A method, system and computer program for analyzing a baked food item

Abstract

A method for analyzing the debris structure of a baked goods article includes the following steps: receiving visual data of the baked goods article at a mobile device, the visual data including real-time distance measurements from the mobile device to the baked goods article, wherein the visual data is collected in a light-controlled environment; segmenting the received visual data to identify the outline of the baked goods article; segmenting regions within the outline of the baked goods article; and using the results of the previous steps, estimating multiple debris structure features of the baked goods article by separating regions within the baked goods article based on pixel intensity, wherein regions with lower intensity correspond to open spaces or pores (i.e., holes, cavities), while regions with higher intensity correspond to debris.

Description

Technical Field

[0001] This invention relates to a method and system for analyzing baked goods. Background Technology

[0002] Baked food items are foods made from dough or batter and cooked by baking, a method of cooking food typically using long hours of dry heat in an oven. Bread is the most commonly baked item, but many other types of food can also be baked.

[0003] Industry professionals should understand that there are various standard parameters used to measure the quality of baked goods, such as the product's volume and structure.

[0004] In the food quality control process, food quality control experts and technicians typically collect and monitor samples of baked goods regularly. They conduct scientific analyses of these samples to determine whether the food quality meets adequate standards. This helps maintain critical consistency in the production of bread and other baked goods.

[0005] One of the existing tools for measuring food quality is the Calibre Control. TM C-Cell TM It provides debris structure data analysis for bread, cakes, and other structured products. C-Cell TM The system uses an entry-level digital imaging system to photograph slices of baked goods in a light-controlled environment (usually within the machine itself), and then processes the data on a computer connected to the device. C-Cell TM The system allows technical, production, and engineering personnel to set their own key performance indicators (KPIs) using objective data. In this way, food quality can be qualitatively determined. The tool requires a specific and constant focal length between the camera and the baked goods for accurate operation. This necessitates meticulous sample preparation, requiring the baked goods to be sliced ​​to precise thicknesses in a laboratory setting. For industrial sliced ​​bread, the system allows slicing in discrete 5mm increments within the device. This introduces a standard deviation for slices that do not correspond to multiples of 5mm, as such slices may not be within the optimal camera focus, leading to inaccurate size conversion. These factors often result in operational inefficiency and increased bias in analytical results. All of the above contribute to increased equipment size, complexity, and cost. Furthermore, those skilled in the art will understand that industrial equipment of this nature is not portable and requires specialized technical training for effective use.

[0006] US2022012467 discloses a multi-sensor analysis of food (such as baked goods) that utilizes various existing detection methods, such as standard cameras on mobile phones, LiDAR (light detection and ranging) sensors, and artificial intelligence engines. This document discloses the study of the caloric content of food by estimating its volume and composition. More specifically, the described techniques include receiving three-dimensional (3D) images, identifying food items within the 3D images, determining the volume of the identified food items based on the 3D images, and estimating the composition of the identified food using millimeter-wave radar.

[0007] However, the cited prior art does not provide a means of analyzing the debris structure within food articles, such as baked goods. Such analysis of the debris structure would provide insights into key debris quality attributes of baked goods, such as porosity level, cell count, slice volume, cell wall thickness, and concavity level. This could become an important component of multifaceted food quality assessment programs.

[0008] Therefore, a method is needed to analyze debris structures in this context.

[0009] Furthermore, one of the challenges in analyzing the crumb structure of industrially sliced ​​baked goods is the surface irregularity of the baked goods, which can be caused by certain processing conditions or formulation factors, such as internal temperature increases during slicing or changes in component quality (such as poor flour quality). Existing crumb structure analysis methods cannot adequately address these surface irregularities.

[0010] The purpose of this invention is to provide a method and system for analyzing baked goods, which offers advantages and solves some of the aforementioned problems and defects. Summary of the Invention

[0011] According to a first aspect of the present invention, a method according to technical solution 1 is provided, as well as a corresponding system and computer program as claimed.

