[54] METHOD AND APPARATUS FOR
DETECTING SURFACE FEATURES OF
TRANSLUCENT OBJECTS
[75] Inventors: Richard D. Heck, Riverside; Henry A.
Affeldt, Victorville, both of Calif.
[73] Assignee: Sunkist Growers, Inc., Ontario, Calif.
[21] Appl. No.: 334,165
[22] Filed: Nov. 3, 1994
[51] Int. CI. G06K 9/00
[52] U.S. CI 382/110; 382/142; 209/588;
348/127; 348/128; 348/131; 348/164
[58] Field of Search 382/110, 141,
382/142; 209/588, 511; 348/127, 128, 131, 164, 89, 91; 356/239, 240, 237, 52, 337,
338
[56] References Cited
U.S. PATENT DOCUMENTS
1,943,278 1/1934 Thompson et al 209/588
3,549,890 12/1970 Keller 382/142
3,768,645 10/1973 Conway et al 209/111.5
3,930,994 1/1976 Conway et al 209/74 M
3,932,042 1/1976 Faani et al 356/240
4,055,834 10/1977 Planke 382/142
4,376,951 3/1983 Miyazawa 356/240
4,608,709 8/1986 Hedler et al 382/142
4,741,042 4/1988 Throop et al 382/110
4,760,270 7/1988 Miller 250/563
5,007,096 4/1991 Yoshida 382/142
5,026,982 6/1991 Stroman 250/223
5,164,795 11/1992 Conway 348/89
FOREIGN PATENT DOCUMENTS
0058028 8/1982 European Pat. Off. .
0566397 10/1993 European Pat. Off. .
Primary Examiner—Christopher S Kelley
Attorney, Agent, or Firm—Linda M. Robb; Daniel L. Dawes
[57] ABSTRACT
The topographic surface features of a translucent object, such as a citrus fruit with a peel, are scanned and evaluated to permit the classification thereof according to its surface features. In the case of citrus fruit, the coarseness or pebbliness, puff and crease, ridge and valley, cuts, punctures, scrapes and splits, clear rot or sour rot of the peel is optically identified through digital analysis of the pixel scans and sorted based upon the peel surface quality. The object is classified by separating the scanned image of the fruit from the background image and removing the background image. A statistical evaluation of the image of the object as a whole, including both hemispheres of the object, is made to determine if there is any surface feature variation which might qualify as a defect or be a suitable basis upon which a classification judgment can be made. If there is, the object image is subject to high frequency or low pass filtering and thresholding of the pixel intensities to derive a refined image. The refined image is then tabulated or organized into neighborhoods contiguous to sharp transitions or gradual transitions to identify specific areas defined as blobs which, when compared against a given minimum area, shape requirement and/or width can be identified as one of the surface imperfections sought.
19 Claims, 20 Drawing Sheets
