Industrial vehicle with feature-based localization and navigation

• US 10,473,469 B2
• Filed: 11/02/2018
• Issued: 11/12/2019
• Est. Priority Date: 09/04/2015
• Status: Active Grant

• Current Assignee: Crown Equipment Corporation
• Original Assignee: Crown Equipment Corporation
• Inventors: Bell, Mark
• Primary Examiner(s): Wang, Carol
• Application Number:
  US16/179,793
• Publication Number:
  US 20190072396A1
• Time in Patent Office: 375 Days
• CPC Class Codes: G01C 21/206 specially adapted for indoor navigation
• CPC Class Codes: G05D 1/0246 using a video camera in combination with image processing means
• CPC Class Codes: G06F 18/213 Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30241 Trajectory
• CPC Class Codes: G06T 2207/30244 Camera pose
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/52 Scale-space analysis, e.g. wavelet analysis multi-scale boundary representations G06V10/42
• CPC Class Codes: G06V 20/10 Terrestrial scenes scenes under surveillance with static cameras G06V20/52; scenes perceived from the exterior of a vehicle G06V20/56; scenes perceived from the interior of a vehicle G06V20/59
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle

Abstract

An industrial vehicle is provided comprising a drive mechanism, a steering mechanism, a vehicle controller, a camera, and a navigation module. The camera is communicatively coupled to the navigation module, the vehicle controller is responsive to commands from the navigation module, and the drive mechanism and the steering mechanism are responsive to commands from the vehicle controller. The camera is configured to capture an input image of a warehouse ceiling comprising elongated skylights characterized by different rates of image intensity change along longitudinal and transverse axial directions, and ceiling lights characterized by a circularly symmetric rate of image intensity change. The navigation module is configured to distinguish between the ceiling lights and the skylights and send commands to the vehicle controller for localization, or to navigate the industrial vehicle through the warehouse based upon valid ceiling light identification, valid skylight identification, or both.

Claims

1. A system comprising a vehicle camera, and a navigation module comprising memory and a processor, wherein:

the vehicle camera is communicatively coupled to the navigation module and is configured to capture input images;

the navigation module is configured to distinguish between object types in the input images by executing machine readable instructions to

run a cascading series of image smoothing and subsampling operations upon the input images to generate a collection of smoothed images,

build a first set of images from the collection of smoothed images, wherein the first set of images has a same size and structure as the smoothed images,

build a second set of images from the collection of smoothed images, wherein the second set of images has the same size and structure as the smoothed images,

utilize the first set of images to identify circular candidate object types in the input images,

utilize the second set of images to identify elongated candidate object types in the input images,

subject the identified circular and elongated candidate object types to candidate feature processing to identify valid object types in the input images, wherein the valid object types include circular object types which are ceiling lights and elongated object types which are skylights, and

generate navigational data based upon the identified valid object types.

2. The industrial vehicle of claim 1 wherein the collection of smoothed images is a Gaussian scale space pyramid.

3. The industrial vehicle of claim 2 wherein the Gaussian scale space pyramid is approximated by convolution with binomial filter kernels.

4. The industrial vehicle of claim 1 wherein the subsampling operations are implemented conditionally as a function of available navigation module computing power.

5. The industrial vehicle of claim 1 wherein

the first set of images is a determinant of Hessian response pyramid formed utilizing a determinant of Hessian;

and

the second set of images is a trace of Hessian response pyramid formed utilizing a trace of Hessian.

6. The industrial vehicle of claim 5 wherein the determinant of Hessian response and trace of Hessian response are calculated based on each scale space image being convolved with second-order partial derivative filter kernels.

7. The industrial vehicle of claim 5 wherein the determinant of Hessian response is calculated by subtraction of a mixed second-order partial derivative term.

8. The industrial vehicle of claim 5 wherein the determinant of Hessian response suppresses responses to objects characterized by different rates of image intensity change along longitudinal and transverse axial directions.

9. The industrial vehicle of claim 8 wherein the determinant of Hessian response is utilized for multiscale non-maximum suppression wherein local maxima are located within a window comprising scale and spatial dimensions.

10. The industrial vehicle of claim 9 wherein an absolute minimum threshold is applied to prevent excessive noisy false positive determinant of Hessian responses.

11. The industrial vehicle of claim 5 wherein the determinant of Hessian response is utilized in a filtering keypoints function for removing candidate points that do not likely correspond to ceiling lights.

12. The industrial vehicle of claim 11 wherein the filtering keypoints function utilizes other filters on candidate points to check against empirically set thresholds.

13. The industrial vehicle of claim 12 wherein the other filters comprise scales at which keypoints are detected, a spatial location of a keypoint, a magnitude of the determinant of Hessian response at the keypoint location, a machine-learning classifier, or an average trace of Hessian response values of a surrounding area.

14. The industrial vehicle of claim 5 wherein the determinant of Hessian response is utilized in a refining keypoints function for refining spatial position and scale of ceiling light candidate keypoints.

15. The industrial vehicle of claim 5 wherein the trace of Hessian response is utilized in a sum large-scale trace of Hessian response images function for summing a selection of trace of Hessian response images from a trace of Hessian response pyramid into an integrated trace of Hessian response image.

16. The industrial vehicle of claim 15 wherein skylight regions are searched for within the trace of Hessian response pyramid.

17. The industrial vehicle of claim 5 wherein the trace of Hessian response is further processed by a threshold integrated trace of Hessian response function, wherein the trace of Hessian response is smoothed and a threshold is applied.

18. The industrial vehicle of claim 5 wherein the trace of Hessian response is utilized in a connected component filtering function for extracting and filtering connected components based on a binary thresholded integrated trace of Hessian response image and the connected component filtering function filters for size and aspect ratio to select substantially rectangular regions.

19. A navigation module comprising memory and a processor, wherein:

the memory is coupled to the processor; and

the processor is configured to distinguish between object types in input images by executing machine readable instructions to

run a cascading series of image smoothing and subsampling operations upon the input images to generate a collection of smoothed images,

build a first set of images from the collection of smoothed images, wherein the first set of images has a same size and structure as the smoothed images,

build a second set of images from the collection of smoothed images, wherein the second set of images has the same size and structure as the smoothed images,

utilize the first set of images to identify circular candidate object types in the input images,

utilize the second set of images to identify elongated candidate object types in the input images,

subject the identified circular and elongated candidate object types to candidate feature processing to identify valid object types in the input images, wherein the valid object types include circular object types which are ceiling lights and elongated object types which are skylights, and

generate navigational data based upon the identified valid object types.