EP-4737020-A1 - COMPUTER-IMPLEMENTED SYSTEM AND METHOD FOR SORTING BANANAS

EP4737020A1EP 4737020 A1EP4737020 A1EP 4737020A1EP-4737020-A1

Abstract

The present invention provides a banana sorting system comprising two conveyor belts (1) separated by a gap (2), whereby a banana group (3) is transferred from one belt (1) to another. It includes imaging means (4) that capture the group from different angles, including the gap (2), allowing a series of images to be obtained during the passage thereof. These means are also designed to reconstruct a single image of the hidden areas of the banana group (3) that are not visible when said banana group (3) is resting on the belts (1). The system has a control system (6) that manages both the belts (1) and the imaging means (4), and determines the quality of the banana from the reconstruction of images and the different viewpoints. Based on this evaluation, the system sorts the banana group (3) according to the assigned quality category, allowing an accurate and automated sorting of the bananas.

Inventors

Carreira Lorenzo, Adrián
Lorenzo Dávila, Alejandro
Fuertes Rodríguez, Miguel Angel
Moreno Pérez, Javier

Assignees

ESAE Vision System S.L.

Dates

Publication Date: 20260506
Application Date: 20241030

Claims (11)

A banana sorting system comprising: - two conveyor belts separated from one another in a conveying direction with a gap between them, the conveyor belts being configured for conveying at least one banana group which is transferred from one belt to the other; - imaging means configured to capture images of the group as it moves along the conveyor belts from different viewpoints including the viewpoint from the gap, such that said imaging means are further configured to capture a series of images of the group as it passes through the gap as seen from said gap; and - control means configured to: - control the conveyor belts and the imaging means; - obtain an image of a hidden part of the group resting against the conveyor belts from reconstructing into a single image the series of images of the group as it passes through the gap as seen from said gap; - determine the category of the group according to the quality of the group established from the images captured by the imaging means from different angles and the reconstructed single image; and - sort the group based on the determined category thereof.
The banana sorting system according to claim 1, wherein, in order to determine the quality of the group from the images captured by the imaging means from different angles and the reconstructed single image, the control means are configured to detect the presence of damage per unit area on the peel of the bananas in the group and categorise the group according to a quantification of said damage, wherein such damage comprises rubbing and/or damage caused by pests and/or the presence of latex.
The banana sorting system according to claim 2, wherein, in order to detect the presence of damage on the peel of the bananas in the group and to categorise the group according to a quantification of said damage, the control means are configured to quantify the damage from: - determining the number of fingers in the group; - determining which fingers are damaged and the relative positions of the damage on each finger; and - determining the amount of damage per unit area of the peel; wherein the category of the group is determined from comparing the quantified damage with pre-established target values.
The banana sorting system according to any of the preceding claims, wherein, from the images captured by the imaging means from different angles and the reconstructed single image, the control means are further configured to: - identifying a crown in the group; and/or - determine the level of ripeness based on the colour of the fingers in the group; and/or - generate a 3D model of the group.
The banana sorting system according to any of the preceding claims, wherein the control means implement an artificial intelligence configured to determine the category of the group according to the quality of the group established from the images captured by the imaging means from different angles and the reconstructed single image, detecting the presence of damage on the peel of the fingers in the group, quantifying it and categorising the group according to the quantification of the damage.
The banana sorting system according to any of the preceding claims, comprising weighing means configured to weigh the group, such that sorting is performed based on the category and weight of the group.
The banana sorting system according to any of the preceding claims, comprising a tunnel through which the conveyor belts run at least partially and in which the imaging means are arranged.
A computer-implemented method for sorting bananas by means of a system for sorting bananas, the system comprising: - two conveyor belts separated from one another in a conveying direction with a gap between them, the conveyor belts being configured for conveying at least one banana group which is transferred from one belt to the other; - imaging means configured to capture images of the group as it moves along the conveyor belts from different viewpoints including the viewpoint from the gap, such that said imaging means are further configured to capture a series of images of the group as it passes through the gap as seen from said gap; and - control means configured to control the conveyor belts and the imaging means, and to obtain an image of a hidden part of the group resting against the conveyor belts from reconstructing into a single image the series of images of the group captured as it passes through the gap as seen from said gap, wherein the control means comprise an artificial intelligence model configured to determine the category of the group according to the quality of the group established from the images captured by the imaging means from different angles and the reconstructed single image; wherein the method comprises, through the control means, the steps of - feeding the artificial intelligence with the images captured by the imaging means from different angles and the reconstructed single image; - determining by artificial intelligence the category of the group according to the quality of the group established from the images captured by the imaging means from different angles and the reconstructed single image; and - obtaining a sorting of the group by artificial intelligence based on the determined category thereof.
The method according to claim 8, comprising, in the step of determining the quality of the group from the images captured by the imaging means from different angles and the reconstructed single image, the step of detecting, by artificial intelligence, the presence of damage per unit area on the peel of the banana in the group and categorising the group according to a quantification of said damage, wherein such damage comprises rubbing and/or damage caused by pests and/or the presence of latex.
The method according to claim 9, wherein the step of detecting, by artificial intelligence, the presence of damage on the peel of the bananas in the banana group and categorizing the group according to a quantification of said damage comprise the step of, by artificial intelligence, quantifying the damage from: - determining the number of fingers in the banana group; - determining which fingers are damaged and the relative positions of the damage on each finger; and - determining the amount of damage per unit area of the peel; wherein the category of the group is determined from comparing the quantified damage with pre-established target values.
The method according to any of claims 8 to 10, wherein in the step from the images captured by the imaging means from different angles and the reconstructed single image, it comprises, by artificial intelligence, the steps of: - identifying a crown in the group; and/or - determining the level of ripeness based on the colour of the fingers in the group.

