spatial data-set database 14 such that pride may provide sufficient motivation.

The foregoing invention has been described in accordance with the relevant legal standards, thus the description is exemplary rather than limiting in nature. Variations and modifications to the disclosed embodiment may become apparent to those skilled in the art and fall within the scope of the invention. Accordingly the scope of legal protection afforded this invention can only be determined by studying the following claims.

Claims

1. A method for assisted map extrapolation comprising the steps of:

establishing a database of geographic images;

establishing a database of geospatial datasets comprised of geospatial object data; providing an unskilled user interface wherein an unskilled user may select a geographic area;

performing a first pass geography extrapolation using an extrapolation algorithm module wherein said geospatial object data is extrapolated from said geographic image of said geographic area using initial seed points;

displaying said first pass geography extrapolation superimposed onto said geographic image of said geographic area;

providing at least one algorithm parameter control wherein the unskilled user may adjust the parameters of said extrapolation algorithm module for a second pass geography extrapolation;

displaying the second pass geography extrapolation superimposed on geographic image of said geographic area;

allowing said unskilled user to adjust said algorithm parameter control until said second pass geography extrapolation accurately reflects said geographic area; and

updating said database of geospatial datasets with the unskilled user adjustments.

2. A method as set forth in claim 1 wherein said initial seed points are obtained by:

utilizing a cursor to identify at least a portion of a geospatial object in said geographic image; and

using said portion as said seed points for said first pass geography extrapolation.

3. A method as set forth in claim 1 wherein said seed points are obtained by: utilizing existing data within said database of geospatial datasets as said seed points.

4. A method as set forth in claim 1 wherein prior to performing said first pass geography extrapolation the method includes:

displaying any pre-existing geospatial object data within said geospatial datasets superimposed onto said geographic image of said selected geographic area.

5. A method as set forth in claim 1 wherein said at least one algorithm parameter control comprises:

adjusting the aggressiveness of said extrapolation algorithm module.

6. A method as set forth in claim 1 wherein said at least one algorithm parameter control comprises:

prompting the unskilled user to identify portions of said geographic image representative of a particular geospatial object, wherein visual parameters of said identified portions are processed by said extrapolation algorithm module.

7. A method as set forth in claim 1 wherein said at least one algorithm parameter control comprises:

a plurality of controls, each of said plurality of controls adjusting a separate algorithm parameter.

8. A method as set forth in claim 7, wherein said plurality of controls adjust algorithm parameters taken from the group of object dimension, intensity, hue, saturation, and object color.

9. A method as set forth in claim 1, further comprising:

allowing said unskilled user to utilize a manual cleanup module wherein said unskilled user can directly input changes to a particular geospatial object.

10. A method as set forth in claim 9, wherein said direct input changes comprise erasing said particular geospatial object, moving said particular geospatial object, or inserting a new geospatial object.

11. A method as set forth in claim 1, further comprising:

providing an unskilled user attribution input wherein said unskilled user can attach non-geographic attributes to a particular geospatial object.

12. A system for assisted map extrapolation comprising:

a database of geographic images;

a database of geospatial datasets comprised of geospatial object data;

an unskilled user interface wherein an unskilled user may select a geographic area;

an extrapolation algorithm module configured to perform a geography extrapolation wherein said geospatial object data is extrapolated from said geographic image using a plurality of algorithm parameters;

at least one algorithm parameter control wherein said unskilled user may adjust said algorithm parameters for a multi pass geography extrapolation, said extrapolation algorithm module configured to perform said multi pass geography extrapolation and to display said multi pass geography extrapolation superimposed on geographic image of said geographic area;

wherein said unskilled user is prompted to adjust said algorithm parameter control until said multi pass geography extrapolation accurately reflects said geographic area; and

updating said database of geospatial datasets with the unskilled user adjustments.

13. A system as described in claim 12, wherein said extrapolation algorithm module is further configured to:

perform a first pass geography extrapolation wherein said geospatial object data is extrapolated from said geographic image of said geographic area using initial seed points; and

display said first pass geography extrapolation superimposed onto said geographic image of said geographic area.

14. A system as described in claim 12, wherein said at least one algorithm parameter control comprises a slider control adjusting the aggressiveness of said extrapolation algorithm module.

15. A system as described in claim 12, wherein said at least one algorithm parameter control comprises:

a plurality of algorithm parameter controls each adjusting a separate algorithm parameter.

16. A method for assisted map extrapolation comprising the steps of:

establishing a database of ortho-rectified geographic images;

displaying one of said ortho-rectified geographic images with geospatial map data superimposed, said geospatial map data comprised of a plurality of geospatial objects;

prompting an unskilled user to identify several point pairs, each comprising a point from the geospatial map data and a point from the ortho-rectified geographic image that should be at the same location;

adjust said geospatial map data utilizing a conflation techniques such that said point pairs coincide and a plurality intermediate points are adjusted on a weighting average relative to nearby point pairs;

repeat said adjustment to maximize the coincidence of point pairs and minimize the distortion of straight lines and curves in said geospatial map data; and

update said geospatial map data with said adjustment.

17. A method as set forth in claim 16 further comprising:

providing an unskilled user attribution input wherein said unskilled user can attach non-geographic attributes to a particular geospatial object.

18. A method as described in claim 16, further comprising:

allowing said unskilled user to utilize a manual cleanup module wherein said unskilled user can directly input changes to a particular geospatial object.

19. A method as described in claim 18, wherein said direct input changes comprise erasing said particular geospatial object, moving said particular geospatial object, or inserting a new geospatial object.