System for characterizing the contents of at least one tank of insect larvae and associated installation and production method

Abstract

System for characterizing the contents of at least one insect larvae tank and associated installation and production method The present invention relates to a system (20) for characterizing the contents (14) of at least one insect larvae tank (12), containing insect larvae and a breeding medium forming the contents (14) of the tank (12). The characterization system (20) comprises: - at least one optical measuring device (22) configured to acquire at least one optical measurement of an upper surface (140) of the contents (14) of the tank (12), and - at least one electronic processing device (24) configured to determine at least one property of the tank (12) representative of a state of the contents (14) of said tank (12) for subsequent industrial processing, the electronic processing device (24) implementing an analysis model capable of determining said property from the acquired optical measurement(s). Figure for abstract: Figure 1

Description

Title of the invention: System for characterizing the contents of at least one tank of insect larvae and associated installation and production method

[0001] The present invention relates to the field of industrial production of insects, in particular insects for food production purposes.

[0002] The present invention relates more particularly to the field of breeding insect larvae, in particular insect larvae in tanks.

[0003] The present invention relates in particular to a system for characterizing the contents of at least one tank of insect larvae, as well as an installation and a method for producing and rearing insect larvae.

[0004] Insects have a number of characteristics that make them well suited for use in animal feed. Insects have a high protein content, while being rich in other beneficial nutrients such as fats, minerals and vitamins. Protein concentration levels in insect meals intended for animal feed vary between 55% and 75%. Insects are characterized by a higher feed conversion rate and can therefore become a very valuable feed source for livestock.

[0005] Furthermore, these products also have a well-balanced nutritional profile to meet human dietary needs.

[0006] These considerations have led to the development of automated mass production of feed from insect farming in industrial sites suitable for the breeding and recovery of mature animals.

[0007] These industrial sites must be optimized to allow industrialization of large volumes of larvae. The costs of building construction, mechanization of operations and installation of storage areas represent a very significant part of the construction budget of industrial sites. In order to reduce these costs, it is essential to increase breeding density, reduce storage time and maximize yields per operation.

[0008] It is known, in order to store large volumes of larvae, to raise them in tanks. Neonate larvae, generally just out of the eggs, are inoculated in tanks in which a feeding substrate has been previously placed which will be consumed at least in part by the larvae during their growth.

[0009] At the end of the larval growth cycle, the tanks then contain mature larvae and a breeding medium, called frass, which comprises a mixture of larval droppings and residues of uneaten substrate.

[0010] One of the critical stages in the industrial production and breeding of larvae in tanks is the recovery of said larvae, and in particular the separation of the larvae from their breeding environment at the end of their growth.

[0011] Thus, poor separation of the larvae from their breeding environment is likely to lead to an interruption of the production chain.

[0012] Several separation methods exist. There are known aeraulic separation methods where the separation is carried out according to a criterion of density or weight / drag force ratio. Said methods are subject to different constraints, in particular the humidity of the material to be separated.

[0013] Another known separation method is sieving. The contents of the tanks are recovered at a sieve-type equipment, the larvae remaining in the sieve while the breeding medium is recovered in another equipment, such as a recovery tank.

[0014] However, sieving is also subject to various constraints. Indeed, the sieve can become clogged depending on the texture and / or the humidity level of the breeding environment, disrupting the separation of the larvae from their breeding environment and potentially leading to an interruption of the chain.

[0015] In order to avoid this, it is known to check the contents of the bins before they are sieved, in order to estimate their sievability.

[0016] By "sievability" is meant here the ability of the contents of a given tank to be sieved. A tank will be said to have good sievability if its contents can be sieved without clogging the sieve used, that is to say without the breeding medium blocking the meshes of the sieve.

[0017] Conventionally, the estimation of the sievability of the contents of the bins is carried out by qualified operators trained for this task. This assessment is typically made on the basis of visual and olfactory observations and on the manual handling of the material contained in the bins. For example, the texture, temperature, color and / or odor of said material are analyzed.

[0018] In case of doubt, additional measurements can be carried out, for example humidity measurements using a TDR (Time Domain Reflectometry) humidity meter. Indeed, the humidity of the breeding medium plays a major role in its sievability, the humidity affecting the texture of the breeding medium and its adhesion properties. Beyond a humidity threshold value, which depends on the equipment and the separation process used, the breeding medium risks disrupting the operation of the separation equipment.

[0019] The conformity of the tanks is also checked by operators. By "conformity" is meant the suitability of the larvae for use in further processing. The conformity of the tanks is, for example, linked to the mortality rate of the larvae. This rate can be estimated using different parameters. For example, the presence of a large amount of uneaten substrate and / or a crust on the surface of the tank indicates a lack of larval activity, implying a high mortality rate.

