Segment-based pattern matching algorithm

1. A computer-implemented method for matching a pattern, said the method comprising:

generating a set of model descriptors representative of segment features of a first object captured in a first image;

generating a set of query descriptors representative of segment features of a second object captured in a second image;

selecting a first model segment to align with the query segments to determine if any of the query segments correspond with the first model segment based on respective model descriptors and query descriptors;

generating a hypothesis by computing a transformation to match the selected first model descriptor with a query descriptor;

validating the hypothesis by performing a model fitting algorithm when comparing other transformed model descriptors with other query descriptors with each query descriptor relative to a segment with a comparable length ratio between a minimum length ratio threshold and a maximum length ratio threshold; and

based on a consensus value from the model fitting algorithm, determining whether a pattern match exists between the first and second images.

2. The method according to claim 1 , further comprising:

defining a termination criteria for the consensus value; and

if the consensus value reaches or crosses the termination criteria, stop validating the hypothesis;

otherwise, continue validating the hypothesis.

3. The method according to claim 1 , further comprising:

sorting the model descriptors by length of the corresponding segments;

retaining a subset of model descriptors; and

wherein selecting a first model segment includes selecting a longest model segment based on corresponding first model descriptors of the retained subset of model descriptors.

4. The method according to claim 1 , wherein validating the hypothesis includes:

searching for a closest query segment based on query descriptors;

comparing the closest query segment with a closest transformed model segment based on transformed model descriptors;

determining whether the closest query and transformed model descriptors match, and if so, increase the consensus value, otherwise, maintain the consensus value.

5. The method according to claim 1 , wherein generating a hypothesis includes generating a scaled rotation matrix including midpoint translation, rotation, and scaling of the first model descriptors to match each set of query descriptors to test correspondence thereof.

6. The method according to claim 5 , further comprising generating hypotheses for each other model descriptor by generating transforms for each model descriptor for matching to each query segment.

7. The method according to claim 1 , further comprising setting a minimum consensus threshold that the consensus value is to equal or cross in determining that a pattern match exists between the first and second images.

8. The method according to claim 1 , wherein the model fitting algorithm is a random sample consensus (RANSAC) algorithm.

9. A computer-implemented method for matching a pattern, the method comprising:

generating a set of model descriptors representative of segment features of a first object captured in a first image including defining a set of model descriptors by determining a location of a midpoint, an orientation angle relative to the x-axis, and a link between endpoints for each model segment;

generating a set of query descriptors representative of segment features of a second object captured in a second image;

selecting a first model segment to align with the query segments to determine if any of the query segments correspond with the first model segment based on respective model descriptors and query descriptors;

generating a hypothesis by computing a transformation to match the selected first model descriptor with a query descriptor;

validating the hypothesis by performing a model fitting algorithm when comparing other transformed model descriptors with other query descriptors; and

based on a consensus value from the model fitting algorithm, determining whether a pattern match exists between the first and second images.

10. The method according to claim 9 , further comprising determining a unique orientation of each model descriptor by performing a cross-product of a segment vector between the midpoint and a first endpoint and a gradient vector.

11. An imaging system, comprising:

an imaging camera;

a processing unit in communication with said imaging camera, and configured to:

generate a set of model descriptors representative of segment features of a first object captured in a first image and define a set of model descriptors by determining a location of a midpoint, an orientation angle relative to the x-axis, and a link between endpoints for each model segment;

generate a set of query descriptors from a second object captured in a second image;

select a first model segment to align with the query segments to determine if any of the query segments correspond with the first model segment based on respective model descriptors and query descriptors;

generate a hypothesis by computing a transformation to match the selected first model descriptor with a query descriptor;

validate the hypothesis by performing a model fitting algorithm when comparing other transformed model descriptors with other query descriptors; and

based on a consensus value from the model fitting algorithm, determine whether a pattern match exists between the first and second images.

12. The system according to claim 11 , wherein the processing unit is further configured to:

define a termination criteria for the consensus value; and

if the consensus value reaches or crosses the termination criteria, stop validating the hypothesis;

otherwise, continue validating the hypothesis.

13. The system according to claim 11 , wherein the processing unit, is further configured to:

sort the model descriptors by length of the corresponding segments;

retain a subset of model descriptors; and

in selecting a first model segment, the processing unit is further configured to select a longest model segment based on corresponding first model descriptors.

14. The system according to claim 11 , wherein the processing unit, in validating the hypothesis, is configured to validate the hypothesis with each query descriptor relative to a segment with a comparable length ratio between a minimum length ratio threshold and a maximum length ratio threshold.

15. The system according to claim 11 , wherein the processing unit, in validating the hypothesis, is configured to:

search for a closest query segment based on query descriptors;

compare the closest query segment with a closest transformed model segment based on transformed model descriptors; and

determine whether the closest query and transformed model descriptors match, and if so, increase the consensus value, otherwise, maintain the consensus value.

16. The system according to claim 11 , wherein the processing unit is further configured to determine a unique orientation of each model descriptor by performing a cross-product of a segment vector between the midpoint and a first endpoint and a gradient vector.

17. The system according to claim 11 , wherein the processing unit, in generating a hypothesis, is configured to generate a scaled rotation matrix including midpoint translation, rotation, and scaling of the first model descriptors to match each set of query descriptors to test correspondence thereof.

18. The system according to claim 17 , wherein the processing unit is further configured to generate hypotheses for each other model descriptor by generating transforms for each model descriptor for matching to each query segment.

19. The system according to claim 11 , wherein the processing unit is further configured to set a minimum consensus threshold that the consensus value is to equal or cross in determining that a pattern match exists between the first and second images.

20. The system according to claim 11 , further comprising a laser marking system configured to perform direct part marking (DPM).