Method of assessing status of a storage organ

Abstract

There is provided a computer implemented method of determining a condition indicative of sprouting of a storage organ, the method comprising, receiving first data indicative of an electrical signal from the storage organ, processing the first data to determine a characteristic of the electrical signal and determining a condition indicative of sprouting of the storage organ based on the characteristic.

Description

[0001] Method of assessing status of a storage organ

[0002] Technical Field

[0003] The present disclosure relates to a method and apparatus for determining a condition indicative of sprouting of a storage organ, such as a root vegetable.

[0004] Background

[0005] Storage organs are a part of a plant used to store energy and / or water and allow the plant to survive during adverse periods of the plant’s life, such as during times of drought. Typically, storage organs grow beneath the surface of the plant’s growth medium (e.g. grow beneath the ground in the soil). Plants that have an underground storage organ are called geophytes. Many storage organs are suitable for human and animal food consumption. For example, storage organs such as potatoes, onions and carrots are cultivated and harvested primarily for human and animal consumption. Such storage organs, along with other underground storage organs which are primarily grown and consumed, are generally referred to as root vegetables.

[0006] Storage organs are usually classified as either a root or a modified stem, although other types of classification are understood. These two classifications of storage organs are further sub-categorised based upon their botanical characteristics. For example, a potato is considered a stem tuber which is a type of modified stem, a carrot is considered a storage taproot which is a type of root, and an onion is considered a bulb which is another type of modified stem.

[0007] Storage organs undergo a process known as sprouting, whereby the storage organ sprouts new shoots from the surface of the storage organ and which further develop into structures of a new plant. The growth of these new structures may include growth of new storage organs. Sprouting is a natural part of the lifecycle of a storage organ, and the timing of sprouting is influenced by many factors, including environmental factors such as temperature.

[0008] During sprouting, storage organs undergo metabolic processes which alter the storage organ’s chemical and / or structural composition. For example, sprouted potatoes contain higher levels of glycoalkaloids than non-sprouted potatoes. Glycoalkaloids are toxic to humans in high doses. Other storage organs may experience different changes in composition which, while not causing any harmful effects to humans if eaten, cause undesirable changes in the storage organ’s consistency and taste. Given that root vegetables and the like are typically harvested over a narrow time period and stored for relatively long time periods, it is desirable to prevent or delay storage organs from undergoing their natural sprouting process.

[0009] For example, a potato crop tends to be harvested during a narrow period over the course of a year. These harvested potatoes must then be stored so as to be used throughout the year until the next harvest. Additionally, processing capacity for processing the potatoes is limited in that it may not be possible to process all potatoes immediately after harvest. Such processing may comprise preparing potato fries, crisps, and chips, for example. Potatoes may therefore need to be stored for extended periods of time, such as up to 10 months before processing. If, during this period, the potatoes start their natural process of sprouting then their viability to be used in further food production is diminished due to spoiling of the potatoes.

[0010] Until relatively recently the problem of inhibiting sprouting was typically managed using chlorpropham (CIPC), which is a growth inhibitor. However, the use of CIPC is not now approved for such use in many countries, such as in the EU and UK. Other alternative chemical substances are available, but their effectiveness can be low and their cost can be high.

[0011] Thus, there is a need to provide an improved method and apparatus for determining a condition indicative of sprouting of a storage organ.

[0012] In a first aspect, there is provided a computer implemented method of determining a condition indicative of sprouting of a storage organ, the method comprising, receiving first data indicative of an electrical signal from the storage organ, processing the first data to determine a characteristic of the electrical signal and determining a condition indicative of sprouting of the storage organ based on the characteristic. Advantageously, by determining the condition of the storage organ, such as a root vegetable, based on the received electrical signal, suitable action may be taken prior to any adverse effects occurring in the storage organ. For example, in the case of potatoes, the presence of a burst of spikes in the received electrical signal is an indication that the potato is beginning the sprouting process, and will shortly develop shoots which extend from the surface of the potato. That is, electrophysiology is used to predict sprouting in the potato, or other storage organ.

[0013] Storage organs, such as potatoes, once harvested are typically stored in warehouses prior to sorting and onward production. Potatoes are usually stored in temperature controlled environments, such as at 4 or 8 degrees C. These temperature controlled environments slow down the development of the potatoes along their lifecycle, allowing them to be stored for longer periods of time without spoiling (such as when they begin to sprout). For example, a potato that is showing visible signs of sprouting (e.g. shoots extending from the potato) cannot usually be used in food products and so usually leads to waste. Therefore, early identification of sprouting can have significant benefit in reducing loss of storage organs. For example, if it is determined that a potato is about to sprout, action may be taken such as applying chemical treatments to the store of potatoes that inhibit sprouting. Therefore, to the extent that chemical treatment must be used, its use is reduced as it is only used when required.

[0014] The condition indicative of sprouting may be considered to be indicative of a temporal position in the lifecycle of the storage organ. For example, the temporal position in the lifecycle of the storage organ may be a position prior to visible signs of sprouting appearing on the root vegetable. The position may be 3-41 days prior to sprouting.

[0015] The electrical signal from the storage organ may be obtained directly from the storage organ. The electrical signal may be generated by the storage organ, and obtained directly from the storage organ using electrodes coupled to the storage organ. That is, the electrical signal is indicative of bioelectrical activity generated within the storage organ as it undergoes sprouting.

[0016] The storage organ may be a sample selected from a crop. That is, the storage organ may belong to a first set, with the remaining storage organs not selected for the sample belonging to a second set. Determining a condition indicative of sprouting of the storage organ based on the characteristic may be implicit. That is, if it is known that a particular characteristic of the electrical signal indicates sprouting, then by determining the characteristic, it is determined that there is a condition indicative of sprouting of the storage organ.

[0017] The steps of receiving first data, processing first data, and determining a condition may be carried out at a processing apparatus comprising a processor. Each step may of course be carried out by the same processing apparatus, or by different processing apparatuses.

[0018] The characteristic of the electrical signal may comprise an event in the electrical signal. The event may be a change from a baseline electrical signal.

[0019] The event in the electrical signal may be a spike in the electrical signal.

[0020] For example, where the electrical signal is indicative of an electric potential difference, an event may be a spike in the electric potential difference, e.g. a rapid change in the value of the electric potential difference from a baseline. For example, a spike may be characterised by a sharp deflection in the voltage, followed by a slower return to the baseline level. The spike in the electrical signal may be indicative of action potential within the storage organ.

[0021] The characteristic of the electrical signal may comprises a burst of events in the electrical signal.

[0022] The burst of events may comprise a predetermined number of events within a predetermined time period.