[0020] A non-compliant tank can be distinguished from a non-sievable tank by the fact that a tank that is non-sievable at the time of characterization may be so later, while a non-compliant tank will remain non-compliant.

[0021] The containers deemed non-compliant are not sieved and follow another industrial cycle.

[0022] However, such control operations do not provide complete satisfaction.

[0023] Indeed, these operations require the immobilization of the tanks on a storage dock accessible by an operator. This implies having specific infrastructure for these operations. The transport of the tanks to these infrastructures and the necessary immobilization time constitute a bottleneck in the larvae production process.

[0024] Operators also need to be trained and have some experience to enable effective characterization of the contents of the bins. In addition, despite the qualifications of the operators, the characterization remains marked by a certain subjectivity, leading to a lack of reproducibility.

[0025] The aim of the present invention is to propose a system and a method for characterizing the contents of insect larvae tanks with a view to their subsequent industrial processing which are faster and more reliable, while being simple to set up in an insect larvae production facility.

[0026] To this end, the invention relates to a system for characterizing the contents of at least one tank of insect larvae, the tank containing insect larvae and a breeding medium forming the contents of the tank, said contents defining an upper surface, the characterization system comprising:

[0027] - at least one optical measuring device configured to acquire at least one optical measurement, e.g. an image, of the upper surface of the contents of the bin, and

[0028] - at least one electronic processing device configured to determine at at least one property of the container, said property being representative of a state of the contents of said container for subsequent industrial processing, the electronic processing device implementing an analysis model capable of determining said property from the optical measurement(s) acquired.

[0029] The characterization system according to the invention allows automated characterization of the contents of insect larvae tanks with a view to their subsequent industrial processing, such as for example their sieving. The characterization system thus allows rapid and reliable control of the tanks. The costs, whether at the level of dedicated infrastructure or operating costs, and in particular the necessary operators, are then reduced.

[0030] According to other advantageous aspects of the invention, the system comprises one or more of the following characteristics, taken individually or in all technically possible combinations:

[0031] - the analysis model implements an artificial intelligence algorithm, preferably a neural network;

[0032] - the optical measuring device comprises a plurality of optical measuring units, each optical measuring unit being configured to acquire at least one optical measurement, for example an image, of the upper surface of the contents of a respective bin of a batch of bins;

[0033] - the electronic processing device is configured to classify the bin among a plurality of distinct bin classes, the electronic processing device implementing an analysis model capable of determining at least one class of the plurality of bin classes from the acquired optical measurement(s), the bin class being representative of a state of the contents of said bin for the purpose of subsequent industrial processing;

[0034] - the electronic processing device is configured to determine a parameter of the contents of the tank, the electronic processing device implementing a regression model capable of determining said parameter from the optical measurement(s) acquired, said parameter being chosen from the time remaining before the contents of the tank can be sieved and the humidity level of the contents of the tank;

[0035] - the system further comprises at least one control device configured to issue a command, for example to a mechanical device for transferring the tray, the command depending on the determined property.

[0036] The invention also relates to an installation for producing insect larvae, the installation comprising:

[0037] - at least one insect larvae tank, the tank containing insect larvae and a breeding medium forming a content of the tank, said content defining an upper surface, and

[0038] - a characterization system of the aforementioned type.

[0039] According to other advantageous aspects of the invention, the installation comprises one or more of the following characteristics, taken individually or in all technically possible combinations:

[0040] - the installation comprises a batch of insect larvae tanks comprising a plurality of insect larvae tanks, the characterization system comprising a plurality of optical measurement units, each optical measurement unit being configured to acquire, preferably simultaneously, at least one optical measurement of the upper surface of the contents of a respective tank of the batch of tanks;

[0041] - the installation comprises a gantry on which the optical device is arranged measurement, the gantry defining a control zone in which the acquisition of optical measurements takes place;

[0042] - the installation further comprises at least one screening device capable of separating insect larvae from the breeding environment contained in the tank.

[0043] The invention also relates to a method for characterizing the contents of at least one tank of insect larvae in an installation of the aforementioned type, the method comprising the following steps:

[0044] - providing at least one bin in a control zone;

[0045] - acquisition of at least one optical measurement, in particular an image, of the surface su upper limit of the contents of said container by the optical measuring device, and

[0046] - determination of at least one property of said container by the electronic device of treatment, said property being representative of a state of the contents of the tank for subsequent industrial treatment.

[0047] According to a particular embodiment, the method further comprises a subsequent step of sorting the container(s) according to the determined property, and optionally a subsequent step of sieving the contents of the container(s) according to the determined property.