[0012] This invention eliminates the need to maintain a fixed distance between the camera device and the baked goods article during visual data capture, thus allowing the use of portable devices to measure the crumb structure characteristics of the baked goods article. For example, as described below, visual data of the baked goods article can be captured handheld using a smartphone for subsequent analysis. This enables analysis to be performed from the desired location and improves operational efficiency by accelerating the analysis process.

[0013] Preferably, the present invention also provides a light-controlled environment comprising a light source configured to emit minimal or no radiation in the near-infrared range, and the light source being further configured to emit light parallel to the surface of a baked food article, wherein the baked food article is located in a light box, wherein the light source is configured to be located within the light box such that the light source emits light from each of a plurality of sides of the baked food article and through the height of the baked food article.

[0014] This solves the problem of uneven surfaces mentioned above, and addresses surface irregularities by emitting light from each side of the analyzed baked goods item placed inside the lightbox and across the entire height of the item. This allows for uniform light distribution even if the quality of the cut of the baked goods item and / or irregularities exist on its surface. Attached Figure Description

[0015] These and other features of the invention will become apparent from the following description of an example described with reference to the accompanying drawings, in which: Figure 1 The location of the smartphone is shown during the analysis, which includes distance measurement methods; Figure 2 The design of a lightbox device according to an aspect of the present invention is shown; Figure 3 The image shows the light distribution from the light panel at the edge of the lightbox device, as well as an image of a slice of bread; Figure 4 An application view of the functional blocks of a system according to a preferred embodiment is shown; Figure 5A The image shows a view of a smartphone device and a mobile application screen that opens before capturing visual data of baked goods. Figure 5B The figure shows an open mobile application screen where the SOD method, specifically the U2-Net model, is used to identify the outlines of baked goods. Figure 6 An example of an image in its binary representation after software processing is shown; Figure 7 The application view shows the results of the analyzed product; Figure 8 A flowchart of a method for analyzing baked food articles according to an aspect of the present invention is shown; and Figure 9 A computer according to an aspect of the invention is shown, comprising a set of instructions for causing the computer to perform any one or more of the methods described herein. Detailed Implementation

[0016] The following description of the invention is provided as a illustrative teaching of preferred embodiments of the invention. Those skilled in the art will recognize that many changes can be made to the described embodiments while still obtaining the beneficial results of the invention. It will also be apparent that some of the intended benefits of the invention can be obtained by selecting some features of the invention without utilizing others. Therefore, those skilled in the art will recognize that modifications and adaptations to the invention are possible, and even desirable in some cases, and are part of the invention. Thus, the following description is provided as an explanation of the principles and limitations of the invention.

[0017] Figure 1 The figure illustrates the positioning of a smartphone 102 relative to a lightbox 104 during the capture of visual data associated with a baked food article 106 (such as a slice of bread) for analyzing the crumb structure of the baked food article 106, as will be described below. In one embodiment of the invention, the visual data is provided in the form of one or more images, one or more videos, and / or one or more three-dimensional (3D) images of the baked food article. Figure 150 depicts a typical sensor configuration on a modern smartphone. A portion of the visual data, such as one or more images and / or one or more videos, is captured by one or more smartphone cameras 107. In one embodiment of the invention, one or more three-dimensional images are provided in the form of three-dimensional image data (such as computer vision data). In another embodiment, the three-dimensional image data is provided in the form of three-dimensional mesh data or point cloud data. In an embodiment, a ranging light detection and ranging (LiDAR) sensor 108 is used to capture three-dimensional image data that includes distance information as part of the visual data. LiDAR pulse 108a is emitted from LiDAR sensor 108 of smartphone 102 and reflects off the surface of the baked goods article, thereby returning pulse 108b to LiDAR sensor 108 of smartphone 102, providing real-time data on the distance from smartphone 102 to baked goods article 106. The smartphone can be held by the user during the analysis. In this embodiment, the distance from smartphone 102 to baked goods article 106 is determined by analyzing LiDAR data received from a rectangular area located at the center 110 of baked goods article 106. The real-time distance data, combined with known camera characteristics, is used to calculate spatial resolution. Spatial resolution is preferably defined as the physical size of pixels representing an image. This is used to estimate debris structure properties such as cell wall size, cell diameter, and other relevant characteristics.