[0023] A predetermined number of events (e.g. spikes) within a predetermined time period may be considered to be a burst of events (e.g. spikes) in the electrical signal. That is, the characteristic of the electrical signal may comprise a burst of events (or spikes) in the electrical signal. A burst may be defined differently for given species of storage organ, but as an example, a burst of spikes in a typical potato may be defined as there being on average more than 2 spikes per hour over a 24 hour period. A burst of events indicates relatively high levels of activity occurring within the storage organ (such as high rates of cell division and biochemical activity).

[0024] The event may indicate an action potential event generated within the storage organ.

[0025] That is, the characteristic of the electrical signal may be caused by an action potential event occurring within the storage organ.

[0026] The electrical signal may be indicative of an electric potential difference.

[0027] For example, the electrical signal may be a measured electric potential difference (voltage) generated by the storage organ. The characteristic of the electrical signal may then be a characteristic of the measured electric potential difference.

[0028] The electric potential difference may be measured between a first electrode coupled to a first portion of the storage organ and a second electrode coupled to a second portion of the storage organ.

[0029] The first portion may be at or near a surface of the storage organ.

[0030] The first electrode being coupled at or near a surface of the storage organ may comprise the first electrode being inserted a relatively short distance into the surface of the potato. For example, the first electrode may be inserted at or about 5 mm or less into the surface of the storage organ.

[0031] The second portion may be at or near a centre portion of the storage organ.

[0032] The second electrode being coupled at or near the centre of the storage organ may comprise the second electrode being inserted relatively deep into the surface of the storage organ such that the electrode is located generally within the centre of the storage organ, the centre being a part of the storage organ that is furthest from the surface.

[0033] The storage organ may belong to a first set of storage organs, and the method may further comprise determining, based on the determination of the condition indicative of sprouting of the storage organ, a condition indicative of sprouting of storage organs belonging to a second set of storage organs, wherein the second set of storage organs are associated with the first set of storage organs.

[0034] For example, the first mentioned storage organ may be selected as a sample set (first set) to be monitored, the selecting being from a total crop of harvested potatoes, with the remaining crop stored in the storage facility being the second set. In this way, the first mentioned storage organ is associated with the storage organs of the second set in that they belong to the same crop. As such, determining that the storage organ in the first set is about to sprout (e.g. signs of shoots emerging from the storage organ will soon be visible), implicitly means determining that the related storage organs in the second set are also about to sprout. That is, electrophysiology is used to predict sprouting in the stored potatoes of the second set.

[0035] Alternatively or additionally, the association between the first and second set may be that both sets may be stored in similar environmental conditions. Alternatively or additionally, the association between the first and second set may be that both sets may comprises storage organs of the same type (e.g. same species and variety) and similar age.

[0036] The method may further comprise outputting second data indicating the condition of the storage organ.

[0037] That is, second data may be generated which is configured to be output to the user, or to further apparatus to automatically control said apparatus.

[0038] The outputting the second data indicating the condition of the storage organ may comprise outputting a warning.

[0039] For example, the warning may be output using visual and / or audio data. The warning may be an alarm. The warning may be sent to an output device, such as a user device. The warning may take any suitable form to alert the user to the condition of the storage organ. For example, a message may be sent to a user device which states that sprouting of a harvest of storage organs is imminent. Outputting second data indicating the condition of the storage organ may comprise outputting a control signal to an environmental control apparatus, the environmental control apparatus configured to automatically change an environment in which the second set of storage organs are stored.

[0040] For example, on receipt of the control signal, the environmental control apparatus may actuate a pump or other dispenser which delivers a chemical substance to the storage organs stored in a storage facility. The chemical substance may be a substance that inhibits sprouting. The control signal may take any suitable form to cause the environmental control apparatus to automatically change the environment.

[0041] The first data may be indicative of an electrical signal from each of a plurality of storage organs and processing the first data may comprise processing the first data to determine the characteristic of at least one of the electrical signals.

[0042] For example, a number of storage organs may be monitored, and a characteristic of one of the electrical signals corresponding to one of the storage organs may be determined. Where the storage organs are a sample selected from a crop, those storage organs selected to be in the sample may comprise storage organs that are more likely to sprout sooner. For example, in the case of potatoes, for a given variety, typically larger potatoes sprout before smaller potatoes. Therefore, the sample may preferentially comprise larger potatoes in the sample than smaller potatoes.

[0043] The storage organ may be a root vegetable. The root vegetable may be any one of a potato, carrot or onion.

[0044] The method may further comprise generating, by segmenting the first data, data indicative of a segmented electrical signal corresponding to a predetermined period of time. Processing the first data to determine the characteristic of the electrical signal may comprise processing data indicative of a segmented electrical signal.

[0045] For example, the electrical signal may be sampled from the storage organ at a frequency of 1 Hz. This continuous data stream may be segmented into non-overlapping windows (i.e. predetermined periods of time). One or more of the individual segmented electrical signals may be processed to determine the characteristic. For example, a characteristic may be generated for one or more of the segmented electrical signals.

[0046] The predetermined period of time may be a 24-hour period of time. For example, a segmented electrical signal may correspond to data recorded from the storage organ over a 24 hour period.

[0047] Processing the first data to determine the characteristic of the electrical signal may comprise generating, by decomposing the electrical signal, third data indicative of a plurality of decomposed electrical signals. Processing the first data to determine the characteristic of the electrical signal may further comprise determining, based upon the third data, one or more features for one or more of the plurality of decomposed electrical signals. Processing the first data to determine the characteristic of the electrical signal may further comprise determining, based upon the one or more features, the characteristic of the electrical signal.

[0048] It will be appreciated that the decomposing of the electrical signal may be of the segmented electrical signal in examples where the electrical signal is segmented. In this example, processing the first data to determine the characteristic of the segmented electrical signal may comprise generating, by decomposing the segmented electrical signal, third data indicative of a plurality of decomposed segmented electrical signals, determining, based upon the third data, one or more features for one or more of the plurality of decomposed segmented electrical signals, and determining, based upon the one or more features, the characteristic of the segmented electrical signal.

[0049] The one or more features may comprise one or more values extracted from the respective decomposed electrical signal. The one or more values may comprise a 5th percentile value, a 25th percentile value, a 75th percentile value, a 95th percentile value, a median value, a mean value, a standard deviation value, a variance value, a root mean square value, an entropy value, a number of zero crossings value, and / or a number of mean crossings value.