[0048] The invention will appear more clearly on reading the description which follows, given solely by way of non-limiting example, and made with reference to the single figure in which:

[0049] [Fig-1] [Fig.l] is a schematic view of a production facility according to the invention, the installation comprising a system for characterizing the contents of at least one tank of insect larvae.

[0050] [Fig.l] schematically shows an installation 10 for producing insect larvae according to the invention.

[0051] The installation 10 is intended for the production and breeding of insect larvae during their growth, from their hatching to the stage of mature larvae.

[0052] The installation 10 is particularly intended for the production of insect larvae in growth modules, also called tanks.

[0053] The installation 10 comprises at least one tank 12 of insect larvae, typically a plurality of tanks 12. Each tank 12 contains insect larvae and a breeding medium forming a content 14 of the tank 12.

[0054] For example, each tray 12 forms an open receptacle having side walls projecting from a bottom. The tray 12 defines an unclosed interior volume, in which the insect larvae and the breeding medium are received. Each tray 12 is intended to receive larvae and a feeding substrate in its receiving volume.

[0055] Advantageously, the larvae injected into a given tank 12 present substantially the same age, that is to say, they hatched at roughly the same time, typically on the same day, preferably at the same time. This allows for good synchronization in the development of the larvae during their growth.

[0056] The substrate typically includes the nutrients necessary for the growth of the larvae. The substrate is intended to be consumed by the larvae during their growth. A portion of the substrate is then transformed into frass, which consists of a mixture of the larval droppings and the residues of the uneaten substrate, dried and optionally fermented.

[0057] As shown in [Fig.l], the contents 14 of the container 12 define an upper surface 140, said surface 140 being visible from the outside of the container.

[0058] Typically, the tanks 12 are distributed in the form of batches of tanks 12 of insect larvae, each batch comprising a plurality of tanks 12 of insect larvae, typically between 5 and 20 tanks.

[0059] Preferably, each tank 12 and / or each batch of tanks 12 is marked by means of an identification system (not shown) chosen from a bar code, a matrix code (commonly referred to as a QR code, for “Quick response code”), a near-field chip (or NFC chip, for “Near Field Communication”), a radio frequency identification chip (or RFID chip, for “Radio Frequency Identification”). The identification system allows logistical traceability throughout the breeding process of the tanks and their contents.

[0060] Advantageously, the trays 12 are suitable for forming a multi-stage module 120, as shown in [Fig.l].

[0061] Said multi-stage module 120 comprises a plurality of tanks 12 arranged one above the other and intended to receive the larvae and the breeding medium.

[0062] For example, said multi-stage module 120 comprises between two and ten trays 12, in particular between four and six trays 12, arranged one above the other.

[0063] A batch of bins comprises, for example, a multi-tiered module 120. According to a particular embodiment, a batch of bins 12 consists of a multi-tiered module 120.

[0064] The installation 10 comprises equipment adapted to the growth of insect larvae, such as heating and ventilation devices (not shown).

[0065] The installation 10 further comprises equipment suitable for recovering larvae after their growth, with a view to subsequent industrial processing.

[0066] By "subsequent industrial treatment" is meant a treatment capable of separating the larvae from their breeding environment at the end of the larval growth.

[0067] For example, the subsequent industrial processing includes a screening step.

[0068] The installation 10 comprises, for example, at least one screening device (schematically represented by the block 19) capable of separating the insect larvae from the breeding environment. The screening device 19 typically comprises a screen having holes whose dimensions are adapted to the larvae of the installation 10, the holes being intended to allow the breeding medium to pass through and to recover the larvae at the end of their growth.

[0069] According to a preferred embodiment, the sieving device 19 comprises means for evaluating the fouling of the sieve, capable of determining a fouling rate of the sieve, and measuring means capable of measuring the humidity rate of the breeding medium received in said sieving device 19. Said means make it possible to determine a single maximum humidity rate of the breeding medium accepted as a function of the desired maximum fouling rate.

[0070] Alternatively or in addition, the subsequent industrial treatment includes a step of measuring the mortality rate of the larvae, separated from their breeding environment, for example for regulatory reasons. Said step is typically followed by the transformation of the larvae into finished products.

[0071] In order to carry out the subsequent industrial processing, the contents of the tanks 12 must have given properties.

[0072] For example, the contents 14 of the tanks 12 must be compliant, that is to say the larvae of said contents must be suitable for use in said subsequent industrial treatment.

[0073] When the subsequent industrial processing includes a sieving step, the contents of the tank 12 must be sievable.