Description

TECHNICAL FIELD OF THE INVENTION The present invention relates to a system for categorising and sorting bananas. The invention also relates to a computer-implemented method for sorting bananas. STATE OF THE ART One of the most delicate and relevant processes in the value and marketing chain of bananas, particularly Canary Island bananas, is the quality-based sorting of bananas. In this sense, Canary Island bananas are recognised for their quality and their presence on the foreign market depends on this. Quality standards are regulated by the regulations dictated by BOE-A-1987-26231 Order of 23 November 1987 for the Spanish market and by the International Standards for Fruit and Vegetables for bananas (OECD, 2023) for international trade. According to these regulations, the quality sorting of bananas is mainly based on the presence of damage on the peel, such as rubbing, damage caused by pests, the presence of latex, etc., parameterised per damaged unit area (cm2/banana) and, at present, the detection of the presence of damage is performed by the same methodology used for many years, that is, by means of visual inspection by qualified operators. This rudimentary sorting-detection technique requires a lot of personnel, making it the highest cost of banana packaging and the second highest cost of the entire value chain of the product, as reflected in the "Estudio de la cadena de valor y formatión de los precios del plátano" (Study of the banana value chain and price formation) (Spanish Ministry of Agriculture, Fish and Food, 2009). Accordingly, for a packaging company to process 30 million kilos of bananas, around 100 people are required for this process alone, achieving a yield of 133 groups/min as efficiently as possible, where at least 5 packing lines are used to achieve this throughput. This leads to a bottleneck in the packaging and marketing process, not being able to deliver the goods requested by the market on time or not being able to process the required production volumes due to the intense harvesting season. As it is a completely manual process, subject only to the objective criteria of an operator, a high percentage of errors are made in the sorting that harm the farmer in the remuneration of the fruit and cause great havoc for the packing company due to the return of lots that do not meet the minimum required and/or legal standards. Therefore, there is a clear need to improve the banana sorting process, making it more efficient while at the same time less dependent on human intervention. DESCRIPTION In order to respond to the detected need, the present invention provides a banana sorting system, hereinafter system, as specified in claim 1. Particular embodiments of the system are set forth in the claims that are dependent on said independent claim 1. Within the scope of the invention, a bunch of bananas is the name given to the set consisting of hands, which are in turn made up of a plurality of fingers (bananas), said hands hanging from a stem, the bunch being made up of a plurality of hands attached to the stem. However, given that a hand can be whole (with all its fingers) or fragmented, referred to as a cluster, in the banana sorting system, sorting will be carried out by processing one group, which is either a hand or a cluster, interchangeably. According to the foregoing, in a first inventive aspect, the banana sorting system comprises: two conveyor belts separated from one another in a conveying direction with a gap between them, the conveyor belts being configured for conveying at least one group consisting of bananas which is transferred from one belt to the other;imaging means configured to capture images of the group as it moves along the conveyor belts from different viewpoints including the viewpoint from the gap, such that said imaging means are further configured to capture a series of images of the group as it passes through the gap as seen from said gap; andcontrol means configured to:control the conveyor belts and the imaging means;obtain an image of a hidden part of the banana group resting against the conveyor belts from reconstructing into a single image the series of images of the banana group as it passes through the gap as seen from said gap;determine the category of the banana group according to the quality of the banana group established from the images captured by the imaging means from different angles and the reconstructed single image;andsort the group based on the determined category thereof. Firstly, it is important to note that, unlike other fruits, an individual banana and the group have an irregular shape and cavities, which prevents them from turning or rotating in a machine in order to evaluate their condition over their entire outer surface. In addition to this fact, due to the delicate nature of this fruit, unnecessary handling of the banana or group may cause additional involvement thereof, which will therefore affect its categorisation and sorting. Due to