[0018] The spatial resolution calculation process is performed in two steps: (a) determining the sensor's true height dimension (hereinafter referred to as "THS"), and (b) calculating the spatial resolution based on the image resolution, once the THS value is known. As determined by the sensor's height, the sensor's true height dimension refers to the actual vertical field of view distance captured by the smartphone camera.

[0019]

[0020]

[0021] For example, the LiDAR sensor has received data indicating that the distance between the smartphone 102 and the baked goods item 106 is 250 mm. The smartphone's camera lens has a focal length of 6.7 mm, and the sensor height is 7.3 mm. The image resolution is 2800×2800 pixels. Then the following calculations are performed:

[0022]

[0023] like Figure 4 As shown, spatial resolution will be used in the third software module 414 (as described below) to calculate specific debris structure features, such as element wall dimensions and element diameters. The results of this calculation will be presented later, as shown in... Figure 7 What is depicted.

[0024] In this embodiment, the LiDAR sensor emits waves with a wavelength of approximately 940 nm. The lightbox 104 is equipped with an electroluminescent (EL) panel that emits cool white light in the wavelength range of 400 to 650 nm, producing minimal near-infrared radiation. This configuration ensures minimal interference with the operation of the LiDAR sensor, thereby optimizing the sensor's performance in capturing accurate real-time distance data. In some embodiments, the filter of the LiDAR sensor can be adapted to any light source within the lightbox. In yet another embodiment, monochromatic light can be used as the light source for the lightbox.

[0025] Figure 2 A lightbox device 200 according to an embodiment of the present invention is shown. The lightbox assembly features a light-emitting EL panel 202, which ensures uniform (consistent) illumination within the lightbox. The EL panel 202 emits light parallel to the surface of the baked goods article placed for analysis, thereby producing accurate shadows in the pores / voids throughout the entire surface. The EL panel is made of a flexible material and is preferably composed of a single piece. The EL panel can be reattached to the base of the lightbox 204, and is preferably connected via Velcro. TMConnectors 206a and 206b are assembled to the base. For structural integrity of the EL panel, additional Velcro connectors are preferably added. TM Connector 208.

[0026] The lightbox assembly also includes a power supply 210 and a custom-designed printed circuit board (PCB) 212. The power supply 210 can be provided in the form of a lightweight rechargeable battery, such as a lithium-polymer (LiPo) battery. The custom PCB 212 is configured to manage power functions, including charging and voltage regulation of the power supply 210. Access to both the PCB 212 and the power supply 210 is protected by a cover 214.

[0027] The lightbox assembly 200 creates a controlled lighting environment, which is highly useful for consistent and accurate analysis of the structure of baked goods crumbs. In some embodiments of the invention, the method includes using a flash to control the light / shadow distribution on the surface of the baked goods article. In alternative embodiments, the base of the lightbox is designed with a predefined color, such as white, to ensure accurate white balance during visual data acquisition.

[0028] Figure 3 An example of light distribution in lightbox 300 is shown. EL lamp panel 302 emits light parallel to the surface of baked goods article 304 located within the lightbox, thus providing illumination from each side and across its entire height. This arrangement ensures that shadows are cast only on the structural pores of the baked goods article, thereby accurately reflecting its crumb structure. For example, image 352 shows the difference in light distribution between a conventional industrial system 354 and an embodiment 356 of the present invention. In the conventional system, due to suboptimal light distribution, visible markings from crust deformation 358 on bread slices and the slicing knife 360 ​​are mistaken for pores. However, when analyzing the same baked goods article using an embodiment of the present invention, no shadows are formed around the crust deformation 362 and the uneven slices 364 of the baked goods article, resulting in improved accuracy in detecting crumb structure by preventing erroneous pore detection. This enhanced light distribution provides more reliable data on the crumb structure characteristics of the baked goods article.

