METHOD AND SYSTEM FOR ASSESSING QUALITY OF COMMODITIES

• US 20200134682A1
• Filed: 04/06/2018
• Published: 04/30/2020
• Est. Priority Date: 04/06/2017
• Status: Active Grant

• Current Assignee: Nebulaa Innovations Private Limited
• Original Assignee: Nebulaa Innovations Private Limited
• Inventors: SETHI, Tanmay, DADHICH, Mohit, GUPTA, Yogesh Kumar, RATHORE, Tapish
• Granted Patent: US 11,397,974 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/51 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06F 18/2413 based on distances to training or reference patterns
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 30/0283 Price estimation or determination
• CPC Class Codes: G06Q 30/0601 Electronic shopping [e-shopping]
• CPC Class Codes: G06Q 30/0625 Directed, with specific intent or strategy
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/30128 Food products
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/136 involving thresholding
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation

Abstract

The present disclosure relates to method and system for assessing quality of commodities. An image of bulk commodity is captured and segmented into one or more segmented images for classification into one of predefined categories of commodities. The method and system classify the commodities based on generalized features created from training images. One or more features in the training images are determined and grouped to obtain the generalized features. A feature score and corresponding weightage score of the generalized feature is then determined to estimate a predetermined regression score. Based on the generalized features and predetermined regression score, a likelihood score of the segmented image is determined that enables the classification of the input image to one of the predefined categories of commodities. Thus, the present disclosure enables quality assessment of commodity by categorizing each commodity of input image into corresponding category with improved accuracy and reduced classification error.

Claims

1. A method of assessing quality of commodities, the method comprising:

receiving, by a processor of a quality assessment system, an input image of commodities captured by at least one image sensor coupled with the processor;

segmenting, by the processor, the input image to generate a plurality of segmented images of the input image;

determining, by the processor, a feature score for each of a plurality of generalized features identified based on random combination of pixels in each segmented image, wherein the plurality of generalized features is generated during training of labelled commodity images;

computing, by the processor, a regression score of each of the plurality of segmented images based on the feature score and a predetermined weightage score corresponding to the plurality of generalized features, wherein the weightage score is dynamically determined during training of labelled commodity images;

determining, by the processor, a likelihood score for each segmented image with respect to each of one of more predefined categories of commodities by comparing the regression score with predetermined regression score of each of the one or more predefined categories of commodities, wherein the predetermined regression score is determined during training of labelled commodity images; and

classifying, by the processor, each of the plurality of segmented images to one of the one or more predefined categories of commodities based on the likelihood score for assessing quality of the commodities.

2. The method as claimed in claim 1, wherein segmenting the input image to generate the plurality of segmented images comprising:

pre-processing the input image to generate a threshold image corresponding to the input image by removing background and related irregularities from the input image;

identifying one or more pixels corresponding to corner points of the commodities in the threshold image;

determining textural value for at least one pair of adjacent pixels of the one or more identified pixels; and

segmenting the threshold image into the plurality of segmented images based on the textural value of pair of adjacent pixels and geometrical shape of the one or more identified pixels.

3. The method as claimed in claim 1, wherein the commodities can be selected from grains, seeds and related objects, and the one or more predefined categories of the grains include at least broken grains, shriveled grains, weevilled grains, vertical broken, damaged, slightly damaged, organic impurities, inorganic impurities, and other edible grains.

4. The method as claimed in claim 1, wherein the training of labelled commodity images comprising:

receiving a plurality of training images of commodities of the one or more predefined categories and associated labels as input, wherein the plurality of training images is stored in an image repository coupled with the processor;

automatically generating the plurality of generalized features from the plurality of training images of each predefined category by repeated processing of random combination of pixels in the training images;

computing a predetermined weightage score corresponding to the plurality of generalized features of each predefined category based on weightage score successively determined for each of the plurality of training images; and

estimating the predetermined regression score for each of the predefined categories of commodities based on the predetermined weightage score and a final feature score determined for the corresponding feature.

5. The method as claimed in claim 4, wherein automatically generating the plurality of generalized features comprising:

deriving a plurality of features from the plurality of training images by random combination of pixels of the training images;

determining feature score for each of the plurality of features of training images based on intensity value of randomly combined pixels;

grouping the plurality of features having similar values of feature score to obtain the plurality of generalized features; and

determining the final feature score based on one of the values of feature score of the plurality of features upon grouping.

6. The method as claimed in claim 4, wherein computing the predetermined weightage score corresponding to the plurality of generalized features of each predefined category, comprising:

assigning random value to weightage score for each feature of a first training image of the plurality of training images;

the computing a regression score for the first training image using the random value of weightage score and feature score, wherein feature score for the first training image is assigned with a non-zero value based on the presence of feature in the first training image;

computing the regression score for a second training image of the plurality of training images using the previously assigned random values of the weightage score;

adjusting weightage score of each feature in the second training image to match the regression score of second training image with the nearest value of computed regression score of the first training image; and

determining the predetermined weightage score based on repeated adjusting of weightage score of each feature in remaining plurality of training images till the regression score matches with the value of computed regression score.

