Method of grading sample

• US 7,142,693 B2
• Filed: 07/10/2002
• Issued: 11/28/2006
• Est. Priority Date: 07/11/2001
• Status: Expired due to Fees

• Current Assignee: Shofner Engineering Associates Incorporated
• Original Assignee: Shofner Engineering Associates Incorporated
• Inventors: Zhang, Yupeng, Shofner, Frederick M., Shofner, Christopher K.
• Primary Examiner(s): Ahmed, Samir
• Assistant Examiner(s): MISTRY, O NEAL RAJAN
• Application Number:
  US10/192,143
• Publication Number:
  US 20030059090A1
• Time in Patent Office: 1,602 Days
• Field of Search: 382/110, 382/111, 356/238.1
• US Class Current: 382/100
• CPC Class Codes: G01N 21/8851 Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges G01N21/8806 and G01N21/93 - G01N21/95692 take precedence; optical measurement of dimensions G01B11/00; optical scanning G02B26/10; image transformation G06T3/00; computerised image enhancement G06T5/00; image processing per se for flaw detection G06T7/0002
• CPC Class Codes: G01N 33/362 material before processing, e.g. bulk cotton or wool
• CPC Class Codes: G06F 18/2321 using statistics or function optimisation, e.g. modelling of probability density functions
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 2207/30124 Fabrics; Textile; Paper
• CPC Class Codes: G06T 7/0006 using a design-rule based approach
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 10/267 by performing operations on regions, e.g. growing, shrinking or watersheds
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 10/56 relating to colour

Abstract

Method of grading a sample. An input color image of the sample, such as a sample of raw cotton, is obtained by a scanning technique. This image is first analyzed by an image segmentation module and a binary image where pixels of a first characteristic are marked as 0 (e.g. cotton lint) and non-lint pixels are marked as 1. In a following particle recognition module, the adjacent non-lint pixels are grouped together as particles, and the color, size, shape and edge strength for each particle are computed. The particle descriptions are analyzed and each particle is assigned a trash type, i.e. leaf, bark, grass, pepper trash or shadow. The data of leaf and pepper trash are analyzed by a leaf grading module and leaf grade is reported. The data of bark/grass are analyzed by a bark/grass grading module and the bark/grass grade is reported.

Claims

1. A method of analysis of a sample of cotton which includes diverse elements comprising the steps of

a. electro-optically capturing and electronically processing a digital image of said sample, said digital image including representations of each of a plurality of the diverse elements of this sample,

i. scanning the sample of cotton to develop a digital image of said sample, said digital image including representations of each of a plurality of the diverse elements of the sample,

ii. identifying substantially each pixel of said digital image as lint pixel or non-lint pixel,

iii. grouping adjacent non-lint pixels into individual ones of said diverse elements,

iv. calculating at least one or more of the color, shape, texture, length/width ratio, and edge strength feature of each of said identified diverse elements,

v. classifying each of said diverse elements as one of leaf, pepper trash, bark, grass or shadow based on the calculated features of each of said diverse elements,

vi. calculating percent area and count per inch² for each class of diverse elements,

b. providing a computer equipped with an artificial neural network,

c. training said artificial neural network employing multiple samples of cotton which have been visually examined and assigned a grade by a human, said training including storing within said computer equipped with said artificial neural network, a digital image of each of said multiple samples and their respective grades which have been assigned thereto by the human,

d. employing said computer and said trained artificial neural network, comparing said existence of, and one or more properties of the sample of cotton to the existence of, and one or more properties of, each of said multiple samples which have been visually examined and assigned a grade by a human and stored in said computer,

e. outputting from said computer a numerical grade for said sample of cotton which is a function of the degree of similarity of said sample of cotton to a human graded one of said multiple samples stored in said computer.

2. The method of claim 1 wherein the sample comprises cotton fibers and said diverse elements comprise trash disposed within the sample.

