Real time automated inspection

• US 4,519,041 A
• Filed: 05/03/1982
• Issued: 05/21/1985
• Est. Priority Date: 05/03/1982
• Status: Expired due to Fees

• Current Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Original Assignee: Honeywell Incorporated (Honeywell International Inc.)
• Inventors: Overland, John E., Ulrich, Franz W., Fant, Karl M., Suresh, Bindinganavle R., Levitt, Tod S., Fundakowski, Richard A.
• Primary Examiner(s): Thomas, James D.
• Assistant Examiner(s): Shaw, Dale M.
• Application Number:
  US06/374,373
• Time in Patent Office: 1,114 Days
• Field of Search: 364/468, 364/469, 364/474, 364/476, 364/550, 364/552, 364/507, 364/472, 382/8, 382/16, 382/21, 382/22, 382/54, 356/429-431
• US Class Current: 382/141
• CPC Class Codes: B21B 38/00 Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product control devices or methods B21B37/00
• CPC Class Codes: G01N 2021/8918 Metal
• CPC Class Codes: G01N 21/89 in moving material, e.g. running paper or textiles G01N21/90, G01N21/91, G01N21/94 take precedence
• CPC Class Codes: G01N 21/8903 using a multiple detector array
• CPC Class Codes: G06T 2207/30136 Metal
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume

Abstract

A method and apparatus relating to the real time automatic detection and classification of characteristic type surface imperfections occurring on the surfaces of material of interest such as moving hot metal slabs produced by a continuous steel caster. A data camera transversely scans continuous lines of such a surface to sense light intensities of scanned pixels and generates corresponding voltage values. The voltage values are converted to corresponding digital values to form a digital image of the surface which is subsequently processed to form an edge-enhanced image having scan lines characterized by intervals corresponding to the edges of the image. The edge-enhanced image is thresholded to segment out the edges and objects formed by the edges are segmented out by interval matching and bin tracking. Features of the objects are derived and such features are utilized to classify the objects into characteristic type surface imperfections.

Claims

1. A method for the real time automatic detection and classification of characteristic type surface imperfections occurring on the surfaces of material of interest comprising the steps of;

transversely scanning a relatively moving surface of said material of interest with data generating means to sense the light intensities of scanned pixels and generating signals indicative of the sensed intensities of said scanned pixels;

converting said signals to corresponding digital values to form a digital image of said surface and storing said digital values;

edge enhancing said digital image formed by said digital values with an edge enhancement operator on the fly to form an edge enhanced image containing intervals of said image corresponding to the edges of said surface imperfections;

thresholding said edge enhanced image to segment out said edges to form a thresholded image;

interval matching said thresholded image to establish connectivity of intervals on adjacent scan lines and bin tracking connected ones of said interval (threshold image) to track and segment out imperfection objects formed by said edges;

forming object feature vectors by computing and accruing directly determinable features including an object identifier of each of said objects; and

classifying said objects with the aid of said features into characteristic type surface imperfections.

2. The method according to claim 1 wherein said data generating means is a data camera and said generated signals are voltage signals.

3. The method according to either of claims 1 or 2 wherein said scanning, edge enhancement and interval matching and bin tracking are accomplished on a scan line basis.

4. The method according to claim 3 wherein said scan lines are contiguous.

6. The method according to either of claims 1 or 5 wherein said step of forming object features further comprises:

forming primary features including a unique identification, pixel area and boundaries thereof;

computing secondary features from said primary features including length, width, and slant of said objects.

7. A method according to either of claims 1 or 5 including the step of outputting data in the form of a report identifying classes of discovered imperfections and information with regard to sizes and locations of said imperfections, wherein said information is in human readable form.

8. A method according to either of claims 1 or 5 wherein said light intensities of said pixels result from reflected light from said surface.

9. A method according to either of claims 1 or 5 including the step of companding said digital image to minimize background clutter.

10. A method according to claim 9 wherein the step of companding includes logarithmically companding said digital image to minimize background clutter by expanding the dynamic range in the lower intensity region harboring imperfections to exaggerate the imperfection edges and compressing the dynamic range harboring background clutter to suppress the clutter edges.

11. A method according to either of claims 1 or 5 wherein pixels equal to or larger than the threshold value are set equal to 1 and to 0 if less than said threshold.

12. A method according to either of claims 1 or 5 wherein variable adaptive thresholding is utilized.

13. A method according to either of claims 1 or 5 wherein said classifying of objects is implemented by a hierarchal tree classifier.