[0074] For example, the contents 14 of a tank 12 are considered to be sievable if the fouling rate after sieving the tank is less than a predefined maximum fouling rate, corresponding to a percentage of blocked sieve meshes. For example, said predefined maximum rate is 5%, preferably 1%, very preferably 0.1%, even more preferably 0.01%.

[0075] Passing a non-sieveable tank through the sieve clogs its mesh and requires maintenance to clean it to make it functional again.

[0076] The installation 10 comprises a characterization system 20 intended to characterize the contents 14 of the tanks 12 of insect larvae, with a view to subsequent industrial treatment.

[0077] The characterization system 20 is capable of determining a given property of the contents of a container 12. Said property is for example: - whether the contents of the given container 12 are compliant; - if the contents of the given bin 12 can be sieved; - a humidity level of the breeding environment; - the presence of surface anomalies, such as the presence of frass balls; - an estimated time before the contents of the bin can be sieved.

[0078] Preferably, the characterization system 20 is capable of determining whether the content 14 of a given 12 bin is compliant and / or sievable.

[0079] The characterization system 20 comprises an optical measuring device 22 configured to acquire at least one optical measurement, for example an image, of the upper surface 140 of the contents 14 of at least one given container 12, and an electronic processing device 24 configured to determine at least one property of said container 12, said property being representative of a state of the contents 14 of said container 12 for subsequent industrial processing. The electronic processing device 24 is configured to implement an analysis model capable of determining said property from the optical measurement(s) acquired by the optical measuring device 22.

[0080] As shown in [Fig. 1], the optical measuring device 22 comprises at least one, preferably a plurality of optical measuring units 220, each optical unit 220 being configured to acquire, preferably simultaneously, at least one optical measurement of the upper surface 140 of the contents 14 of a respective bin 12 of a batch 120 of bins 12.

[0081] The acquisition of optical measurements of the surface 140 of the tanks 12 takes place in a zone 25 called “control zone 25”.

[0082] For example, as shown in [Fig. 1], the optical device 22 comprises a plurality of optical measurement units 220, each optical unit 220 being configured to acquire, preferably simultaneously, at least one optical measurement of the upper surface 140 of the contents 14 of a respective bin 12 of a multi-stage module 120.

[0083] Each optical measuring unit 220 allows optical detection of the surface 140 of the contents 14 of a container 12.

[0084] Said detection is for example carried out in the visible or infrared range.

[0085] The optical measurement is for example a black and white image, a color image or an image acquired with a camera capable of imaging wavelengths up to short-wave infrared (or “SWIR”, for Short-Wave Infrared).

[0086] For example, each optical unit 220 comprises a sensor, which is for example capable of detecting the light transmitted by the surface 140.

[0087] The sensor is preferably coupled to a reflector making it possible to collect the light reflected by the entire surface 140 of the tank or by a significant fraction of the surface 140 of the tank.

[0088] The sensor is for example a camera operating in the visible range and / or in the infrared and capable of forming an image of the detected surface. The resolution of the image obtained makes it possible, for example, to identify the larvae and the breeding environment. The resolution preferably makes it possible to identify the presence of texture defects, such as pellets or frass crusts, the presence of uneaten substrate on the surface, the abnormal presence of molts and / or to identify the color of the surface 140.

[0089] Alternatively, the sensor is a multispectral and hyperspectral camera operating in the visible and / or infrared domain allowing simultaneous acquisition of an image in several more or less narrow frequency bands.

[0090] Alternatively, the sensor is a short-wave infrared photodiode, or SWIR (for “Short Wave Infrared Range”) photodiode, capable of detecting a precise wavelength range, typically 1400 nm, identified as being sensitive to humidity for the medium.

[0091] Alternatively, the sensor is an array of several short-wave infrared photodiodes, each photodiode operating over different wavelength ranges.

[0092] According to a particular embodiment, the sensor is configured to acquire an optical measurement, in particular an image, of the entire surface 140 of the contents 14 of a container 12.

[0093] Alternatively, the sensor is configured to acquire at least one optical measurement, in particular a partial image, of the surface 140 of the contents 14 of a tray 12. This partial optical measurement corresponds to the part of the surface 140 imaged by the optical unit 220 for a relative position. Thus, the optical unit 220 is designed to acquire different partial optical measurements of the surface 140 during a two-dimensional scan of said surface 140 by the optical unit 220.

[0094] By "two-dimensional scanning by the optical unit" is meant the relative displacement of the optical unit 220 with respect to the surface 140 in at least two directions of said surface 140.

[0095] Preferably, the characterization system 20 further comprises a lighting device 26 capable of illuminating the surface 140 of the content 14 to be characterized.