3. The method of claim 1 wherein said diverse elements include at least one of leaf, bark, grass, and/or pepper trash.

4. The method of claim 3 wherein one of said diverse elements includes shadows.

5. The method of claim 1 wherein said steps of identifying substantially each pixel of said digital image as lint pixel or non-lint pixel and grouping adjacent non-lint pixels into individual particles, includes evaluation of said pixels employing a Bayesian Weighted K-Means Clustering algorithm.

6. A method of analysis of a sample of raw cotton which includes diverse elements comprising the steps of

capturing a two-dimensional color digital image of the sample,

converting said digital image to a binary image wherein each of the pixels of said digital image is assigned a value which identifies the pixel as lint or non-lint,

grouping at least said non-lint pixels into individual classes of trash having like properties comprising,

a. calculating at least one or more of the color, shape, texture, length/width ratio, and edge strength feature of each of said diverse elements,

b. classifying each of said diverse elements as one of leaf, pepper trash, bark, grass or shadow based on the calculated features of each of said diverse elements,

c. calculating percent area and count per inch² for each class of diverse elements.

providing a computer equipped with an artificial neural network,

training said artificial neural network employing multiple samples of cotton which have been visually examined and assigned a grade by a human, said training including storing within said computer equipped with said artificial neural network, a digital image of each of said multiple samples and their respective grades which have been assigned thereto by the human,

employing said computer and said trained artificial neural network, comparing said existence of, and one or more properties of the sample of cotton to the existence of, and one or more properties of, each of said multiple samples which have been visually examined and assigned a grade by a human and stored in said computer,

outputting from said computer a numerical grade for said sample of cotton which is a function of the degree of similarity of said sample of cotton to a human graded one of said multiple samples stored in said computer.

7. The method of claim 6 wherein said samples and associated grades stored in said artificial neural network includes one or more computer-graded samples of cotton and their associated grades.

8. The method of claim 6 wherein said steps of converting said digital image to a binary image wherein each of the pixels of said digital image is assigned a value which identifies the pixel as lint or non-lint, grouping at least said non-lint pixels into individual classes of trash having like properties includes evaluation of said pixels employing a Bayesian Weighted K-Means Clustering algorithm.

9. A method for the grading of a cotton sample which includes diverse trash elements comprising the steps of

a. developing a two-dimensional color digital image of the sample,

b. creating a CIELAB color space,

c. employing a Bayesian Weighted K-Means Clustering Algorithm, classifying the data points in the CIELAB color space to convert said digital image to a binary image,

d. assigning values to each of the pixels of said binary image,

e. grouping like ones of said sets into two or more groups,

f. categorizing and analyzing at least one selected group of said groups employing a connected components labeling algorithm to divide said binary image into sets of coherent regions or objects,

g. creating a RGB color space employing said two or more groups of values,

h. analyzing at least one of the color size, shape and edge strength of each categorized selected group,

i. marking each categorized group as a function of its detected characteristics,

j. providing a computer equipped with an artificial neural network,

k. training said artificial neural network employing multiple samples of cotton which have been visually examined and assigned a grade by a human, said training including storing within said computer equipped with said artificial neural network, a digital image of each of said multiple samples and their respective grades which have been assigned thereto by the human,

l. employing said computer and said trained artificial neural network, comparing said existence of, and one or more properties of the sample of cotton to the existence of, and one or more properties of, each of said multiple samples which have been visually examined and assigned a grade by a human and stored in said computer,

m. outputting from said computer a numerical grade for said sample of cotton which is a function of the degree of similarity of said sample of cotton to a human graded one of said multiple samples stored in said computer.

