Device for sorting camelid fibres

Abstract

The present invention relates to a method and device for sorting alpaca fibres, in which an automated device sorts the portions of fleece, which have been previously separated manually according to the diameter range, using an identification belt with transport rollers that move and align the portions. A pressure roller and a transparent glass plate ensure that the portions remain on a single horizontal plane for uniform analysis. An industrial photographic camera with a magnifying lens and a light source is used to capture microscopic images that are processed by artificial intelligence (AI) to sort the fibres according to their quality. The sorting belt, equipped with presence sensors and mechanical pistons, guides the portions to the correct containers, while a robotic arm collects and redirects incorrect samples for re-sorting.

Description

CAMELID FIBER CLASSIFICATION EQUIPMENT

[0001] The following invention is developed in the technical field of industrial automation, especially in the industrial textile classification of camelid fibers using artificial intelligence.

[0002] It is currently known that the classic classification of alpaca fibers consists of separating the fleece into portions or pieces that possess similar characteristics, separating fine parts from coarse ones, identifying them within quality groups, transporting the identified portion of the fleece, and then storing it. The classification is carried out from the fleece (which generally arrives packaged as a "drum"), which is separated into portions or pieces that have similar color, breed origin, length, softness, and whose fiber diameter falls within a specific range (Zarate, 2012). This process is performed at the Classification Plants in accordance with the Peruvian technical standard (NTP) 231.301 (INACAL, 2022).It is carried out by specialized personnel who use sight and touch to determine fiber quality (Saldana, 2017), and those in charge are called "master classifiers," who, according to their experience, can be: a) Representative or review master classifiers, who guide, control, and supervise the progress of the classification work; and b) Master Classifiers of the Sorting Process, who classify the alpaca fiber according to the technical standard (Lencinas & Torres, 2010). Therefore, currently, alpaca fleece classification is done manually, using touch and sight, which has many drawbacks from the point of view of labor, cost, hygiene, and the requirement for personnel with extensive training, making the work subjective and dependent on the qualified experience of the staff.

[0003] Due to the inherent nature of the work, recruiting skilled sorters is not easy. The process is labor-intensive, labor costs have continued to rise, and stringent testing standards mean that the consequences of errors are severe, further increasing the experience requirements for technicians (Lingzhong & Jingbo, 2020). Faced with a large quantity of alpaca fiber urgently needing sorting, traditional labor problems such as low screening efficiency, high screening standards, and high labor costs, as well as health issues stemming from the fiber's dust and other contaminants, current methods can no longer meet market demands. Therefore, the industry, fiber traders, and wholesalers are urgently calling for the mechanization and automation of this process.

[0004] There are inventions such as patent CN114044280A “Garbage sorter using image recognition through artificial intelligence” that perform the dispersal of garbage and classify it into only two groups; Recyclable and non-recyclable garbage; while invention CN111196454A “Garbage sorting system based on machine learning”, which consists of a camera that captures the image of the garbage and is integrated into the image recognition module that allows the type of garbage to be identified, while the results processing module provides the result on the screen and by voice, and at the same time controls the switch of the garbage can, opens the garbage disposal port of the corresponding type for garbage delivery;However, these inventions refer specifically to the classification of garbage, but not of alpaca fibers; these inventions do not perform an analysis of the grouped fibers through photographs and classifying them into up to seven types.

[0005] On the other hand, patent CN204310226U refers to the utility model for “automatic wool sorting and a collecting structure with an identification function.” This refers to a structure that allows for the automatic movement and collection of pre-sorted fibers into their respective containers using baskets, which are identified by a digital tag. This invention is intended for wool and does not perform any sorting from the fleece, and even less so does it work with alpaca fibers. However, it facilitates the trimming of wool for sorting and direct storage, thus reducing the workload of a sorter. It also has an electronic chip that stores the content of the tags, facilitating subsequent wool processing and ensuring that the amount of wool cut is accurate and precise (Tianjiao, 2015).