[0096] For example, as shown in [Fig.l], the lighting device 26 comprises a plurality of lighting units 260, each lighting unit 260 being configured to illuminate the surface 140 of a respective tray 12, here of a respective tray 12 of the module 120.

[0097] According to a preferred embodiment, shown in [Fig.l], the installation 10 comprises a structure on which the optical measuring device 22 is mounted.

[0098] For example, said structure is a gantry 28 defining the control zone 25 of the contents of the bins 12 in which the acquisition of optical measurements, in particular images, of the surface 140 of the bins 12 takes place.

[0099] Each optical measuring unit 220 is mounted on the gantry 28.

[0100] When the characterization system 20 comprises a lighting device 26, said device 26 is advantageously mounted on the gantry 28.

[0101] For example, as shown in [Fig.l], the optical measurement units 220 and the lighting units 260 are mounted on a respective support 280 of the gantry 28, the trays 12 being located between said supports during the optical measurement.

[0102] Preferably, the characterization system 20 further comprises a system 30 of detection capable of detecting the presence of the bins 12, the acquisition of the optical measurements being advantageously controlled by the detection of the bins.

[0103] The detection system 30 typically comprises a presence sensor, for example mounted on the gantry 28, configured to detect the presence of one or more bins 12 in the control zone 25.

[0104] Advantageously, the detection system 30 is capable of identifying each bin 12 or batch of bins 12, for example by means of its identification system, and of transmitting said identity to the electronic processing device 24. As a variant, the identification of the bins 12 or batches of bins 12 is carried out by a different device.

[0105] The electronic processing device 24 is capable of receiving signals from the optical measuring device 22, for example by wired transmission means 32, and of processing them in order to determine at least one property representative of the contents 14 of a tank 12 with a view to subsequent industrial processing.

[0106] For example, the electronic processing device 24 comprises a memory 40 and a microprocessor 42. The electronic processing device 24 is capable of determining the property(ies) representative of the contents 14 of the container 12 on the basis of the optical measurement(s), in particular images, acquired by the optical measuring device 22.

[0107] The microprocessor 42 is configured to implement the analysis model capable of determining the property(ies) from the acquired optical measurement(s).

[0108] For example, the analysis model implements an artificial intelligence algorithm.

[0109] For example, an input variable of the artificial intelligence algorithm is the optical measurement, for example the image or a part of the image, taken of the contents 14 of a tray 12, and at least one output variable of the artificial intelligence algorithm is an indication relating to a property of said contents 14.

[0110] Advantageously, the algorithm is an artificial neural network.

[0111] The artificial neural network used by the electronic processing device 24 may comprise a system trained on the basis of optical measurements, in particular images, or samples which have been previously classified by a human expert.

[0112] According to a particular embodiment, the neural network is a convolutional neural network, also sometimes called a convolutional neural network or by the acronym CNN which refers to the English term for “Convolutional Neural Networks”. In a convolutional neural network, each neuron of the same layer has exactly the same connection pattern as its neighboring neurons, but at different input positions. The connection pattern is called a convolution kernel or, more often, “kernel” in reference to the English term cor- respondent.

[0113] Advantageously, the artificial neural network has been previously trained using training data. The training data may comprise optical measurements, in particular images, of bin contents and the associated properties which have been previously assigned by an expert.

[0114] For example, the training data may include optical measurements, including images, classified as corresponding to compliant bins and images classified as corresponding to non-compliant bins.

[0115] For example, the training data may comprise images classified following visual observations made by operators and / or following measurements made by operators, such as measurements of the humidity level of the breeding environment, said level having been measured using a TDR humidity meter.

[0116] Once the neural network has been trained, it can for example be used to identify or predict a property of the contents 14 of a bin 12 represented in an image. For example, an uncharacterized image is provided to the network and the network determines a property.

[0117] For example, the network is capable of estimating the humidity level of the contents 14 based on its color and possibly determining whether the contents 14 can be sieved or not.

[0118] The memory 40 of the electronic processing device 24 is capable of storing determination software configured to determine, via the analysis model, in the optical measurement, in particular the image, taken by the optical measurement device 22, at least one property of the content 14 to be characterized. The microprocessor is then capable of executing the determination software.

[0119] According to a preferred embodiment, the analysis model is a classification model.

[0120] The electronic processing device 24 is configured to classify the bins 12 among a plurality of distinct bin classes. The electronic processing device 24 implements an analysis model capable of determining at least one class of the plurality of bin classes from the optical measurement(s), in particular the images, acquired, the class of the bin 12 being representative of a state of the contents 16 of said bin 12 for the purpose of subsequent industrial processing.

[0121] The electronic processing device 24 is capable of deciding on the classification of each processed bin 12 on the basis of the optical measurements obtained by the optical measuring device 22.