[0050] The 5th Percentile is a measure of the lower bound of the data distribution. The 25th Percentile is a measure of the first quartile, indicating the lower 25% of the data. The 75th Percentile is a measure of the third quartile, indicating the upper 25% of the data. The 95th Percentile is a measure of the upper bound of the data distribution. The median is the middle value of the data, providing a measure of central tendency. The mean is value obtained by dividing the sum of the values in the dataset by the number of values in the dataset, indicating the central point of the data distribution. The Standard Deviation is a measure of the data's dispersion around the mean. The variance is the square of the standard deviation, providing a measure of data variability. The Root Mean Square (RMS) is a measure of the magnitude of the signal. The entropy value may be the entropy of the probability distribution of a list of unique values, e.g. a measure of the randomness or complexity of the data. The number of zero crossings is the count of times the signal crosses the zero axis, indicative of signal frequency. The number of mean crossings is the count of times the signal crosses its mean value, providing insight into signal variability.

[0051] In some examples, all listed features are extracted from each decomposed electrical signal. In other examples, only a subset are extracted from each decomposed electrical signal. For example, the subset may include a 5th percentile value, a 75th percentile value, a 95th percentile value, a mean value, a standard deviation value, a root mean square value, and a number of zero crossings value. Other subsets are possible.

[0052] Determining the condition indicative of sprouting of the storage organ based on the characteristic may comprise generating fourth data by providing the characteristic of the electrical signal as input to a trained machine learning model. Determining the condition indicative of sprouting of the storage organ based on the characteristic may further comprise determining the condition based upon the fourth data.

[0053] For example, the input to the machine learning model may be one, more or all of the features extracted from each respective decomposed electrical signal, for a single time period, e.g. for a 24 hour time period. The fourth data may be output by the machine learning model, and may be a prediction of the number of days until sprouting. This process may be repeated for subsequent 24 hour time periods. That is, each day, the electrical signal recorded over the previous 24 hours is decomposed into a plurality of decomposed electrical signals, features are extracted from the decomposed electrical signals, which are then input into the machine learning model to output, for example, a predicted number of days until sprouting. The user can then be provided with an estimate each day for the predicted date of sprouting. While a 24 hour time period has been described, it will be appreciated that other time periods could be used, such as a 12 hour time period, 6 hour time period, etc.

[0054] The trained machine learning model may be an extreme gradient boosting regressor (XG Boost) model.

[0055] Determining the condition indicative of sprouting of the storage organ based on the characteristic may further comprise generating fifth data by providing the characteristic of the electrical signal as input to an additional trained machine learning model. The trained machine learning model may have been trained on a random subset of a training dataset. The additional trained machine learning model may have been trained on a different random subset of the training dataset. Determining the condition may be based upon the fourth data and fifth data.

[0056] That is, a bagging (or bootstrap aggregating) method can be used to improve the reliability of the predicted sprouting date. The additional machine learning model may be identical to the first mentioned machine learning model in terms of architecture. While an additional machine learning model is described, it will be appreciated that more may be used, such as ten models.

[0057] The fourth and fifth data may be averaged to determine the condition indicative of sprouting. That is, for a given input, the output from the machine learning model (fourth data) may indicate that there are 60 days until sprouting, whereas for the same given input, the output for from the additional machine learning model may indicate that there are 70 days until sprouting. An average of these may be taken to determine that there are 65 days until sprouting. Similarly, where more models are used, the outputs of all of these models may be averaged.

[0058] Processing the data indicative of a segmented electrical signal to determine the characteristic of the electrical signal and determining the condition indicative of sprouting of the storage organ may be repeated over a plurality of periods of time to determine a plurality of conditions indicative of sprouting of the storage organ. Each condition may correspond to a respective one of the plurality of periods of time. Each condition may indicate a predicted date in which the storage organ is predicted to sprout. For example, over each predetermined period of time (such as 24 hours), the electrical signal obtained from the storage organ is segmented to cover the predetermined period of time, and the data indicative of this segmented electrical signal is processed to determine the condition indicative of sprouting for that particular period of time. This process can be repeated sequentially, such that for each sequential predetermined period of time (e.g. each 24 hours), a new predicted date in which the storage organ is predicted to sprout can be generated.

[0059] The trained machine learning model and / or the additional training machine learning model may have been trained by obtaining training data indicative of a plurality of characteristics of a plurality of electrical signals obtained from a training storage organ. The trained machine learning model and / or the additional training machine learning model may have been trained by processing the training data with the respective machine learning model to obtain a plurality of outputs. The plurality of outputs may indicate a plurality of predicted dates. Each predicted date may be a date in which the training storage organ is predicted to sprout. The trained machine learning model and / or the additional training machine learning model may have been trained by determining an average value based upon the plurality of predicted dates. The trained machine learning model and / or the additional training machine learning model may have been trained by comparing the average value to a ground truth value indicative of the date in which the training storage organ sprouts to determine an error value. The trained machine learning model and / or the additional training machine learning model may have been trained by updating the machine learning model based on the error value.

[0060] The characteristic of the electrical signal may comprise the one or more features of the plurality of decomposed electrical signals. Decomposing the electrical signal may comprise performing a discrete wavelet decomposition. Performing the discrete wavelet decomposition may comprise applying a sym11 wavelet.

[0061] For example, a sym11 discrete wavelet decomposition may be applied to each predetermined time period (e.g. each 24 hour window) of the electrical signal. The electrical signal may be decomposed into wavelet coefficients at a pre-determined number of different levels (e.g. 16) of decomposition, capturing both high-frequency and low- frequency components of the signal. In a second aspect there is provided an apparatus comprising a processor, a memory comprising computer readable instructions, that when executed by the processor, cause the processor to carry out the computer implemented method of the first aspect or fourth aspect.

[0062] In a third aspect there is provided a computer readable storage medium comprising computer readable instructions, the computer readable instructions, when executed by a processor, cause the processor to carry out the computer implemented method of the first aspect or fourth aspect.

[0063] In a fourth aspect there is provided a computer implemented method of determining a condition indicative of sprouting of a potato, the method comprising, receiving, at a processor, first data comprising an electrical signal obtained from the potato, the electrical signal obtained via first and second electrodes attached to the potato, the electrical signal indicating an electric potential difference between two parts of the potato, determining, by the processor, a plurality of spikes in the electric potential difference, determining, by the processor, that there are a predetermined number of the plurality of spikes in the electric potential difference within a predetermined time period, the predetermined number of the plurality of spikes in the electric potential difference within the predetermined time period indicating a condition indicative of sprouting of the potato, and outputting, by the processor, second data indicating the condition indicative of sprouting of the potato.

[0064] Optional features of each aspect may be combined. For example, optional features of the first aspect may be combined with the fourth aspect.