[0006] Also, there is patent document CN112304229A (Lingzhong & Jingbo, 2020), which refers to a method and system for the automatic analysis of fibers, such as alpaca fiber. The invention relates to an automated method and system for analyzing textile fibers, in particular to an artificial intelligence method and system for: a) automatically measuring the diameter of textile fibers, b) automatically discriminating the composition of textile fibers, and c) automatically calculating the textile fiber content. However, it does not classify fibers by quality according to any technical standard, but it does use artificial intelligence to discriminate between different types of fibers.

[0007] The Peruvian Technical Standard (NTP) 231.301:2022 was published in 2022 and defines nine grades of alpaca fiber, determined primarily by the average fiber diameter (AFD). These grades range from Ultrafine Alpaca, with a fineness of 18 µm or less, to Coarse Alpaca, with a diameter greater than 31.5 µm. It also includes Short Alpaca, characterized by shorter fibers between 20 and 50 mm in length, and MP Alpaca, composed of short, coarse, and damaged fibers.

[0008] Each category also specifies a minimum staple length: 65 mm for the finest fibers, such as Ultrafine, Superfine, and Extrafine, and 70 mm for the coarsest fibers, such as Fine, Semifine, Semicoarse, and Coarse. In the case of Short Alpaca, the fibers range in length from 20 to 50 mm. This classification is essential for determining the intended use and application of the fibers based on their quality. It is well known that there is significant variability in DMF (Digital Fiber Mass) within alpaca fleeces (Aylan-Parker & McGregor, 2002; McGregor et al., 2012); therefore, a single fleece can contain several different qualities of alpaca fiber. Description of the Invention

[0009] To overcome the limitations mentioned above, a CAMELID FIBER CLASSIFYING EQUIPMENT, called FIBER CLASS, has been developed. This equipment allows the classification of camelid fibers such as alpaca, vicuña, llama, among others, according to specific parameters that comply with the Peruvian Technical Standard (NTP) 231.301:2022, or with any other set of parameters related to the DMF, through an automated system that optimizes both the accuracy and efficiency of the process, without the need for intensive manual intervention, thus optimizing both the speed and precision of the process.

[0010] For the implementation of the classification system, a DMF variation spectrum (Figure 1) was developed based on the evaluation of several alpaca fleeces from different ages and sexes. As a result, each fleece is divided into seven zones: neck, upper back, middle back, lower back, belly, front region, and rear region. This information on the distribution of fibers in different zones serves as the basis for the design of the classification system, allowing the fleece to be classified without requiring the operator to have specialized skills in identifying the different fiber qualities.

[0011] The sorting process begins when the fleece is placed on the adjustable table, where the portions are separated according to the seven zones mentioned. Portions containing short or damaged fibers are identified and manually sorted as "Short Alpaca" or "Alpaca MP," while the rest of the fleece continues with the automated process.

[0012] After obtaining the fleece portions (except those corresponding to Short Alpaca and MP Alpaca), these are placed on the identification belt, which is equipped with two transport rollers. As these rollers rotate, they move the fleece portions to the middle of the belt. At this point, the fleece portion is pushed under a metal plate by a pressure roller, which allows it, despite its initial volume, to fit under the transparent glass, reducing its height and ensuring that the surface remains in a single horizontal plane, always in contact with the glass.

[0013] Once in this position, at least one industrial camera and a magnifying lens, joined by a custom-sized spacer, are used to ensure an ideal depth of field for obtaining microscopic photographs (B) of the fleece portion. The light source provides adequate illumination, and the camera performs the scan, capturing at least 80 images in no more than 40 seconds. These images are crucial for the subsequent identification of fiber quality using the artificial intelligence (AI)-based system.

[0014] The captured images are processed by an artificial intelligence (AI) system, which analyzes them to determine the quality of the fibers present in each portion of fleece, classifying them according to their fineness and specific characteristics. This process is carried out quickly and accurately, with the aim of classifying each portion of fleece according to its category. The finest fibers, such as Ultrafine, Superfine, and Extrafine, are 65 mm long, while the coarsest fibers, such as Fine, Semifine, Semicoarse, and Coarse, are 70 mm long. In the case of Short Alpaca, the fibers have a length that varies between 20 and 50 mm. This classification is essential for defining the destination and use of the fibers according to their quality, as indicated in NTP 231.301:2022.

