# US 12,118,710 B2 · Process for the detection of bitter almonds based on the processing of digital images and a device associated therewith

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,118,710 B2
- Document ID: US12118710B2
- Status: Granted
- Kind: B2
- Application number: 17/506,718
- Filed: 2021-10-21
- Published: 2022-02-10
- Granted: 2024-10-15
- Earliest claimed: 2019-06-20
- Assignees: Universitat D'Alacant/Universidad De Alicante
- Applicants: Universitat D'Alacant/Universidad De Alicante
- Inventors: Guillermo Grindlay Lledó; Luis Gras Garcia; Juan Mora Pastor; Marta Navas García
- Web page: https://patentgrep.com/patent/US12118710B2

## CPC

- G01N 21/6456: {Spatial resolved fluorescence measurements; Imaging}
- G06T 7/0008: {checking presence/absence}
- B07C 5/342: according to optical properties, e.g. colour {(according to radiation transmittivity B07C5/3416)}
- G01N 21/6486: {Measuring fluorescence of biological material, e.g. DNA, RNA, cells (G01N21/6428 takes precedence)}
- G01N 21/85: Investigating moving fluids or granular solids
- G01N 33/025: {Fruits or vegetables}
- G06T 7/0004: {Industrial image inspection}
- G06T 7/10: Segmentation; Edge detection (motion-based segmentation G06T7/215)
- G06T 7/90: Determination of colour characteristics
- B07C 2501/0081: Sorting of food items
- G01N 2021/845: {Objects on a conveyor}
- G01N 2021/8466: {Investigation of vegetal material, e.g. leaves, plants, fruits}
- G01N 2021/8592: {Grain or other flowing solid samples}
- G01N 2201/0221: Portable; cableless; compact; hand-held
- G01N 2201/129: Using chemometrical methods
- G06T 2207/10024: Color image
- G06T 2207/10064: Fluorescence image
- G06T 2207/30128: Food products
- G01N: INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES (measuring or testing apparatus or processes other than immunoassay, involving enzymes or microorganisms C12M, C12Q)
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- B07C: POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING (specially adapted for a specific purpose covered by another class, see the relevant place, e.g. A43D33/06, B23Q7/12)

## Abstract

Procedure for the detection of bitter almonds based on the processing of digital images, and a system and device associated therewith. Detection procedure and system for the automated classification of sweet and bitter almonds based on the processing of digital images. The fluorescence of the cyanogenic compounds naturally present in almonds generates a clear difference in colour between sweet and bitter almonds which subsequently is analysed and classified by means of a computer program. The invention also includes the device, either portable or automatic, for carrying out the classification of bitter or sweet almonds. This device will be necessary during the goods reception process and in the validation/verification of the quality of the finished product, prior to the loading and transport process.

## Claims

### Claim 1

A procedure for the detection of bitter almonds, which comprises the following stages:
(a) placing at least one almond on a surface by means of a manual or automatic supply system;
(b) illuminating the almond with a source of ultraviolet (UV) radiation;
(c) acquiring an image of the almond with a photographic camera and storing said image in an internal or external system for subsequent analysis;
(d) processing said acquired image with a computer system that comprises a program for applying a discriminative model;
(e) classifying the almond as bitter or non-bitter according to said discriminative model; and
(f) withdrawing the almond by means of a manual or automatic withdrawal system, wherein stage (d) in turn comprises the following stages:
i. processing the data of the acquired image and segmentation of the image to establish the RGB colour parameters of a sample;
ii. transforming the RGB colour parameters obtained in stage i into at least one L*a*b* parameter of the CIE L*a*b* space;
iii. determining at least one of the L*, a* and/or b* parameters of the CIE L*a*b* space and its corresponding value in said image; and
iv. interpolating the value obtained in stage iii in said discriminative model generated from the values obtained with different sample types of bitter and non-bitter almonds acquired under predetermined conditions.

### Claim 2 (depends on 1)

The procedure for the detection of bitter almonds according to claim 1, wherein in stage (a), the sample is placed in a closed cell without penetration of sunlight.

### Claim 3 (depends on 1)

The procedure for the detection of bitter almonds according to claim 1, wherein the illumination of a cell is carried out with UV radiation having wavelengths<400 nm at a distance between the light source and the sample of between 0.1 m and 1 m.

