System and method for object recognition and classification using a three-dimensional system with adaptive feature detectors

• US 8,452,078 B2
• Filed: 10/15/2008
• Issued: 05/28/2013
• Est. Priority Date: 10/15/2008
• Status: Active Grant

• Current Assignee: Toyota Motor Corporation
• Original Assignee: Toyota Motor Engineering & Manufacturing North America Incorporated (Toyota Motor Corporation)
• Inventors: Prokhorov, Danil V.
• Primary Examiner(s): Lu, Tom Y
• Assistant Examiner(s): CONWAY, THOMAS A
• Application Number:
  US12/252,080
• Publication Number:
  US 20100092073A1
• Time in Patent Office: 1,686 Days
• Field of Search: None
• US Class Current: 382/154
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
• CPC Class Codes: G06V 20/653 by matching three-dimensional models, e.g. conformal mapping of Riemann surfaces

Abstract

A method including imaging an object in three-dimensions; binning data of the imaged object into three-dimensional regions having a predetermined size; determining a density value p of the data in each bin; inputting the p density values of the bins into a first layer of a computational system including a corresponding processing element for each of the bins; calculating an output O of the processing elements of the computational system while restricting the processing elements to have weights Wc1 connecting the processing elements to the corresponding p density values; and communicating an estimated class of the scanned object based on the calculated system outputs.

Claims

1. A method, comprising:

imaging an object in three-dimensions by collecting three-dimensional data points of the object;

binning the data points of the imaged object into three-dimensional bins having a predetermined three-dimensional size to create a binned scanned object in a three-dimensional bin space;

determining a density value of the data in each of the bins by calculating a number of the data points in each of the bins;

inputting the density value of each of the bins into a first layer of a computational system including a corresponding processing element for each of the bins;

calculating outputs of the processing elements, wherein all of the processing elements have weights of a same value connecting the processing elements to corresponding density values; and

communicating an estimated class of the scanned object based on the calculated outputs, wherein the calculating includes;

performing a nonlinear transformation on a product of a connected weight and the density value for each bin;

sub-sampling the outputs by averaging the outputs over two adjacent processing elements in three dimensions; and

repeating the calculating process using additional layers of the computational system, the output of the computational system connected to all processing elements of a last layer of the computational system.

2. The method of claim 1, wherein the data of the imaged object in three-dimensions is represented in a Cartesian coordinate system.

3. The method of claim 1, wherein the data of the imaged object in three-dimensions is represented in a spherical system.

4. The method of claim 1, wherein the nonlinear transformation includes applying a hyperbolic tangent to the product of the connected weight and the density value for each bin.

5. The method of claim 1, wherein a number of the weights depends on a connectivity pattern between the input density values and the processing elements in the first layer of the computational system.

6. The method of claim 1, wherein the estimated class of the scanned object is selected by the computational system from a group including at least two candidate classes.

7. The method of claim 1, wherein the imaging is accomplished with a three-dimensional scanning system.

8. The method of claim 1, wherein the imaging is accomplished with a three-dimensional non-scanning system.

9. A non-transitory computer readable medium having computer code stored thereon, the computer code, when executed by a computer, causing the computer to implement the method according to claim 1.

10. The method of claim 1, further comprising:

rotating the binned scanned object, creating N additional views of the binned scanned object, each view including the bins of the data points; and

determining weights for each of the N+1 views, wherein

the computational system includes three-dimensional maps,

each map includes processing elements for one of the N+1 views, and

each of the processing elements for one of the maps is connected to the corresponding density values of the corresponding view via the determined weights of the corresponding map.

11. The method of claim 10, further comprising:

normalizing the density values separately for each of the N+1 views.

12. A method, comprising:

imaging an object in three-dimensions by collecting three-dimensional data points of the object;

binning the data points of the imaged object into three-dimensional bins having a predetermined three-dimensional size to create a binned scanned object in a three-dimensional bin space;

determining a density value of the data in each of the bins by calculating a number of the data points in each of the bins;

inputting the density value of each of the bins into a first layer of a computational system including a corresponding processing element for each of the bins; and

calculating outputs of the processing elements, wherein all of the processing elements have weights of a same value connecting the processing elements to corresponding density values, wherein

the weights are obtained through unsupervised training of the computational system, and the unsupervised training includes;

initializing a plurality of weights Wc1(num) for num =1, 2,,... Nu, wherein Nu is a predetermined integer;

for each num, computing an output O for a randomly selected bin of data;

determining a num_max corresponding to a maximum computed output O;

adjusting the weight Wc1(num_max) to implement an unsupervised training algorithm; and

repeating the computing, determining, and adjusting steps until changes in the weights Wc1 are smaller than a predetermined value.

13. The method of claim 12, wherein the adjusting includes implementing a form of the Hebbian adaptation rule.

14. The method of claim 12, wherein the computing the output O includes:

computing a scalar product for each density value of bins adjacent the randomly selected bin of data, referred to as p(ind), by calculating Wc1(num)*p(ind); and

computing Euclidian norms of Wc1(num) and p(ind),

wherein the computed output O of the randomly selected bin is a function of the computed scalar products and the computed Euclidian norms.

15. The method of claim 12, wherein the calculating includes:

performing a nonlinear transformation on a product of a connected weight and the density value for each bin;

sub-sampling the outputs by averaging the outputs over two adjacent processing elements in three dimensions; and

repeating the calculating process using additional layers of the computational system, the output of the computational system connected to all processing elements of a last layer of the computational system.

16. A non-transitory computer readable medium having computer code stored thereon, the computer code, when executed by a computer, causing the computer to implement the method according to claim 12.

17. A method, comprising:

imaging an object in three-dimensions by collecting three-dimensional data points of the object;

binning the data points of the imaged object into three-dimensional bins having a predetermined three-dimensional size to create a binned scanned object in a three-dimensional bin space;

determining a density value of the data in each of the bins by calculating a number of the data points in each of the bins;

inputting the density value of each of the bins into a first layer of a computational system including a corresponding processing element for each of the bins; and

calculating outputs of the processing elements, wherein all of the processing elements have weights of a same value connecting the processing elements to corresponding density values, wherein

the weights are obtained through unsupervised training of the computational system, and the supervised training includes;

obtaining a representative set of examples of objects with the classes;

training the system using a supervised training algorithm until a predetermined error value is reached;

testing the trained system on a set of unknown objects to obtain error value;

determining whether the error value is within tolerance of a training error; and

repeating the supervised training process until the error value is within the tolerance of the training error.

18. The method of claim 17, wherein the supervised training algorithm is a form of an incremental optimization algorithm.

19. The method of claim 17, wherein the calculating includes:

performing a nonlinear transformation on a product of a connected weight and the density value for each bin;

sub-sampling the outputs by averaging the outputs over two adjacent processing elements in three dimensions; and

repeating the calculating process using additional layers of the computational system, the output of the computational system connected to all processing elements of a last layer of the computational system.

20. A non-transitory computer readable medium having computer code stored thereon, the computer code, when executed by a computer, causing the computer to implement the method according to claim 17.