[0065] Brief Description of Figures

[0066] Embodiments are now described, by way of example only, with reference to the accompanying drawings, in which:

[0067] Figure 1 schematically illustrates a system for monitoring a storage organ;

[0068] Figure 2 schematically illustrates a storage organ monitoring device; Figure 3 illustrates a plot of an electrical signal obtained from a potato and containing a spike in the electrical signal;

[0069] Figure 4 illustrates a plot of an electrical signal obtained from a potato and showing a burst of spikes in the electrical signal, indicating imminent sprouting;

[0070] Figure 5 illustrates a plot of several electrical signals, each from an individual potato, demonstrating the synchronous nature of sprouting;

[0071] Figure 6 is a flowchart showing steps carried out to determine a condition indicative of sprouting of a storage organ;

[0072] Figure 7 schematically illustrates a discrete wavelet decomposition process;

[0073] Figure 8 schematically illustrates a trained machine learning model for processing characteristics of an electrical signal; and

[0074] Figure 9 illustrates a plot of a data indicating predicted and actual time to sprouting of a storage organ.

[0075] Detailed Description

[0076] Figure 1 shows a schematic representation of a system 1 for determining a condition indicative of sprouting of a storage organ 100, the system 1 comprising a storage organ monitoring device 110, and optionally an output device 130 and an environmental control apparatus 140. The storage organ 100 may be a root vegetable, such as a potato, carrot, or onion. That is, the storage organ 100 is a part of a plant that normally grows underground and is typically harvested for food consumption. For the forgoing example, the storage organ 100 is a potato. The potato 100 has been harvested (e.g. is out of the ground and has been severed from the rest of the plant of which it was a part) and is stored in an environment 150. The environment 150 may have predefined characteristics, such as specific temperature, humidity, etc. The potato 100 is associated with a second set of potatoes 108 stored at a storage facility in a second environment 160, the second environment 160 is the same or similar to environment 150. For example, the potato 100 may be a sample potato selected from a crop of harvested potatoes, the remaining crop of harvested potatoes stored in the storage facility. In this way, the sample potato 100 (or sample potatoes) may be considered to belong to a first set of potatoes 107, with the remaining crop stored in the storage facility being the second set of potatoes 108. The first set of potatoes 107 may be located in the same physical location as the second set of potatoes 108. For example, the first set 107 may be located in the same storage facility as the second set 108. Alternatively, the first set 107 may be located in a different location to the second set 108. For example, the first set of potatoes 107 may be moved to a laboratory for monitoring. Additionally, while a storage facility is referred to as storing the second set 108, it will be appreciated that the second set may be stored over multiple storage facilities having the same environment.

[0077] Environment 150 may be a controlled environment, configured to maintain an environment in which the potato 100 is stored. In particular, environment 150 may be controlled so as to mimic the second environment 160 in which the second set of potatoes 108 are stored (e.g. the environment of the storage facility). Of course, this may be achieved naturally if the sample potato (i.e. first set 107) is located in the same part of the storage facility as the rest of the crop (i.e. second set 108). In this way, the environment 150 in which the first set of potatoes 107 are stored corresponds to the second environment 160 in which the second set of potatoes 108 are stored. For potatoes, it is typical to keep potatoes at 4 Degrees C or 8 Degrees C. Other root vegetables may be kept at other temperatures. Keeping root vegetables in, for example, a temperature controlled environment helps slow down their natural aging process. While the environments 150, 160 are shown as encompassing the first and second sets 107,

[0078] 108 in Figure 1 , it will be appreciated that the environments 150, 160 may also encompass any one or more of the storage organ monitoring device 110, output device 130, and environmental control apparatus 140.

[0079] The storage organ monitoring device 110 is arranged to receive an electrical signal from the potato 100, the received electrical signal being generated by the potato 100. For example the electrical signal may be an action potential event generated within the potato. The received electrical signal indicates an electric potential difference (e.g. a voltage) generated by the potato 100. In the example of Figure 1 , the storage organ monitoring device 110 receives the electrical signal from the potato 100 using a first electrode 102 and a second electrode 103. As is shown in Figure 1 , the first electrode 102 and the second electrode 103 are electrically coupled to the storage organ monitoring device 110 via a first lead 111 and second lead 112 respectively. That is, the first electrode 102 is coupled to the storage organ monitoring device 110 by the first lead 111 and the second electrode 103 is coupled to the storage organ monitoring device 110 by the second lead 112. The leads 111 , 112 may be any suitable leads being electrically conductive. For example, the leads 111 , 112 may be coaxial cables which are arranged to connect each of the electrodes 102, 103 to the storage organ monitoring device 110. The coaxial cables may be useful for shielding the leads 111 , 112 from external sources of electromagnetic radiation.

[0080] The first electrode 102 is coupled to a first portion 100a of the potato 100. In the example shown in Figure 1 , the first portion 100a is at or near a surface of the potato 100. The first electrode 102 being coupled at or near the surface of the potato 100 may comprise the first electrode 102 being inserted a relatively short distance into the surface of the potato 100. For example, the first electrode 102 may be inserted at or about 5 mm (or less) into the surface.

[0081] The surface of the potato 100 to which the first electrode 102 is coupled may be any surface of the potato 100. Preferably, the surface is a portion of the potato 100 generally opposite a stem end of the potato 100. That is, the surface may be the location from where an apical bud will typically extend when sprouting, or from a region close to where the apical bud will typically extend. The apical bud usually extends from the top of the potato 100. Regions of a potato that sprout, such as where the apical bud sprouts from, exhibit relatively high rates of cell division and biochemical activity during sprouting, and so placing the first electrode 102 in this region increases the likelihood of detecting electrical signals from the potato 100.

[0082] The second electrode 103 is coupled to a second portion 100b of the potato 100. In the example shown in Figure 1 , the second portion 100b is at or near a centre (or centroid) of the potato 100. The second electrode 103 being coupled at or near the centre of the potato 100 may comprise the second electrode 103 being inserted relatively deep into the surface of the potato such that the electrode 103 is located generally within the centre of the potato 100. With the first electrode 102 located generally at the surface of the potato 100, it is preferable to place the second electrode 103 at a position furthest from the surface of the potato 100 (e.g. the centre), the centre being a location with relatively low electrical activity. In order to ensure that any detected electrical signal comes from the centre and not from the surface of the potato 100 through which the second electrode extends, the second electrode is partially insulated (not shown), leaving only a tip exposed to make electrical contact with the centre of the potato 100. For example, the tip may extend for about 3-5 mm beyond an insulated sleeve of the second electrode 103. The insulated sleeve therefore electrically isolates other parts of the potato 100, such as the surface of the potato 100, from the second electrode 103. The first electrode 102 may be a capture electrode, and the second electrode 103 may be a reference electrode.