10. An image-based method of grading cotton samples, which samples include cotton lint and diverse, non-lint elements, said diverse non-lint elements having localized, differently-responding image features from cotton lint, and comprising the steps of

a. capturing a digital image of the cotton sample, said digital image including a multiplicity of pixels, each of said pixels having unique and localized spatial position and an amplitude response in proportion to light intensity received thereupon, said multiplicity of pixels comprising an original image file stored in and processed by a digital computer,

b. employing said digital computer, determining, from the amplitude response for each of said pixels in said digital image of the cotton sample, whether each pixel represents cotton lint or a non-lint element, and assigning binary values of “

0”

to lint and “

1”

to non-lint elements disposed within the cotton sample,

c. grouping one or more adjacent ones of said non-lint pixels into a binary image of said non-lint elements, and storing said binary image within said computer,

d. employing said binary image and said original image file in said computer, determining one or more image features for each of said non-lint elements,

e. employing said one or more determined image features, grouping said non-lint elements into one or more classes of cotton trash,

f. employing said stored original binary image and said determined image files, calculating count per inch² and percentage area for said one or more determined image features for each of said non-lint elements,

g. providing a computer having an artificial neural network associated therewith,

h. introducing to said computer a multiplicity of cotton samples having respective ones of a wide range of human-assigned grades for each class of said non-lint elements, executing steps a-f hereinabove, to thereby train said artificial neural network by measurement of image feature data for each such non-lint classification and associating said measured image feature data with said human-assigned grades, and

i. thereafter, for an unknown cotton sample, using said measurement data of image features and said trained artificial neural network, outputting instrumental estimates of the grade of said unknown cotton sample corresponding to said diverse, non-lint elements in said unknown sample.

11. The method of claim 10 wherein said steps a-e include analysis of data accumulated in said computer employing a Bayesian K-Means Clustering Algorithm.

12. The method of claim 10 wherein said diverse non-lint elements comprise leaf, pepper trash, bark, grass and shadows.

13. A method of analysis and outputting of a grade for a sample of cotton which includes diverse elements comprising the steps of

a. providing a computer equipped with an artificial neural network,

b. training said artificial neural network employing multiple samples of cotton, each of which has been visually examined and assigned at least one grade by at least one human, said training comprising the steps of:

i. electro-optically capturing and electronically processing a digital image of each of said multiple samples, each of said digital images including representations of each of a plurality of the respective diverse elements of said multiple samples,

ii for each of said multiple samples, identifying substantially each pixel of each of said digital images as a lint pixel or a non-lint pixel,

iii. grouping adjacent non-lint pixels of respective ones of said multiple samples into individual ones of said diverse elements,

iv. calculating at least one of the color, area, shape, texture, length/width ratio, and edge strength feature of each of said identified diverse elements of each of said multiple samples,

v. classifying each of said calculated diverse elements for each of said multiple samples as one of leaf, pepper trash, bark, grass or shadow based on the calculated features of each of said elements for each of said multiple samples,

vi. calculating percent area and count per inch² for each class of diverse elements for each of said multiple samples,

vii. storing within said computer equipped with said artificial neural network the percent area and counts/in² associated with each classification of said diverse elements for each of said training samples and the respective grades of each of said multiple samples which have been assigned thereto by the at least one human for each of said diverse elements,

viii. combining said human grades and said image data to produce a trained response, and

ix. storing said trained response in said computer for subsequent analysis and outputting of an instrument grade for unknown samples;

c. analyzing an unknown sample of cotton, comprising repeating steps b-i through b-vi hereinabove;

d. employing said trained artificial neural network in said computer, comparing said percent area and count/in² determined from said digital image of said unknown sample for each class of diverse elements to the stored percent area and count/in² of said multiple samples and outputting a grade for each of said classes for said unknown sample.

14. A method for the analysis and assignment of a grade to a sample of cotton comprising the steps of

a. providing a computer equipped with an artificial network,

b. training said computer using multiple samples of cotton with known Human Classer grades,

c. using said trained computer, analyzing a sample of cotton having an unknown grade, and,

d. outputting at least one of leaf, bark, grass, and pepper trash grade for said sample of unknown grade.