[0029] The above description led to the collection of visual data about baked goods using a smartphone. The collected information was then fed into a series of software modules, such as... Figure 4 As shown in the image.

[0030] exist Figure 4 In accordance with a preferred embodiment of the present invention, a system for analyzing baked food articles is generally described with reference to reference numeral 400.

[0031] System 400 is provided as an example of a mobile device 40, which includes a camera 402, a LiDAR sensor 404, and software 406. System 400 also includes a database 408.

[0032] As will be described below, in use, software 406 (comprising a first software module 410, a second software module 412, and a third software module 414, as described below) receives visual data from camera 402 and LiDAR sensor 404, and processes such data to perform an analysis of the crumb structure of baked goods.

[0033] The first software module 410 segments the visual data received from the camera 402 to identify the outline of the baked goods item 106. This means identifying the position of the baked goods item 106 across all or one or more images or videos, and separating or “segmenting” the position of the baked goods item from the rest of the content of one or more images or videos. In one embodiment, the step of segmenting the visual data includes using a salient object detection (SOD) method. In this embodiment, the salient object detection method is provided as an example of an artificial intelligence (AI) tool operable to extract the outline of the analyzed baked goods item. In this embodiment, the AI ​​tool is provided as an example of a U-2 Net model. It should be understood that different neural networks can also be used for image segmentation.

[0034] The identified outline of the baked goods item is further sent to a second software module 412, where another stage of segmentation of the region within the outline is performed. In one embodiment of the invention, the step of segmenting the region within the outline of the baked goods item involves dividing the image into multiple segments or regions based on similarity characteristics. In one embodiment, image segmentation is performed using a thresholding technique. The threshold in the thresholding technique is defined as a specific intensity value, which serves as a boundary for classifying pixels within the image. The threshold is selected such that it minimizes the variance within intensity groups and maximizes the variance between intensity groups. Once the threshold is applied, the image is converted to a duotone representation (also known as a binary representation), as shown in... Figure 6 As shown in the diagram, pixels with intensity values ​​below a threshold correspond to pores or open spaces, such as holes or cavities 602, while pixels with higher intensity values ​​represent solid or filled areas, such as debris 604. This binary transformation allows for clear identification and separation of regions.

[0035] In another embodiment, the step of segmenting the region within the contour includes applying a clustering algorithm to the image, such as the k-means clustering algorithm. The k-means clustering algorithm is a well-known vector quantization method that divides (n) observations into (k) clusters, where each observation belongs to the cluster with the nearest mean (as a prototype of the cluster). The clustering algorithm groups a set of data into clusters, where the data arrays consist of pixels with different intensities. Depending on the structure and segmentation of the image data, each cluster formed by the k-means clustering algorithm can include multiple data arrays. A single data array can include pixels with different intensities because the clustering algorithm evaluates pixels and assigns them to clusters based on intensity similarity relative to the centroid. In this context, the centroid represents the average value of the features (e.g., pixel intensity) of the data points within the cluster. This allows pixels to be grouped into clusters, thereby facilitating the segmentation of different regions within the image. For example, the k-means algorithm where k=2 produces a cluster similar to... Figure 6 The binary file shown.

[0036] In some embodiments, an image preprocessing step can be used to improve image quality, reduce noise, enhance contrast, and prepare the image for analysis and further processing. This image preprocessing step can include one or more of the following techniques: color normalization, contrast preprocessing, and / or noise reduction. The image preprocessing step can also be used for image normalization.