7. The method as claimed in claim 1, wherein classifying the plurality of segmented images comprising:

identifying highest value of the likelihood score for each segmented image among the likelihood scores determined with respect to predefined categories of commodities; and

classifying the segmented image to one of the predefined categories associated with the highest value of the likelihood score.

8. The method as claimed in claim 1, further comprising generating a quality analysis report indicating count of predefined categories of commodities in the input image based on classification of the plurality of segmented images.

9. A quality assessment system, comprising:

a processor;

at least one image sensor coupled with the processor; and

a memory communicatively coupled with the processor, wherein the memory stores processor-executable instructions, which on execution cause the processor to;

receive an input image of commodities captured by at least one image sensor coupled with the processor;

segment the input image to generate a plurality of segmented images of the input image;

determine a feature score for each of a plurality of generalized features identified based on random combination of pixels in each segmented image, wherein the plurality of generalized features is generated during training of labelled commodity images;

compute a regression score of each of the plurality of segmented images based on the feature score and a predetermined weightage score corresponding to the plurality of generalized features, wherein the weightage score is dynamically determined during training of labelled commodity images;

determine a likelihood score for each segmented image with respect to each of one of more predefined categories of commodities by comparing the regression score with predetermined regression score of each of the one or more predefined categories of commodities, wherein the predetermined regression score is determined during training of labelled commodity images; and

classify each of the plurality of segmented images to one of the one or more predefined categories of commodities based on the likelihood score for assessing quality of the commodities.

10. The system as claimed in claim 9, wherein the processor is configured to segment the input image to generate the plurality of segmented images by:

pre-processing the input image to generate a threshold image corresponding to the input image by removing background and related irregularities from the input image;

identifying one or more pixels corresponding to corner points of the commodities in the threshold image;

determining textural value for at least one pair of adjacent pixels of the one or more identified pixels; and

segmenting the threshold image into the plurality of segmented images based on the textural value of pair of adjacent pixels and geometrical shape of the one or more identified pixels.

11. The system as claimed in claim 9, wherein the commodities can be selected from grains, seeds and related objects, and the one or more predefined categories of the grains include at least broken grains, shriveled grains, weevilled grains, vertical broken, damaged, slightly damaged, organic impurities, inorganic impurities, and other edible grains.

12. The system as claimed in claim 9, wherein the processor is configured to train the labelled commodity images by:

receiving a plurality of training images of commodities of the one or more predefined categories and associated labels as input, wherein the plurality of training images is stored in an image repository coupled with the processor;

automatically generating the plurality of generalized features from the plurality of training images of each predefined category by repeated processing of random combination of pixels in the training images;

computing a predetermined weightage score corresponding to the plurality of generalized features of each predefined category based on weightage score successively determined for each of the plurality of training images; and

estimating the predetermined regression score for each of the predefined categories of commodities based on the predetermined weightage score and a final feature score determined for the corresponding feature.

13. The system as claimed in claim 12, wherein the processor is configured to automatically generate the plurality of generalized features by:

deriving a plurality of features from the plurality of training images by random combination of pixels of the training images;

determining feature score for each of the plurality of features of training images based on intensity value of randomly combined pixels;

grouping the plurality of features having similar values of feature score to obtain the plurality of generalized features; and

determining the final feature score based on one of the values of feature score of the plurality of features upon grouping.

14. The system as claimed in claim 12, wherein the processor is configured to compute the predetermined weightage score by:

assigning random value to weightage score for each feature of a first training image of the plurality of training images;

computing a regression score for the first training image using the random value of weightage score and feature score, wherein the feature score for the first training image is assigned with a non-zero value based on the presence of feature in the first training image;

computing the regression score for a second training image of the plurality of training images using the previously assigned random values of the weightage score;

adjusting weightage score of each feature in the second training image to match the regression score of second training image with the nearest value of computed regression score of the first training image; and

determining the predetermined weightage score based on repeated adjusting of weightage score of each feature in remaining plurality of training images till the regression score matches with the value of computed regression score.

15. The system as claimed in claim 9, wherein the processor is configured to classify the plurality of segmented images by:

identifying highest value of the likelihood score for each segmented image among the likelihood scores determined with respect to predefined categories of commodities; and

classifying the segmented image to one of the predefined categories associated with the highest value of the likelihood score.

16. The system as claimed in claim 9, wherein the processor is further configured to generate a quality analysis report indicating count of predefined categories of commodities in the input image based on classification of the plurality of segmented images.