[0083] The electrical signal received from the potato 100 at the storage organ monitoring device 110 may be characterised as first data indicative of an electrical signal from the potato 100 (or more generally storage organs in the case of other storage organs). The first data is processed by the storage organ monitoring device 110 to determine a characteristic of the electrical signal. The characteristic of the electrical signal may comprise an event in the electrical signal. The event may be a spike in the electrical signal. An example spike 171 is shown in Figure 3, which shows a plot 170 obtained experimentally by testing a particular potato using the storage organ monitoring device 110 as described herein. The plot 170 depicts a voltage trace of voltage (in mV) against time (in hours and minutes), where the measured voltage is a received electrical signal from a potato. As can be seen in Figure 3, between 08:10 and 08:26 the measured voltage is a relatively flat baseline voltage. At about 08:26 there is a sharp downwards deflection in the voltage followed by a slower increase in the voltage returning to the baseline level. It has been observed that the sharp deflection in the voltage typically lasts between tenths of seconds to a few seconds and typically has an amplitude change of > 1 mV, whereas the return to the baseline level may last tenths of seconds to minutes. The spike 171 can be described as a relatively rapid change to the baseline voltage followed by a return to the baseline voltage. However, the exact form of the spike 171 (e.g. its shape, duration, etc.) will vary depending on the specific storage organ, and the family / variety of the storage organ. While the spike is shown as comprising a sharp downwards deflection (decrease) in the voltage, followed by a slower increase, the spike may also comprise a sharp upwards deflection (increase) in the voltage, followed by a slower decrease to the baseline. The spike 171 indicates electrical activity occurring within the potato, such as action potential generated within the potato.

[0084] Detection of spikes in the electrical signal may be carried out using any suitable method. For example, a machine learning model may be trained to detect the spikes. That is, the first data may be processed by a machine learning model running on the storage organ monitoring device 110 to determine the presence of one or more spikes. The machine learning model may have been trained in any suitable way, such as using supervised learning in which labelled training data comprising electrical signals and associated labels identifying the presence (or lack thereof) of one or more spikes are provided to the machine learning model during training.

[0085] The machine learning model may be an artificial neural network, comprising one or more hidden layers and a plurality of weights, the value of the weights being determined during training. The machine learning model may be a deep neural network having any suitable architecture and any suitable input.

[0086] A specific, non-limiting, example neural network along with its input, is provided as follows. An electrical signal (voltage) is measured from the potato over a 10 minute period at 1 sample / second (e.g. such that there are 600 samples in a window). The samples can be filtered, such as by taking the median value for each sample and subtracting to remove the zero drift. A time-frequency decomposition is performed between the frequencies 1 / 300 Hz and 1 / 10 Hz with a resolution of 39 frequency steps to obtain a matrix having dimensions 39x600. The matrix is flattened to obtain a 1 dimensional array of 23400 elements (39*600), where the values of each element represent the power of a specific frequency at a specific point in time in the 10 minute window. The 1 dimensional array of 23400 elements is the input to the neural network. That is, the 1 dimensional array is processed by the neural network to determine the presence of a spike in the electrical signal within the 10 minute window. The neural network is a deep neural network with the following architecture, where the order of the layers indicated below represents their order in the network:

[0087] A dense input layer of 256 units, “relu” activation function, 23400 for the input dimension;

[0088] A dropout layer with dropout rate of 0.15;

[0089] A batch Normalization layer;

[0090] A dense layer of 128 units, “relu” activation function;

[0091] A dense layer of 64 units, “relu” activation function;

[0092] A dropout layer with dropout rate of 0.15;

[0093] A dense layer of 48 units, “relu” activation function;

[0094] A dense layer of 32 units, “relu” activation function;

[0095] A dense layer of 16 units, “relu” activation function; and

[0096] An output dense layer with 2 units and a “Sigmoid” activation function The above neural network was compiled by the present inventors using the “Adam” Optimizer with a learning rate of 0.001 and “binary cross entropy” as loss function, where the metric tracked is the accuracy of the output. The neural network has 6,038,082 parameters, with 6,037,570 trainable parameters and 512 non-trainable parameters.

[0097] During testing, the neural network was trained with labelled data which comprised 75% of a total of an observed dataset (with the remaining 25% used for testing). The total observed data was split such that data from a given potato belongs only to the training dataset or the test dataset. The data was randomized before the training to avoid presenting data from the same potato consecutively. During the training of the model, an internal validation split was performed. The validation split was applied on the training dataset only (the 75% of the total dataset). An example split is an 80 / 20 random split. For every training iteration (epoch), the training dataset was further subdivided into 80% of training and 20% of validation. Additional parameters used during training were a “batch size” of 64 and 100 “epochs”, and where training was terminated if the validation loss had not changed for 10 consecutive epochs. The validation accuracy (average of last 3 training epoch) was found by the inventors to be 83%, while on the test dataset (data unseen by the model), the accuracy was 84.8%.

[0098] Of course, while a specific example of machine learning model has been provided, it will be appreciated that other machine learning models, or changes to the specific architecture, input or training regime, may be used. For example, greater or fewer layers may be used, with greater or fewer parameters. The input to a machine learning model may instead be an image of a voltage trace (such as that shown in Figure 3), and a convolutional neural network may be used to process the image for the detection of spikes. Furthermore, a machine learning model such as that described above is not essential, and other signal processing techniques may be used to identify spikes. For example, characteristics of the electrical signal for a given storage organ (such as duration and / or amplitude change) may be used by an algorithm to determine the presence of spikes. In another example, signature matching techniques may be used, where the electrical signal is compared to a reference signal containing a spike to determine the presence of a spike in the electrical signal.

[0099] As noted above, the first data (e.g. the electrical signal) is processed by the storage organ monitoring device 110 to determine a characteristic of the electrical signal (e.g. a spike in the electrical signal). The characteristic may also comprise a burst of spikes in the electrical signal. A burst of spikes may be defined as a relatively large number of spikes within a relatively short time period, e.g. a predetermined number of spikes within a predetermined time period, where the predetermined values may depend on the storage organ and variety. In particular, it has been found by the inventors that the presence of a burst of spikes in the electrical signal obtained from a storage organ precedes sprouting of the storage organs. As such, the presence of a burst of spikes can be used as an indicator of sprouting (e.g. that the potato 100 is undergoing internal biological changes necessary to sprout).