[0037] The final segmentation data received by LiDAR sensor 404, combined with distance data from smartphone 40 to the baked goods item, is forwarded to third software module 414 for estimating multiple debris structure features, including but not limited to porosity, cell wall size, and cell diameter. In the current embodiment, the spatial resolution calculation described above is performed in third module 414. In another embodiment, the spatial resolution calculation can be performed and stored at any earlier stage until the debris structure features need to be estimated. The estimation of multiple debris structure features is performed using known image processing methods (including but not limited to methods based on image processing libraries such as OpenCV). The output of third software module 414 is multiple debris structure features, such as porosity level, cell diameter, cell wall size, concavity level, etc., and in Figure 7 The diagram is further illustrated in the text.

[0038] During operation, database 408 contains product quality parameters, which can be compared with results generated by software 406 to provide product quality assessment results.

[0039] Because light boxes were used during the acquisition of visual data (such as in Figure 3As depicted in Figure 352, software module 412 receives high-quality visual data regarding the light-shadow distribution on the surface of the baked goods article 106. Simultaneously, the low near-infrared radiation emitted by the EL panel 302, used as a light source, does not interfere with the LiDAR pulses 108a-b, thus allowing accurate acquisition of visual data related to the distance between the baked goods article 106 and the smartphone 102, such as... Figure 1 As shown in the diagram. In the current embodiment, all software modules execute on the smartphone. However, in alternative embodiments, the modules may execute individually, jointly, or in any combination thereof, either in the cloud or on the smartphone.

[0040] Figure 5A The diagram shows a view of a smartphone device 500 and a mobile application screen 502 that opens before capturing visual data of a baked goods item 504. A visible wireframe 506 on the screen guides the user to adjust the distance between the smartphone 500 and the baked goods item 504 so that it visually aligns with the wireframe 506. A capture button 508 is provided, enabling the user to initiate the acquisition of visual data of the baked goods item and identify its outline.

[0041] Figure 5B A smartphone device 550 with an open mobile application screen 552 is shown, illustrating the outline 554 of a baked goods item and its internal area. A confirmation button 556 is provided, allowing the user to verify the accuracy of the outline segmentation and submit data about the outline and its internal area for further processing. If outline segmentation must be re-performed, the user can do so by clicking button 558, which will take the user to... Figure 5A The preceding process steps described above. The received data regarding the real-time distance 560° between the smartphone and the baked goods is displayed.

[0042] Figure 6 The illustration shows an image of a baked goods item 600 after the user activates the confirmation button 556. This is an example of the image in a binary representation after processing by software module 412, where pixels with different intensities are grouped together. As a result, black areas 602 in the image represent pores, such as holes or cavities, while white areas 604 represent debris. This image is further sent to software module 414 for estimating multiple debris structural features and displaying the analysis results to the user, as described below.

[0043] Figure 7A smartphone device 700 with a mobile application screen 702 is shown, displaying the results of a debris structure analysis of baked goods articles 704a-704f. The debris structure characteristics of each analyzed baked goods article 704a-704f are estimated and displayed. For example, the porosity 706 of slice 704d is shown. Furthermore, the concavity level 708 and unit wall size 710 of baked goods articles 704d and 704e are presented respectively. In one embodiment, the images are processed in real time. In an alternative embodiment, the images are queued for later analysis.

[0044] exist Figure 8 The figure shows a flowchart of a method for analyzing baked food articles according to one aspect of the present invention, generally described with reference to reference numeral 800.

[0045] The method includes a first step of opening a mobile application in box 802, and a second step of creating an item (e.g., a folder) in box 804 to store visual data and debris structure feature estimation results. This step can be performed before or after the item analysis. In this embodiment, the item is stored on cloud storage. In yet another embodiment, the item can be stored on a smartphone.

[0046] The next step in box 806 involves placing the baked goods item inside the lightbox. The next step in box 808 involves activating the camera and positioning it above the baked goods item. It should be understood that a 180° angle between the surface of the product being analyzed and the mobile device is ideal. In this context, the camera on the mobile device provides focus on the object with / without LiDAR sensor support.