### Claim 4 (depends on 1)

The procedure for the detection of bitter almonds according to claim 1, wherein the photography of the sample is performed at a distance between the camera and the sample of between 0.1 m and 1 m, under conditions of the absence of sunlight and application of UV radiation.

### Claim 5 (depends on 1)

The procedure for the detection of bitter almonds according to claim 1,
wherein:
said almond is a skinned almond;
stage (d) in turn comprises the following stages:
(i) segmenting the image into RGB channels and obtaining the average RGB values over all pixels of said image;
(ii) transforming the average RGB values into the corresponding CIE L*a*b* values;
(iii) interpolating:
a value Va from the discriminative model using the L* and a* values of said almond when said discriminative model is a combination of the L* and a* values of each member of a population comprising bitter almonds and non-bitter almonds, wherein said combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population, wherein bitter almonds are assigned to a range Ra1 of said combination and non-bitter almonds are assigned to a range Ra2 of said combination; or
a value Vb from the discriminative model using the L* and b* values of said almond, when said discriminative model is a combination of the L* and b* values of each member of said population, wherein said combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population, wherein bitter almonds are assigned to a range Rb1 of said combination and non-bitter almonds are assigned to a range Rb2 of said combination; and
stage (e) classifies the almond using the value Va or Vb obtained according to the discriminative model of stage (d), wherein said almond is classified as a bitter almond when:
Va falls inside range Ra1;
Va falls outside range Ra2;
Vb falls inside range Rb1; or
Vb falls outside range Rb2.

### Claim 6 (depends on 5)

The procedure for the detection of bitter almonds according to claim 5,
wherein
said discriminative model is developed by:
(I) performing stages (a) to (c) for each almond in said population comprising bitter almonds and non-bitter almonds;
(II) segmenting each image acquired for each almond in said population into RGB channels and obtaining the average RGB values over all pixels of each image;
(III) transforming the average RGB values of each image acquired for each almond in said population into the corresponding CIE L*a*b* values;
(IV) calculating a combination of:
the L* and a* values of each almond in said population; or
the L* and b* values of each almond in said population
wherein each combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population.

### Claim 7 (depends on 6)

The procedure for the detection of bitter almonds according to claim 6, which additionally comprises the following stage after stage (III) and before stage (IV) of performing multivariate data analysis by linear discriminant analysis which obtains:
a discriminant function when using the R, G and B parameters of each almond in said population; or
a discriminant function when using L*, a* and b* parameters of each almond in said population,
wherein:
the value of the discriminant function determined using the R, G and/or B parameters of each almond in said population; or
the value of the discriminant function determined using the L*, a* and/or b* parameters of each almond in said population discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population.

### Claim 8 (depends on 1)

The procedure for the detection of bitter almonds according to claim 1, wherein stage (c) comprises acquiring an image of multiple almonds, wherein each image of each almond is separated therefrom by image recognition software.

### Claim 9 (depends on 1)

A device for carrying out the procedure of claim 1,
wherein said device comprises at least:
a cell where the sample to be analyzed is placed according to step (a);
a UV light radiation system with at least one source of UV light configured to carry out step (b);
a positioned photographic camera which photographs the sample according to step (c) for subsequent analysis by means of a computer system; and the computer system is configured to preform classification of the sample into sweet and bitter almonds depending on a result obtained upon applying a discriminative model according to steps (d) and (e).

### Claim 10 (depends on 9)

The device for the detection of bitter almonds according to claim 9, wherein the UV light radiation system comprises a source of UV light that is applied with wavelengths<400 nm at a distance of between 0.1 m and 1 m between the light source and the sample.

### Claim 11 (depends on 9)

The device for the detection of bitter almonds according to claim 9, wherein the computer system comprises a computer programme which controls and processes the detection of the bitter almonds.