[0015] Once the AI ​​has determined the fiber quality, the fleece portions are transferred to a sorting system, which activates mechanical pistons to direct the portions to the appropriate containers based on their quality. If the system detects an error or misclassification, a robotic arm retrieves the incorrect samples and returns them to the process for reclassification.

[0016] This equipment, called FIBER CLASS, automates the sorting of alpaca fibers, reducing manual intervention and ensuring accurate and rapid classification, thus improving the efficiency of high-quality yarn and fabric production. Furthermore, the equipment's modular design, equipped with presence sensors and a robotic arm to correct potential errors, guarantees efficient and error-free sorting, adapting to process needs and optimizing the use of available resources.

[0017] Spectrum of diameter variation in alpaca fleece, where the right side corresponds to the neck, and the red area corresponds to the belly of the animal. This visualization is comparable to a graphic representation of the animal's skin sections under analysis.

[0018] Process diagram of the operation of the camelid fiber sorting equipment.

[0019] Animal fiber sorting equipment, showing its different parts.

[0020] Identification band, showing its different parts

[0021] Identification module (interior view), showing its different parts, with assembled light source.

[0022] Identification module (interior view), showing its different parts, without a light source to observe other parts that make it up.

[0023] Classification band, showing its different parts.

[0024] Robotic arm with camera and computer

[0025] Table of fineness ranges in µm for each fiber group classification PREFERRED EMBODIMENT OF THE INVENTION

[0026] The camelid fiber sorting equipment includes an adjustable table (3) for placing the fleece for analysis. The table height can be adjusted by moving the legs and securing them with a clamping mechanism. The table surface incorporates a mesh (4) with a grid-like structure, the dimensions of which do not exceed 5 cm per side, designed to filter contaminants such as dust and plant debris. This process can be optimized by manual vibrations and / or repeatedly rotating the fleece. Subsequently, the fleece is manually divided into seven fleece portions (2): neck, upper back, middle back, lower back, belly, front region, and rear region, according to a DMF variation spectrum determined after evaluating multiple fleeces from alpacas of different ages and sexes.The portions of the lower limbs, called Short Alpaca, and the short, thick or deteriorated fibers, classified as MP Alpaca (Figure 9: table), are excluded from the analysis.

[0027] The selected portions are transported by a continuous identification belt (5) equipped with two transport rollers (30) that push the fleece towards a metal plate (6). Here, a pressure roller (7) flattens the portions, allowing them to pass under a transparent glass (8), ensuring that they lie on a uniform horizontal plane for optical analysis.

[0028] An industrial camera (9) and a magnifying lens (10), connected by a spacer (11) that ensures the appropriate depth of field, capture microscopic images under the illumination provided by an LED light source (12). The system can scan and capture up to 120 images in an estimated time of 40 seconds.

[0029] A trapezoidal screw (13) allows the initial working distance of the industrial camera (9) to be calibrated and fixed by turning a knob (29) clockwise or counterclockwise. Furthermore, the strategic connection between the camera support (14) and the metal plate (6) also ensures that the working distance remains fixed and the focus point constant. If the horizontal plane of the metal plate changes in the future due to pressure or other factors, the camera support will adapt to this change or deformation, maintaining the focus point and the initially established working distance to obtain sharp, focused photographs.

[0030] A presence sensor (15)(A1), located on the identification belt (5), detects (A) the entry of a portion of fleece (2) into the identification module (16) and sends a signal to the electronic control module (18), which in turn communicates with a computer (17). This process automatically activates the image capture system and the artificial intelligence (AI)-based analysis.

[0031] On the computer (17), the images are processed (C) by an Artificial Intelligence model that resizes and transforms them into tensors for analysis. Convolutions are applied to extract relevant fiber features, ultimately selecting 1280 key attributes that allow the fleece to be classified (D) into one of seven predefined categories (Figure 9: table). Processing capacity depends on the available hardware, and can handle individual images or batches using specialized CPUs or GPUs with tensor cores.