[0100] Determining the presence of a burst of spikes comprises monitoring the electrical signal and counting how many spikes occur over a particular time period. For example, the number of spikes may be determined each hour (spike rate every hour), and the average spike rate over the last 24 hours may be determined each hour (rolling average over 24 hours). It has been found for potatoes that, if there are on average more than 2 spikes detected per hour over the last 24 hours, the potato will likely sprout within the next 3-41 days. Of course, for different storage organ species, and different varieties, the number of spikes per hour over the last 24 hours which would indicate imminent sprouting may differ. That is, how a burst is defined is dependent on the particular species of storage organ, and its variety. The definition of a burst can be obtained empirically by testing a particular storage organ species / variety. For example, an electrical signal may be recorded from a test storage organ before visual signs of sprouting have occurred in the test storage organ. During recording, the test storage organ is regularly visually inspected. When a first shoot is observed sprouting from the test storage organ, this can be taken as the date of sprouting, and the electrical signal prior to that date may then be processed to identify burst of spikes in the electrical signal prior to the date of sprouting. It will be found that a few days prior to the date of sprouting, there will be an increase in the frequency of the spikes (e.g. a burst of spikes). The characteristics of this identified burst of spikes can therefore be determined. The determined characteristics can then subsequently be used for detecting sprouting in other storage organs of the same species as the test storage organ. Of course, when empirically testing, multiple test storage organs of the same species / variety may be monitored, allowing averages to be calculated (e.g. average characteristics of the burst, such as average value of frequency of bursts). As noted above, the presence of a burst of spikes may be determined by counting how many spikes occur over a particular time period, and comparing this to reference data (e.g. more than 2 spikes per hour over the last 24 hours). However, other methods are possible. For example, the presence of a burst of spikes may be determined using a signature matching technique. For example, the electrical signal may be periodically compared with one or more stored electrical signals, the one or more stored electrical signals having the characteristic (e.g. having the burst of spikes). For example, a database may store electrical signals obtained previously from test storage organs that underwent sprouting while their electrical signals were being monitored. The stored electrical signals may therefore act as a reference to which a presently received electrical signal may be compared. Where the electrical signal matches the one or more stored electrical signals, it may be determined that the presently received electrical signal exhibits the characteristic. In another example, a machine learning technique may be used. For example, a classifier may be used to process the electrical signal and output whether the electrical signal exhibits the characteristic, e.g. a burst of spikes. The classifier may be trained on labelled training data comprising electrical signals obtained previously by the test storage organs mentioned above and which label the presence of a burst of spikes.

[0101] Processing of the first data may comprise performing signal conditioning on the electrical signal. Performing signal conditioning may comprise performing analogue and / or digital signal conditioning. Said signal conditioning may comprise one or more of: amplifying, filtering, normalising, and / or down-sampling the electrical signal. For example, the signal may be downsampled to 1 sample / second.

[0102] Once the characteristic is determined, a condition indicative of sprouting of the potato 100 based on the characteristic is determined. For example, once it has been determined that there has been a burst of spikes in the electrical signal received from the potato 100, it is determined that sprouting is beginning / about to begin within the potato 100 and a shoot will emerge from the surface of the potato 100 shortly (e.g. within the next few days). In this way, the condition may be thought of as an indication of the temporal position in the lifecycle of the potato 100, e.g. the condition of the potato 100 is that it is at a stage in life where shoots are about to sprout from its surface. The time period between the burst of spikes and the first visual signs of sprouting depend on the particular storage organ and variety, as well as the environment, e.g. temperature. However, it has been observed by the inventors that a positive detection of a burst of spikes in the electrical signal is followed 3 to 41 days later by the first visual signs of a shoot, the average being 21 days later.

[0103] When a condition indicative of sprouting has been determined (e.g. based on a detection of a burst of spikes), action may be taken. For example, second data may be output indicating the condition of the potato 100. For example, the storage organ monitoring device 110 may be configured to output the second data when it is determined that a burst of spikes has been observed. The second data may comprise a message or alert that indicates that the spikes have been observed and indicates that a shoot will shortly emerge from the surface of the potato 100. The second data may be output to the output device 130 for output to a user. The output device 130 may be configured to provide feedback to a user in the form of an audio, haptic and / or visual alert. For example, the output device 130 may comprise a speaker, display, light source, haptics, or any other component configured to output audio, visual and / or haptic data. The output device 130 may be a user device, such as a smart phone, smart watch, computer, tablet, etc. Alternatively, or additionally, the output device 130 may comprise a speaker or alarm installed at the storage facility containing the potatoes (e.g. the second set 108) associated with the potato 100. The storage organ monitoring device 110 may transmit the second data to the output device 130 using any suitable means. For example, the second data may be transmitted to the output device 130 via a wired or wireless connection 120, using any suitable protocol.

[0104] The second data may be used as an indication to take action regarding the potatoes in the storage facility (e.g. the second set 108). For example, on receiving the second data, the user may apply a chemical substance to the potatoes in the storage facility to inhibit sprouting. In this way, monitoring of an electrical signal generated by a sample potato (e.g. the first set 107) can be used to protect the integrity of the potatoes in the storage facility (e.g. the second set 108).

[0105] Alternatively or additionally, the second data may comprise a control signal. For example, the second data may comprise a control signal that is configured to cause an apparatus to take automatic action upon receipt of the control signal. In an example, the control signal may be output to an environmental control apparatus 140 that causes the environmental control apparatus 140 to automatically make a change to the second environment 160 of the second set of potatoes 108. For example, the environmental control apparatus 140 may comprise a chemical substance dispenser. The chemical substance dispenser may be configured to spray or otherwise apply a chemical substance to the potatoes in storage (e.g. the potatoes of the second set 108). The chemical substance may prevent sprouting, such as a growth inhibitor. The chemical substance may be chlorpropham, ethylene, or spearmint fogs (where their use is permitted). Other example chemical substances are 3-decen-2-one, 1 ,4- dimethylnapthalene [1 ,4-DMN], ethylene alone, ethylene in combination with 1- methylcyclopropene (1-MCP), L-carvone, and D-limonene. The storage organ monitoring device 110 may transmit the second data to the environmental control apparatus 140 using any suitable means. For example, the second data may be transmitted to the environmental control apparatus 140 via a wired or wireless connection 125 using any suitable protocol.

[0106] While Figure 1 shows a single potato 100 in the first set 107 being monitored, it will be appreciated that the first set 107 may comprise multiple potatoes, e.g. multiple sample potatoes may be monitored. For example, the storage organ monitoring device 110 may have multiple pairs of electrodes / leads such that each pair electrodes / leads may be coupled to a different potato in the first set 107. In this way, a sample comprising multiple potatoes may be selected from the total crop of potatoes in storage. In an example, the first set 107 may comprise eight potatoes, with the second set 108 comprising the remaining potatoes of the crop.