[0122] The classification criteria are for example recorded in the memory 40. The microprocessor 42 is capable of comparing the optical measurements with the parameters of the classification criteria.

[0123] Advantageously, the microprocessor is configured to implement an artificial intelligence algorithm in order to classify the bins 12.

[0124] For example, an input variable of the artificial intelligence algorithm is the optical measurement, in particular the image taken or a part of the image taken, and at least one output variable of the artificial intelligence algorithm is an indication relating to a class of the bac 12.

[0125] The class of the tank 12 is representative of a state of the contents of said tank 12 for subsequent industrial processing.

[0126] For example, the electronic processing device 24 is configured to classify the bins 12 between a first class corresponding to the bins whose contents 14 are considered to be compliant (compliant bins) and a second class corresponding to the bins whose contents 14 are considered to be non-compliant (non-compliant bins). The electronic processing device 24 is thus capable of classifying each bin 12 as a compliant bin or a non-compliant bin.

[0127] For example, the electronic processing device 24 is configured to classify the bins 12 between a first class corresponding to the bins whose contents 14 are judged to be sievable (sievable bins) and a second class corresponding to the bins whose contents 14 are judged to be not sievable (non-sievable bins). The electronic processing device 24 is thus capable of classifying each bin 12 as a sievable bin or a non-sievable bin.

[0128] According to a preferred embodiment, the electronic processing device 24 is configured to classify the bins 12 between a first class corresponding to the bins whose contents 14 are judged to be sievable (sievable bins) and a plurality of additional classes, each additional class corresponding to the bins 12 whose contents 14 are judged to be not yet sievable but whose contents 14 will be sievable in N days, N being a natural integer typically between 1 (inclusive) and 7 (inclusive).

[0129] For example, the electronic processing device 24 is configured to classify the bins 12 between a first class corresponding to the bins whose contents 14 are judged to be sievable (sievable bins), a second class corresponding to the bins whose contents 14 are judged to be not yet sievable but whose contents 14 will be sievable in 1 day and a third class corresponding to the bins whose contents 14 are judged to be not yet sievable but whose contents 14 will be sievable in 2 days.

[0130] According to a very preferred embodiment, the electronic processing device 24 is configured to classify the bins 12 between a first class corresponding to the bins whose contents 14 are judged to be sievable (sievable bins) and a plurality of additional classes, each additional class corresponding to the bins 12. corresponding to the bins 12 whose contents 14 are judged not to be yet sievable but whose contents 14 will be sievable in N hours, N being a natural integer. Such a device 24 thus allows an even more precise characterization of the bins.

[0131] Advantageously, the artificial intelligence algorithm is an artificial neural network, for example a convolutional neural network.

[0132] The memory 40 is capable of storing determination software configured to determine, via the artificial neural network, in the optical measurement, in particular the image, taken by the optical device 22, the class of each bin 12 represented in the image. The microprocessor 42 is then capable of executing the determination software.

[0133] For example, the electronic processing device 24 is configured to determine the class of each bin 12 represented in the image, an input variable of the artificial neural network being the image or a part of the image taken by the optical measuring device 22, and at least one output variable of the neural network being an indication relating to the class of the bin 12.

[0134] Advantageously, the artificial neural network has been previously trained using training data. The training data may comprise optical measurements, in particular images, of contents 14 of bins 12 and their designated classes which have been previously assigned by an expert. For example, the training data may comprise images classified as corresponding to compliant bins and portions of images classified as corresponding to non-compliant bins.

[0135] Once the neural network has been trained, it can for example be used to identify or predict the class of a bin 12 whose contents 14 are represented in an image. For example, an unclassified image is provided to the network and the network produces a classification.

[0136] Alternatively or additionally, the analysis model is a regression model. The electronic processing device 24 is configured to determine a parameter of the contents 14 of the tank 12, said parameter being chosen from the time remaining before the contents 14 of the tank 12 can be sieved, the mortality rate of the larvae or the humidity of the contents 14 of the tank 12.

[0137] The regression model is for example suitable for determining, from the optical measurement(s) received, in how many days or hours the contents 14 of a tank 12 will be sievable.

[0138] Preferably, the electronic processing device 24 is capable of simultaneously determining several properties of a given container 12. For example, the electronic processing device 24 is capable of determining whether the contents 16 of the container 12 are siftable and compliant.

[0139] Preferably, the electronic processing device 24 further comprises a communication module (not shown), capable of communicating with a database to record the determined properties there and associate them with the respective bins 12.

[0140] Preferably, as shown in [Fig.l], the characterization system 20 further comprises at least one control device 50, capable of communicating with the electronic processing device 24.