[0032] Once the quality has been determined by the computer (17), the classified portions are redirected by a drop guide (19) towards the sorting belt (20). Presence sensors (15 and 21) (D1) monitor the movement (D2) of the portions and send signals (D3) to the electronic control module (18), which regulates the flow of the process (E)(E1).

[0033] If errors or a combination of classes are detected in a portion, a robotic arm (26) (D4), equipped with a camera (27), picks up the material and repositions it on the adjustable table or places it in the corresponding container (28).

[0034] Pistons (22) activate the mechanical arm (23), which orients (G) each portion towards the assigned container (25), ensuring an efficient and precise process that integrates mechanical, optical and artificial intelligence technologies.

[0035] Additional presence sensors (24), located on the sorting conveyor (20), ensure that each portion is directed and falls into the correct container (25) (F)(G1). After a few seconds, the robotic arm (23) returns to its initial position (H), ready to process the next portion of fleece (2).

Claims

An automated camelid fiber sorting machine, of the type comprising a feeding zone and a sorting zone, characterized in that: it has an identification belt (5) configured to transport fleece portions (2); a flattening system comprising a metal plate (6), a pressure roller (7) and a transparent glass (8), disposed on the identification belt (5), configured to exert pressure on the fleece portions (2) and generate a uniform horizontal viewing plane through said transparent glass (8); an image capture module comprising at least one industrial camera (9) with a magnifying lens (10) and an LED light source (12), disposed on the transparent glass (8) to capture microscopic images of the flattened fibers;a computer (17) configured with artificial intelligence algorithms to analyze said images and determine the fiber quality according to pre-established parameters; and a sorting belt (20) connected to the output of the identification belt, which has a plurality of containers (25) around it, and on the sides pistons (22) that drive and retract mechanical arms (23) controlled by the computer (17) to direct each portion of fleece to the corresponding container according to its determined quality.; The equipment according to claim 1, characterized in that it further comprises a robotic arm (26) with a camera (27) configured to identify and remove portions of fleece that are incorrectly sorted or have a mixture of qualities for reintroduction into the process. The equipment according to claim 1, characterized in that the industrial camera (9) is physically linked to the metal plate (6) by means of a camera support (14) that maintains a constant focal distance regardless of the vertical movement of the plate caused by the volume of the fleece. The equipment according to claim 1, characterized in that it further comprises, arranged prior to the identification strip (5), an adjustable table (3) whose surface incorporates a mesh (4) configured to allow manual separation of the fleece portions and preliminary filtering of contaminants. The equipment according to claim 4, characterized in that the preliminary filtration of contaminants can be carried out by manual stirring, or by incorporating a mechanical stirring mechanism. A method for the automated classification of camelid fibers using the equipment of claim 1, characterized in that it comprises the following steps: A) Detecting the entry of a portion of fleece (2) into an identification zone by means of a presence sensor (15), activating a capture sequence; B) Flattening and capturing a plurality of microscopic images of the fibers through transparent glass (8) and by means of an industrial camera (9) synchronized with the movement of the transport, ensuring a uniform focal plane; C) Digitally processing the captured images by means of an artificial intelligence model based on convolutional neural networks, configured to automatically extract discriminating visual patterns and features from the fibers, and classifying said fibers into pre-established quality groups based on said learned patterns;D) Determine the quality group of the fleece portion (2) based on the result of the automatic classification obtained by the artificial intelligence model; E) Regulate the process flow by activating and deactivating the sorting belt (20); F) Monitor the physical position of the sorted fleece portion along the sorting belt (20) using additional presence sensors (24); G) Selectively activate a diverter mechanism consisting of a piston (22) and an arm (23) when the fleece portion reaches the position corresponding to its determined category, directing the portion towards a specific container (25); and H) Deactivate the diverter mechanism by closing the arm (23) once the additional presence sensor (24) detects that the fleece portion has passed and been deposited in the specific container (25).

Images