[0107] The storage organ monitoring device 110 may be located at the same location as the potato 100. Alternatively, the storage organ monitoring device 110 may be located elsewhere. For example, the storage organ monitoring device 110 may be hosted in the cloud, e.g. on a remote server. In such cases, the leads 111 and 112 may be connected to one or more intermediate computer devices (not shown in Figure 1), the one or more intermediate computer devices (such as computers and / or routers) being configured to route the first data to the storage organ monitoring device 110 using any suitable network (such as the internet) and / or protocol.

[0108] While the storage organ monitoring device 110 is shown as a single box in Figure 1 , it will be appreciated that functions of the storage organ monitoring device 110 may be split up amongst different computer devices. For example, the storage organ monitoring device 110 may comprise multiple computer devices. A first computer device may be located locally to the sample potato 100 and may be configured to receive the first data. The first data may then be transmitted by the first computer device to a second computer device. The transmission may be over any suitable network, such as the internet. The second computer device may be a remote server. On receipt of the first data, processing of the first data takes place at the second computer device. The second data may be generated by the second computer device and then transmitted over a network, such as the internet, by the second computer device to the output device 130 and / or environmental control apparatus 140, for example. In other examples, the receiving of the first data, processing of the first data, and outputting of the second data may be carried out by a single computer device.

[0109] The storage organ monitoring device 110 may be implemented using any suitable computing apparatus. An example of a suitable computing apparatus is shown in more detail in Figure 2, where the storage organ monitoring device 110 comprises a data acquisition module 10 and a controller 12. The leads 111 , 112 are connected to inputs of the data acquisition module 10. The data acquisition module 10 measures voltage potential differences present between each pair of the electrodes 102, 103 inserted in the storage organ 100. Electrical signals may be recorded in mV level as a function of time and may be recorded at a rate of 240 Hz (i.e. , 240 samples per second) by the data acquisition module 10. The recording frequency of the data acquisition module 10 may, for example, be any value between 1 Hz to 10 KHz.

[0110] In more detail, the data acquisition module 10 comprises an analog filter 9, an amplifier 13, and an analog-to-digital converter (ADC) 14. The analog filter 9 may be a low-pass filter. In an example, the analog filter 9 may be a DC-30Hz filter with a gentle 6dB / octave roll off. Such a filter is useful for minimal ringing so that the transient waveforms have minimal distortion in the time domain. The amplifier 13 may be an analog, non-switching, instrumentation amplifier and may provide an amplification factor between 0 and 100. In an example, the amplifier 13 provides an amplification factor of 4.

[0111] The ADC 14 may be of a Successive Approximation Register (SAR) design. In an example, the ADC 14 may be an 18-bit SAR ADC capable of processing 100K samples per second. In particular, the ADC 14 has a sample-and-hold input and may return a full 18-bit signed value over a + / - 2.048V range for each input. The ADC 14 may be implemented using ADS8777 ADC made by Texas Instruments.

[0112] The controller 12 comprises a processor 12a which is configured to read and execute instructions stored in a volatile memory 12b which takes the form of a random access memory. The volatile memory 12b stores instructions for execution by the processor 12a and data used by those instructions. For example, in use, the data acquired by the data acquisition module 10 may be stored in the volatile memory 12b.

[0113] The controller 12 further comprises non-volatile storage in the form of a hard disc drive 12c. The data acquired by the data acquisition module 10 may be stored on the hard disc drive 12c. The controller 12 further comprises an I / O interface 12d to which are connected data capture and peripheral devices used in connection with the controller 12. In the example shown, a display 12e is connected to the I / O interface 12d to display output from the controller 12. The display 12e may, for example, display a representation of the data acquired by the data acquisition module 10. The display 12e may be provided locally to the storage organ monitoring device 110 (e.g. as a screen), or remotely from the storage organ monitoring device 110. For example, a display associated with a separate device (e.g. a mobile computing device) may be used as a display for the storage organ monitoring device 110. Additionally, the display 12e may display images generated by processing of the data acquired by the data acquisition module 10. The display 12e may display the first data and / or second data. Additionally or alternatively, a touchscreen associated with the display 12e may operate as a user input device, so as to allow a user to interact with the controller 12. Alternatively or additionally, separate input devices may be also connected to the I / O interface 12d, such as a mouse and / or keyboard. A network interface 12f allows the controller 12 to be connected to an appropriate computer network so as to receive and transmit data from and to other computing devices. While not shown, the output device 130 and / or the environmental control apparatus 140 may be connected to the network interface 12f such that the second data may be transmitted to the output device 120 and / or the environmental control apparatus 140. The processor 12a, volatile memory 12b, hard disc drive 12c, I / O interface 12d, and network interface 12f, are connected together by a bus 12g.

[0114] In addition to the peripheral devices described above being connected to the I / O interface 12d, the output of the data acquisition module 10 is also connected to the I / O interface 12d. By virtue of these connections, potential differences sensed at the electrodes 102, 103 can be processed and converted to a digital signal by the data acquisition module 10 and subsequently processed by the processor 12a and stored in the hard disc drive 12c.

[0115] The controller 12 may be connected to an external computer / server via the network interface 12f. In that case, the external computer / service may further process the digitalised signals obtained by the controller 12. In an example, the digitalised signals may be extracted and processed using a data processing software by the external computer / server. Further or alternatively, the controller 12 may be connected to a single board computer via the network interface 12f, such that the digitalised signals obtained by the controller are collected into the single board computer.

[0116] The controller 12 may comprise a microcontroller to which the output of the ADC 14 is fed. In an example, the controller 12 may comprise a single board computer which provides a STM32F103 microcontroller. The STM32F103 microcontroller uses the ARM M3 processor design and runs at 72MHz. The STM32F103 microcontroller also provides USB connectivity.

[0117] Figure 4 shows a plot 2 of obtained experimental results. In particular, Figure 4 shows an electrical signal 252 in the form of a voltage trace. The electrical signal 252 is an example of the electrical signal as referred to above with respect to Figure 1 . That is, the voltage trace 252 may be typical of that which is observed by the storage organ monitoring device 110 of Figure 1. The voltage trace 252 comprises a voltage (in millivolts) measured from a potato (Innovator) over a period of time. The rate at which the voltage was measured from the potato was 256 times per second and the voltage was measured on a 15 to -15 Millivolt scale 230. However, the value of the voltage plotted is the value recorded each 5 minutes. That is, the resolution of the plot in time is 5 minutes. The voltage trace 252 was processed to determine the presence of spikes as described above. Note that the scale of the voltage trace 252 is at too low a resolution to show spikes, such as that shown in Figure 2. However, each individual occurrence of a spike in the electrical signal is plotted along line 260, shown as circles on the plot, each circle corresponding to the detection of a single spike. The x-axis is the same as that used to plot the voltage trace 252. Note that the position of each circle along the y-axis (e.g. along the vertical axis when looking at the figure) has no significance. The positioning along the y-axis is simply for helping to visualise the individual data points plot along line 260. An example of a specific spike occurring at a specific point in time is data point 261.