[0141] The control device 50 is capable of sorting the bins 12 according to the determined property, optionally the determined class.

[0142] Thus, depending on the property determined by the electronic processing device 24, the tank 12 is then subjected to different operations.

[0143] The control device 50 is for example configured to issue a command. The command depends on the determined property, optionally on the determined class.

[0144] For example, if a tank 12 has been characterized as being compliant and sievable, a sieving instruction is issued and the tank 12 is then transferred to the sieving device 19.

[0145] If a tank 12 has been characterized as being compliant and not sievable, a transfer instruction is issued and said tank 12 is transferred to a waiting area, before a new subsequent characterization.

[0146] If a container 12 has been characterized as being non-compliant, a rejection instruction is issued and said container 12 is transferred to a rejection device 70 (schematically represented by block 70).

[0147] Preferably, as shown in [Fig.l], the installation 10 comprises a mechanical transfer device 60 capable of moving the bins 12 inside the installation 10.

[0148] The mechanical transfer device 60 comprises, for example, conveying means, such as conveyor belts.

[0149] The mechanical transfer device 60 is notably configured to bring the bins 12 to the level of the control zone 25.

[0150] For example, in the case of [Fig.l], the mechanical transfer device 60 is configured to bring the bins 12 into the control zone under the gantry 28.

[0151] Advantageously, the mechanical transfer device 60 is also configured to transport the tanks 12 after their characterization, for example towards the screening device.

[0152] A method for characterizing the contents 14 of at least one tank 12 of insect larvae according to the invention will now be described.

[0153] Said method is intended to be implemented in an installation 10 as described previously.

[0154] In order to check the status of the contents 14 of a tray 12 for further processing, at least one tray 12 is provided.

[0155] Said tank 12 contains insect larvae and a breeding medium forming the contents 14 of the tank. Preferably, the tank 12 is provided at the end of a larval growth phase.

[0156] At the end of the growth phase, the tank 12 is transported to the control zone 25, for example by means of the transfer device 60.

[0157] For example, as shown in [Fig.l], the tray 12, here the multi-stage module 120 comprising said tray 12, is conveyed to the gantry 28.

[0158] During a step of acquiring optical measurements, for example images, at least one optical measurement of the upper surface 140 of the contents 14 of said container 12 is acquired by the optical measuring device 22, in particular by an optical measuring unit 220.

[0159] In the example of [Fig.l], an optical measurement of the upper surface 140 of the contents 14 of each bin 12 of the module 120 is acquired by a respective optical measurement unit 220.

[0160] The image(s) are then transmitted to the electronic processing device 24.

[0161] Said electronic processing device 24 determines at least one property of the or each tank 12, said property being representative of a state of the contents 14 of the tank 12 given for the purpose of subsequent industrial processing.

[0162] Advantageously, the determination is carried out by means of an artificial intelligence algorithm, preferably by an artificial neural network. An input variable of the artificial intelligence algorithm is an optical measurement, in particular an image, of the contents 14 of a given tray 12, and at least one output variable of the artificial intelligence algorithm is an indication relating to a property of said contents 14.

[0163] The method preferably comprises a preliminary step of training the model. During this preliminary step, an expert annotates a data set comprising optical measurements, in particular images, of bin contents and their associated properties, then the model is trained on this data set.

[0164] For example, the determination step comprises a step of classifying the or each bin 12. During the classification step, the optical measurement, for example the image, is analyzed and compared to the classification criteria. The electronic processing device 24 determines to which class each bin 12 belongs.

[0165] The classification is automated. Advantageously, the classification is carried out at by means of an artificial intelligence algorithm, preferably by an artificial neural network. An input variable of the artificial intelligence model is an optical measurement of the content 14 of a given bin 12, and at least one output variable of the artificial intelligence model is an indication relating to a class of the bin 12.

[0166] The method preferably comprises a preliminary step of training the model. During this preliminary step, an expert annotates a data set comprising optical measurements, in particular images, of bin contents and their designated classes, then the model is trained on this data set.

[0167] Preferably, the method comprises a subsequent step of sorting the bin(s) 12 according to the determined property(ies), for example the determined class.

[0168] For example, when the method comprises a subsequent sieving step, the determined property is for example representative of the sievability and / or the conformity of the contents 14 of the tank(s) 12.

[0169] According to an advantageous embodiment, the steps of acquiring images and determining at least one property of said tank 12 are repeated at least once, increasing the reliability of the characterization.

[0170] The characterization system and method according to the invention allow rapid and reliable control of the contents of tanks containing insect larvae with a view to their subsequent industrial treatment.