[0047] Once the area being analyzed is in focus, visual data can be acquired in frame 810. There may be a 0-30 s delay in receiving the visual data so that the LiDAR sensor can identify the distance to the baked goods item. In this embodiment, both an RGB camera (810a) and a LiDAR sensor (810b) are preferably used to perform the operation of receiving visual data. Alternatively, the focus between two or more cameras can be used for distance measurement, or a combination of distance data provided by the LiDAR sensor and data from two or more cameras (i.e., stereo cameras) can be used.

[0048] In box 812, the received visual data is segmented using the U2-Net model (812a), and the outlines of the analyzed baked goods items are identified.

[0049] In box 814, the step of segmenting the region within the identified contour (as described in module 414) involves dividing the image into multiple segments or regions based on similar features.

[0050] By allowing the estimation of multiple debris structural features, the process in block 816 enables further analysis of baked goods articles. Various structural features, including those requiring spatial resolution values, can be quantified by explicitly distinguishing between debris and pores or open spaces (i.e., holes, cavities). In the current embodiment, spatial resolution is calculated in block 816a.

[0051] In box 818, the visual data and analysis results are saved to the database.

[0052] In the current embodiment, also as Figure 7 As shown in the diagram, the analysis results are further displayed to the user in box 820.

[0053] exist Figure 9 A computer is provided, in which a set of instructions can be executed to cause the computer to perform any or more methods described herein. According to embodiments of the invention, the computer is generally described with reference to reference numeral 900.

[0054] In a networked deployment, computer 900 can operate as a server or client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. Computer 900 can be a personal computer (PC), tablet computer, set-top box (STB), personal digital assistant (PDA), cellular phone, web device, network router, switch, or bridge, or any computer 900 capable of executing a set of instructions (sequentially or otherwise) specifying the actions to be taken by computer 900.

[0055] Furthermore, although only a single computer 900 is illustrated, the term "computer" should also be considered as including any collection of computers that individually or jointly execute a set (or more) of instructions to perform any one or more of the methods discussed herein.

[0056] Example computer system 900 includes processors 902 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), or both) communicating with each other via bus 901, main memory 904, and static memory 906. Computer 900 may also include a video display unit 910 (e.g., a liquid crystal display (LCD)). Computer 900 also includes an alphanumeric input device 912 (e.g., a keyboard), a user interface (UI) navigation device 914 (e.g., a mouse), a disk drive unit 916, a signal generation device 918 (e.g., a speaker), and a network interface device 908.

[0057] The disk drive unit 916 includes a computer-readable medium 922 on which one or more sets of instructions and data structures (e.g., software 924) are stored, implementing or utilizing any one or more methods or functions described herein. During execution of software 924 by computer system 900, software 924 may also reside wholly or at least partially within main memory 904 and / or processor, which also constitute computer-readable media. Therefore, for clarity, it should be noted that during execution of software 924 by computer system 900, if software 924 is not in main memory 904 and / or processor, it will reside in a cloud-based storage location or a remote storage location and can be executed directly from there.

[0058] The software 924 can also be sent or received via network interface device 908 through network 926 using any of a variety of well-known transmission protocols (e.g., secure HTTP, FTP).

[0059] Although the computer-readable medium 922 shown in the example embodiment is a single medium, the term "computer-readable medium" should be considered to include a single medium or multiple media (e.g., a centralized or distributed database and / or associated caches and servers) that store one or more sets of instructions.

[0060] The term "computer-readable medium" should also be considered to include any medium capable of storing, encoding, or carrying a set of instructions for execution by computer 900 and enabling computer 900 to perform any one or more methods of this embodiment, or any medium capable of storing, encoding, or carrying data structures used by or associated with such set of instructions.