### Claim 12

A system for the detection of bitter almonds, which comprises the following:
(a) means for placing at least one almond on a surface by means of a manual or automatic supply system;
(b) means for illuminating the almond with a source of ultraviolet (UV) radiation;
(c) means for acquiring an image of the almond with a photographic camera and storing said image in an internal or external system for subsequent analysis;
(d) means for processing said acquired image with a computer system that comprises a programme for applying a discriminative model;
(e) means for classifying the almond as bitter or non-bitter according to said discriminative model; and
(f) means for withdrawing the almond by means of a manual or automatic withdrawal system,
wherein (d) in turn comprises the following:
i. means for processing the data of the acquired image and segmentation of the image to establish the RGB colour parameters of a sample;
ii. means for transforming the RGB colour parameters obtained in stage i into at least one L*a*b* parameter of the CIE L*a*b* space;
iii. means for determining at least one of the L*, a* and/or b* parameters of the CIE L*a*b* space and its corresponding value in said image; and
iv. means for interpolating the value obtained by iii in said discriminative model generated from the values obtained with different sample types of bitter and non-bitter almonds acquired under predetermined conditions.

### Claim 13 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein in (a), the sample is placed in a closed cell without penetration of sunlight.

### Claim 14 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein the illumination of a cell is carried out with UV radiation having wavelengths<400 nm at a distance between the light source and the sample of between 0.1 m and 1 m.

### Claim 15 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein the photography of the sample is performed at a distance between the camera and the sample of between 0.1 m and 1 m, under conditions of the absence of sunlight and application of UV radiation.

### Claim 16 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein: said almond is a skinned almond;
the processing of (d) in turn comprises the following:
(i) segmenting the image into RGB channels and obtaining the average RGB values over all pixels of said image;
(ii) transforming the average RGB values into the corresponding CIE L*a*b* values;
(iii) interpolating:
a value Va from the discriminative model using the L* and a* values of said almond when said discriminative model is a combination of the L* and a* values of each member of a population comprising bitter almonds and non-bitter almonds, wherein said combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population, wherein bitter almonds are assigned to a range Ra1 of said combination and non-bitter almonds are assigned to a range Ra2 of said combination; or
a value Vb from the discriminative model using the L* and b* values of said almond, when said discriminative model is a combination of the L* and b* values of each member of said population, wherein said combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population, wherein bitter almonds are assigned to a range Rb1 of said combination and non-bitter almonds are assigned to a range Rb2 of said combination; and
(e) classifies the almond using the value Va or Vb obtained according to the discriminative model of (d), wherein said almond is classified as a bitter almond when:
Va falls inside range Ra1;
Va falls outside range Ra2;
Vb falls inside range Rb1; or
Vb falls outside range Rb2.

### Claim 17 (depends on 16)

The system for the detection of bitter almonds according to claim 16, wherein said discriminative model is developed by:
(I) means for performing (a) to (c) for each almond in said population comprising bitter almonds and non-bitter almonds;
(II) means for segmenting each image acquired for each almond in said population into RGB channels and obtaining the average RGB values over all pixels of each image;
(III) means for transforming the average RGB values of each image acquired for each almond in said population into the corresponding CIE L*a*b* values;
(IV) means for calculating a combination of:
the L* and a* values of each almond in said population; or
the L* and b* values of each almond in said population
wherein each combination discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population.

### Claim 18 (depends on 17)

The system for the detection of bitter almonds according to claim 17, which, after transforming using (III) and before calculating using (IV), additionally comprises performing multivariate data analysis by linear discriminant analysis which obtains:
a discriminant function when using the R, G and B parameters of each almond in said population; or
a discriminant function when using L*, a* and b* parameters of each almond in said population,
wherein:
the value of the discriminant function determined using the R, G and/or B parameters of each almond in said population; or
the value of the discriminant function determined using the L*, a* and/or b* parameters of each almond in said population discriminates between the sub-population of bitter almonds and the sub-population of non-bitter almonds in said population.

### Claim 19 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein (c) comprises means for acquiring an image of multiple almonds, wherein each image of each almond is separated therefrom by image recognition software.

### Claim 20 (depends on 12)

The system for the detection of bitter almonds according to claim 12, wherein:
the means for placing at least one skinned almond on a surface comprises a manual or automatic supply or dispensing means;
the means for illuminating the almond with a source of ultraviolet (UV) radiation comprises a cell where the sample to be analysed is positioned and a UV light radiation system with at least one source of UV light;
the means for acquiring an image of the almond comprises a positioned photographic camera which photographs the sample for subsequent analysis by means of the computer system; and
the means for processing said acquired image comprises a computer system which comprises a programme for applying a discriminative model for classification of the sample into sweet and bitter almonds depending on a result obtained upon applying the discriminative model.