14. A method according to claim 13 wherein said hierarchal tree classifier has binary statistical classifiers.

15. A method according to either of claims 1 or 5 wherein said material of interest is metal in slab form and further comprising the step of automatically making slab disposition decisions, said step comprising;

establishing slab disposition determining dimensions of the physical attributes of different kinds of said imperfections and storing said dimensions; and

comparing said computed features with said dimensions to make an automatic slab disposition decision based on the results thereof.

16. A method according to either of claims 1 or 5 including the step of comparing classified objects on the basis of proximity searches to make a determination of the presence of sister components of larger objects and combining revealed related sister components to form larger objects upon finding sets of related sister components;

and

appropriately classifying newly revealed ones of said larger objects.

17. A method according to claim 15 including the step of comparing classified objects on the basis of proximity searches to make a determination of the presence of sister components of larger objects and combining revealed related sister components to form larger objects upon finding sets of related sister components;

and

appropriately classifying newly revealed ones of said larger objects.

18. A method according to claim 16 wherein syntactic/semantic rules are used in connection with said determination.

19. A method according to claim 17 wherein syntactic/semantic rules are used in connection with said determination.

20. A method according to claim 18 wherein strings of potential objects are formed and spatial tests are made for said strings as prerequisites for combining objects.

21. A method according to either of claims 1 or 5 wherein one of the said feature extraction techniques involves an approximating technique wherein recorded interval data is discarded on a line skipping basis in proportion to the length of an object being tracked and during the time the object is tracked.

22. A method according to either of claims 1 or 5 wherein said interval matching and bin tracking involves an approximating technique wherein for each interval recorded only the data corresponding to the midpoint of each said interval is retained to represent an edge of a tracked object.

5. A method for the real time automatic detection and classification of characteristic type surface imperfections occurring on the surfaces of material of interest, comprising the steps of;

transversely scanning lines of a relatively moving surface of said material with a data camera to sense the light intensities of scanned pixels and generating voltage values which correspond to the sensed intensities of said scanned pixels;

converting said voltage values to corresponding digital values to form a digital image of said surface and storing said digital values;

edge enhancing said digital image formed by said digital values with an edge enhancement operator on the fly on a scan line basis to form an edge enhanced image comprising scan lines containing intervals of said image corresponding to the edges of said imperfections;

thresholding said edge enhanced image to segment out said edges and form a thresholded image;

interval matching said threshold image to establish connectivity of intervals on adjacent scan lines and bin tracking connected one of said intervals (threshold image) on a scan line basis to track and segment out objects formed by said edges;

forming object feature vectors by computing and accruing directly determinable features including an object identifier of each of said objects including a unique identification, pixel area and boundaries thereof; and

classifying said objects with the aid of said features into characteristic type surface imperfections.

23. An apparatus for the real time automatic detection and classification of characteristic type surface imperfections occurring on the surfaces of material of interest comprising;

means for transversely scanning contiguous lines of a relatively moving surface of said material of interest with a data generating means to sense the light intensities of scanned pixels and generating signals indicative of the sensed intensities of said scanned pixels;

means for converting said signals to corresponding digital values to form a digital image of said surface and storing said digital values;

means for edge enhancing said digital image formed by said digital values with an edge enhancement operator on the fly to form an edge enhanced image containing intervals of said image corresponding to the edges of said surface imperfections;

means for thresholding said edge enhanced image to segment out said edges to form a thresholded image;

means for interval matching said thresholded image to establish connectivity of intervals on adjacent scan lines and means for bin tracking connected ones of said intervals (thresholded image) to track and segment out imperfection objects formed by said edges;

means for forming object feature vectors by computing and accruing directly determinable features including an object identifier of each of said objects;

means for classifying said objects with the aid of said features into characteristic type surface imperfections.

24. An apparatus for the real time automatic detection and classification of characteristic type surface imperfections occuring on the surfaces of material of interest, comprising of the following elements;

means for transversely scanning lines of a relatively moving surface of said material with a data camera to sense the light intensities of scanned pixels and generating voltage values which correspond to the sensed intensities of said scanned pixels;

means for converting said voltage values to corresponding digital values to form a digital image of said surface and storing said digital values;

means for edge enhancing said digital image formed by said digital values with an edge enhancement operator on the fly on a scan line basis to form an edge enhanced image comprising scan lines containing intervals of said image corresponding to the edges of said imperfections;

means for thresholding said edge enhanced image to segment out said edges to form a thresholded image;

means for interval matching said thresholded image to establish connectivity of intervals on adjacent scan lines and bin tracking connected ones of said intervals (thresholded image) on a scan line basis to track and segment out objects formed by said edges;

means for forming object feature vectors by computing and accruing directly determinable features of each of said objects including a unique identification, pixel area and boundaries thereof; and

means for classifying said objects with the aid of said features into characteristic type surface imperfections.