[0061] Therefore, the term "computer-readable media" should be considered to include, but is not limited to, solid-state storage, optical media, magnetic media, and cloud storage options such as Amazon Web services. TM Microsoft Azure TM wait).

[0062] It should be understood that the present invention is not limited to the specific details given herein by way of example only, and various modifications and alterations may be made without departing from the scope of the invention as defined in the appended claims.

Claims

1. A method for analyzing the crumb structure of baked goods, comprising the following steps: a. Receives visual data of the baked goods item at a mobile device, the visual data including real-time distance measurements from the mobile device to the baked goods item, wherein the visual data is collected in a light-controlled environment; b. Segment the received visual data to identify the outline of the baked food item; c. Divide the region within the outline of the baked food article; as well as Using the results of step c, multiple debris structure features of the baked food article are estimated by separating regions within the baked food article based on pixel intensity, wherein regions with lower intensity correspond to open spaces or pores, while regions with higher intensity correspond to debris.

2. The method according to claim 1, wherein, Several debris characteristics include the porosity level, unit wall size, and concavity level of the baked food article.

3. The method according to any one of the preceding claims, wherein, The step of receiving visual data of the baked goods item includes accessing the camera and LiDAR sensor on the mobile device to capture one or more images of the item and / or determine the distance to the item.

4. The method according to claim 3, wherein, The steps of receiving visual data include activating the camera on the mobile device and positioning it above the object in a horizontal position.

5. The method according to claim 1, wherein, Step d is performed using the calculated spatial resolution.

6. The method according to claim 5, wherein, The spatial resolution is calculated based on the sensor's actual height and image resolution.

7. The method according to any one of the preceding claims, wherein, The method includes using a flash to control the light / shadow distribution on the surface of the baked food article.

8. The method according to claim 7, wherein, The method includes using one or more application algorithms to calculate flash mode parameters for real-time image processing.

9. The method according to any one of the preceding claims, wherein, Step b includes using a salient object detection (SOD) method.

10. The method according to claim 9, wherein, The salient object detection method is provided as an example of an artificial intelligence tool, i.e., an AI tool, which is operable to extract one or more boundaries of the analyzed baked goods article.

11. The method according to claim 10, wherein, The AI ​​tools are provided as an example of the U-2 Net model.

12. The method according to claim 1, wherein, Step c applies a thresholding technique, wherein a threshold is selected, and the threshold is defined as a specific intensity value used as a boundary for classifying pixels within an image.

13. The method according to claim 12, wherein, After applying the thresholding technique, the image is converted to a duotone representation, wherein the segmentation algorithm groups the pixels according to the intensity level of the pixels associated with the selected threshold.

14. The method according to claim 1, wherein, Step c uses the k-means clustering algorithm.

15. The method according to claim 14, wherein, The k-means clustering algorithm is a vector quantization method that divides n observations into k clusters, where each observation belongs to the cluster closest to the mean.

16. The method according to any one of claims 14 or 15, wherein, The clustering algorithm groups a set of data into multiple clusters, where each cluster has multiple data arrays, and each data array includes pixels representing various intensities.

17. The method according to claim 16, wherein, The clustering algorithm assigns pixels to clusters based on the similarity of intensity values ​​related to the centroid.

18. The method according to claim 17, wherein, The centroid represents the average pixel intensity of the data points within the cluster.

19. The method according to claim 1, wherein, The light-controlled environment includes a light source configured to emit minimal or no radiation in the near-infrared range.

20. The method according to claim 19, wherein, The light source is configured to emit light parallel to the surface of the baked food article, wherein the baked food article is located in a light box, wherein the light source is located in the light box in such a way that the light source emits light from each of a plurality of sides of the baked food article and through the height of the baked food article.

21. A system comprising means adapted to perform all steps of the method according to any one of the preceding method claims.

22. A computer program comprising instructions for performing all steps of the method according to any one of the preceding method claims when the computer program is executed on a computer system.

Images