Smooth surface defect detection method based on rotating platform

Abstract

This invention discloses a method for detecting defects on smooth surfaces based on a rotating platform. The method includes a first placement step, a second placement step, a startup step, a line segment feature extraction step, a line segment change analysis and defect judgment step, and a defect result output and visualization storage step. Therefore, by using a rotating platform and a fixed long straight detection target to generate a reflective image, and analyzing the continuity of the reflective line segments through the image captured by a camera, surface defects can be accurately detected. This achieves the effect of eliminating the need for labeling and model training, and can be applied in real-time to process monitoring.

Claims

1. A method for detecting defects on a smooth surface based on a rotating platform, comprising: First placement step: Align the first center of the object to be tested with the second center of the rotating platform; A second placement step: A long, straight detection target is fixed at an outer periphery of the rotating platform, so that an edge of the long, straight detection target can reflect through a surface of the object under test to form a reflection area; and a camera is mounted at an angle slightly above the long, straight detection target to capture the entire reflection area and make the reflection area within the camera's field of view; A startup step: The rotating platform is started to rotate, and the camera is started to set the frames per second (FPS) according to actual needs to perform a 360-degree full-area scan of the surface of the object under test and continuously capture an image of the object under test; A line segment feature extraction step: Each time the camera captures a frame of the image of the object under test, a line segment detection processing procedure is performed on the image; The process of analyzing line segment changes and determining defects involves comparing a main line segment in consecutive frames. If a line segment is found to suddenly shorten in length, break into more than two segments, or have an abnormal center position, then that frame is marked as a potential defect. The process of outputting and visualizing defects involves automatically capturing a defective image and a difference mask, drawing a red frame around the broken area and a reflection line, and storing an analysis report and image data.

2. The method for detecting smooth surface defects based on a rotating platform as described in claim 1, wherein, The test object is a wafer.

3. The method for detecting smooth surface defects based on a rotating platform as described in claim 1, wherein, The long, straight object to be tested is a straight light-emitting diode (LED) or a straight ruler.

4. The method for detecting smooth surface defects based on a rotating platform as described in claim 1, wherein, The outer perimeter is above.

5. The method for detecting smooth surface defects based on a rotating platform as described in claim 1, wherein, The line segment detection processing procedure includes at least the Line Segment Detector (LSD) algorithm, Hough Transform, and Edge Drawing Lines (EDLines).

6. The method for detecting smooth surface defects based on a rotating platform as described in claim 5, wherein, When the line segment detection processing procedure is a Line Segment Detection (LSD) algorithm, the line detection step includes a reflection line segment detection sub-step and a main line segment extraction sub-step. The reflection line segment detection sub-step includes converting the input image into a grayscale image, using the Line Segment Detection (LSD) algorithm to extract the origin coordinates and end coordinates of all line segments in the image, and calculating the length of each line segment. The main line segment extraction sub-step includes retaining only the line segments that meet a main line segment selection condition from all detected line segments as the main observation line segments of the frame. The main line segment selection condition includes a center penetration condition and a maximum length condition. The center penetration condition requires that each line segment must cross a center point ±Δ region of the image. The maximum length condition requires that the length of the line segment in the set of line segments is the longest and that the length must exceed a preset minimum length threshold to exclude small noise or background texture.

7. The method for detecting smooth surface defects based on a rotating platform as described in claim 6, wherein, The ±Δ area is no more than ±50 pixels.

8. The method for detecting smooth surface defects based on a rotating platform as described in claim 6, wherein, The preset minimum length threshold is set to the length of the target object being detected, which occupies more than half of the image pixels.

9. The method for detecting smooth surface defects based on a rotating platform as described in claim 5, wherein, The line segment detection and processing procedure is the Hough Transform, which involves transforming to the parameter space, extracting the line with the highest cumulative parameter value, and then restoring its endpoints.

10. The method for detecting smooth surface defects based on a rotating platform as described in claim 5, wherein, The line segment detection processing program is called Edge Drawing Lines (EDLines), which means that line segments are fitted after concatenating edge points.

Images