[0171] The automated detection of tanks 12 deviating from the predetermined conditions is more reliable and takes less time than if it were carried out manually by operators. The use of an electronic processing device 24 also makes it possible to improve the reproducibility of the characterization of the tanks.

[0172] The tanks 12 deviating from the predetermined conditions can thus be identified and removed automatically and quickly, avoiding the formation of a bottleneck in the larvae production process. The characterization system thus allows efficient and rapid sorting of the tanks 12 according to their contents 14, improving the throughput of the process.

[0173] Furthermore, the system and the characterization method according to the invention allow a reduction in costs, both in terms of infrastructure and operating costs, and in particular the necessary operators, making it possible to produce competitive products.

[0174] The use of an artificial intelligence algorithm, such as a neural network, is particularly reliable in avoiding characterization errors (false positives or false negatives). Indeed, the model used can be easily optimized by adapting the training set to situations of false positives or false negatives. The model can, for example, be continuously optimized by evaluating the fouling rate of the sieve depending on the humidity level of the breeding environment.

Claims

Claims

1. System (20) for characterizing a content (14) of at least one tank (12) of insect larvae, the tank (12) containing insect larvae and a breeding medium forming the content (14) of the tank (12), said content (14) defining an upper surface (140), the characterization system (20) comprising: • at least one optical measuring device (22) configured to acquire at least one optical measurement, for example an image, of the upper surface (140) of the content (14) of the tank (12), and • at least one electronic processing device (24) configured to determine at least one property of the tank (12), said property being representative of a state of the content (14) of said tank (12) for subsequent industrial processing, the electronic processing device (24) implementing an analysis model capable of determining said property from the acquired optical measurement(s).

2. The system (20) of claim 1, wherein the analysis model implements an artificial intelligence algorithm, preferably a neural network.

3. A system (20) according to any preceding claim, wherein the optical measuring device (22) comprises a plurality of optical measuring units (220), each optical measuring unit (220) being configured to acquire at least one optical measurement, for example an image, of the upper surface (140) of the contents (14) of a respective bin (12) of a batch of bins (12).

4. System (20) according to any one of the preceding claims, in which the electronic processing device (24) is configured to classify the bin (12) among a plurality of distinct classes of bins (12), the electronic processing device (24) implementing an analysis model capable of determining at least one class of the plurality of bin classes from the acquired optical measurement(s), the class of the bin (12) being representative of a state of the contents (14) of said bin (12) for subsequent industrial processing.

5. System (20) according to any one of the preceding claims, in which the electronic processing device (24) is configured to determine a parameter of the contents (14) of the tank (12), the electronic processing device (24) implementing a regression model capable of determining said parameter from the acquired optical measurement(s), said parameter being chosen from the time remaining before the contents (14) of the tank (12) can be sieved and the humidity level of the contents (14) of the tank (12).

6. System (20) according to any one of the preceding claims, further comprising at least one control device (50) configured to issue a command, for example to a mechanical device (60) for transferring the tray (12), the command depending on the determined property.

7. Installation (10) for producing insect larvae, the installation (10) comprising: • at least one tank (12) of insect larvae, the tank (12) containing insect larvae and a breeding medium forming a content (14) of the tank (12), said content (14) defining an upper surface (140), and • a characterization system (20) according to any one of the preceding claims.

8. Installation (10) according to claim 7, the installation (10) comprising a batch of insect larvae tanks (12) comprising a plurality of insect larvae tanks (12), the characterization system (20) comprising a plurality of optical measurement units (220), each optical measurement unit (220) being configured to acquire, preferably simultaneously, at least one optical measurement of the upper surface (140) of the contents (14) of a respective tank (12) of the batch of tanks (12).

9. Installation (10) according to claim 7 or 8, comprising a gantry (28) on which the optical measuring device (22) is arranged, the gantry (28) defining a control zone (25) in which the acquisition of optical measurements takes place.

10. Installation (10) according to any one of claims 7 to 9, further comprising at least one sieving device (19) capable of separating the insect larvae from the breeding medium contained in the tank (12).

11. Method for characterizing a content (14) of at least one container (12) of

12. insect larvae in an installation (10) according to any one of claims 7 to 10, the method comprising the following steps: • provision of at least one bin (12) in a control zone (25); • acquisition of at least one optical measurement, in particular an image, of the upper surface (140) of the contents (14) of said container (12) by the optical measuring device (22), and • determination of at least one property of said container (12) by the electronic processing device (24), said property being representative of a state of the contents (14) of the container (12) with a view to subsequent industrial processing. Method according to claim 11, further comprising a subsequent step of sorting the container(s) (12) according to the determined property, and optionally a subsequent step of sieving the contents (14) of the container(s) (12) according to the determined property.

Images