System and method of surface inspection of an object using mulitplexed optical coherence tomography

1. A method of detecting defects in a surface of an object using multiplexed optical coherence tomography (OCT) and machine learning, the method comprising:

imaging the object using a multiplexed OCT imaging technique, the imaging including:

moving the object relative to two or more scanner heads along a direction of travel;

alternatingly directing a sample beam to each of the two or more scanner heads;

when the sample beam is directed at a first of the scanner heads, scanning comprising:

steering the sample beam from the first of the scanner heads to an unscanned region on the surface of the object; and

performing an A-scan of the object; and

when the sample beam is directed at a subsequent one of the scanner heads, scanning comprising:

steering the sample beam from the subsequent one of the scanner heads to an adjacent unscanned region adjacent to the most-recently scanned region along a direction opposite the direction of travel; and

performing an A-scan of the object;

forming an OCT image from the A-scans; and

analyzing the OCT image to determine whether a defect is present in the surface of the object, the analyzing including:

providing the OCT image to an input layer of a neural network trained to detect and classify defects in the OCT image, the neural network trained using imaging data for which defect location and defect existence is known;

determining whether a defect is present in the OCT image using the neural network, the determining including generating a defect classification for each detected defect at an output layer of the neural network; and

for each detected defect, determining via a processor a feature localization for the detected defect.

2. The method of claim 1 , wherein the sample beam is steered at a first unscanned region or an unscanned region adjacent to a most-recently scanned region along a direction opposite the direction of travel.

3. The method of claim 1 , wherein, when the sample beam is directed at each respective scanner head, the scanning further comprising:

determining whether the surface of the unscanned region is in focus; and

where the surface of the unscanned region is not in focus, adjusting an optical path in the OCT such that the surface of the unscanned region is in focus.

4. The method of claim 3 , wherein determining whether the surface of the unscanned region is in focus comprises measuring a distance between the respective scanner head and the surface of the object at the unscanned region to determine whether the surface is within a focal length of the OCT with the respective scanner head.

5. The method of claim 3 , further comprising retrieving a surface geometry of the object, wherein determining whether the surface of the unscanned region is in focus comprises determining whether the surface is within a focal length of the OCT with the respective scanner head using the surface geometry of the object to determine a distance between the respective scanner head and the surface.

6. The method of claim 1 , wherein, when the sample beam is directed at each respective scanner head, the scanning further comprising:

determining whether a difference between a reference path of the OCT and a sample path of the OCT at the unscanned region are within a coherence length; and where the difference is not within the coherence length, adjusting an optical path in the OCT such that the difference is within the coherence length.

7. The method of claim 1 , wherein successively scanned regions are nonadjacent.