[0118] As can be seen in the plot along line 260, there is a cluster of data points 201 that occur approximately between 1 March and 6 March. The cluster indicates a burst of spikes. That is, between the 1 March and 6 March there was a burst of spikes in the electrical signal observed from the potato. Along with the individual spikes plotted along line 260, plot 2 also shows, as a bar chart located at the bottom of plot 2, a count per hour of spikes. As can be seen, the cluster of data points 201 corresponds with an increase 200 in the count per hour of spikes.

[0119] Plot 2 indicates at line 240 the time at which the occurrence of the first visible signs of sprouting in the potato were observed, e.g. the first shoot was observed extending from the surface of the potato. As can be seen, the cluster of data points 201 occurred a few days prior to visible sprouting (indicated by line 240). These experimental results show that a burst of spikes generated by the potato (represented by the cluster of data points 201) provides an indication that sprouting of the potato is due to occur.

[0120] Figure 5 shows a plot 3 depicting further obtained experimental data. In particular, the plot 3 shows 16 individual voltage traces observed from 16 individual potatoes of the same variety (Innovator), and plotted against time. Each one of the voltage traces may be typical of that which is observed by the storage organ monitoring device 110 of Figure 1. Each one of the voltage traces was detected from each one of their respective potatoes. Not all individual voltage traces have been labelled in Figure 4 for clarity, but as an example, one of the traces 320i depicted on plot 3 is labelled and corresponds to one of the 16 voltage traces corresponding to one of the 16 potatoes. Each one of the 16 potatoes was stored at 8 Degrees C for the duration of the experiment. Each one of the 16 potatoes originated from the same crop as every other one of the 16 potatoes.

[0121] Plot 3 further depicts 16 vertical dotted lines (310a, 310b, 310c, 31 Od, 31 Oe, 31 Of, 310g, 31 Oh, 31 Oi, 31 Oj, 31 Oj, 310k, 3101, 310m, 310n, 310o, and 310p) along each one of the voltage traces. Each one of the 16 vertical dotted lines corresponds to one of the voltage traces and its respective one of the 16 potatoes. For example, vertical dotted line 310i corresponds to voltage trace 320i, which in turn corresponds to one of the 16 potatoes. Each one of the vertical dotted lines indicates a point in time, for its respective potato, that the potato showed first visible signs of sprouting activity, as detected through timelapse video acquisition. For example, vertical dotted line 31 Oi indicates that, for voltage trace 320i and its respective potato, the respective potato showed first visible signs of sprouting activity approximately on January 17th. Note that the resolution of the voltage trace, like that in Figure 4, is too low to show the characteristic burst of spikes prior to visible signs of sprouting.

[0122] The experimental results shown in plot 3 demonstrate that the first visible signs of sprouting activity for each one of the 16 potatoes occurred within a relatively narrow time frame 300. The time frame 300 was approximately one month, with most potatoes sprouting within 2-3 weeks of the first potato sprouting. It is therefore clear that each one of the 16 potatoes which were stored at the same temperature and originated from the same crop have developed at a similar rate along their respective lifecycle. This demonstrates that other potatoes also stored at the same temperature and originating from the same crop as each one of the 16 potatoes will likely also develop at a similar rate. Therefore, it can be seen that by monitoring a sample set of potatoes taken from a crop (e.g. the first set 107), reliable data regarding sprouting of the remaining potatoes of the crop (e.g. the second set 108) can be inferred. This is advantageous because a relatively small sample of potatoes originating from a relatively large crop may be monitored, rather than monitoring every one of the potatoes in the crop, which in most cases would be impractical.

[0123] With reference to Figure 6, there is now described a method of determining a condition indicative of sprouting of a storage organ. The method may be implemented using the system 1 as described above.

[0124] At step S1 , first data is received, the first data indicative of an electrical signal from the storage organ. The electrical signal may be indicative of an electric potential difference within the storage organ, as described above.

[0125] At step S2, the first data is processed to determine a characteristic of the electrical signal. For example, the electrical signal may be processed to determine a presence of a predetermined number of events (such as spikes in the electrical signal) within a predetermined time period, as described above. At step S3, a condition is determined, the condition indicative of sprouting of the storage organ based on the characteristic. That is, based on the presence of the predetermined number of events within the predetermined time period, it may be determined that the storage organ will shortly exhibit shoots extending from the surface of the storage organ (e.g. the storage organ is about to sprout), as described above.

[0126] At step S4, second data is output indicating the condition of the storage organ. For example, an alert may be output to alert a user to the detected condition. Alternatively or additionally, a control signal may be output, the control signal configured to cause a change in the environment of the second set of potatoes, as described above.

[0127] While it has been described that the characteristic may correspond to a burst of spikes, in another implementation, the characteristic may be one or more statistical features extracted from the electrical signal. The one or more statistical features extracted from the electrical signal may be processed by a trained machine learning model configured to predict when sprouting will occur. Such an example method is now described.

[0128] The first data indicative of the electrical signal, described above, may be indicative of an electrical signal that corresponds to a predetermined period of time. For example, the electrical signal may correspond to a 24-hour period, or window, of time, i.e. the electrical signals from the storage organ were monitored for 24 hours. In some embodiments, the first data may be segmented (i.e. divided into chunks) to determine data indicative of a segmented electrical signal (i.e. a data segment). Each data segment may correspond to the predetermined period of time. The data segment may then be provided as the data processed to determine characteristics of the signal. For example, an electrical signal recorded over 96 hours may be segmented into four 24-hour signals, each for processing separately. That is, at least one of the data segments (i.e. a data segment for a 24-hour signal) may be processed to determine the characteristic of the electrical signal. The inventors have realised that the accuracy of the techniques described herein tend to increase the longer the window used to segment the electrical signal, a 24-hour window being found to be optimal. The segmented electrical signal may be non-overlapping with other segments of the electrical signal. Segmentation ensures that temporal features are preserved in the data for each period of time. For example, in a specific embodiment, an electrical signal is sampled from one or more tubers at a frequency of 1 Hz. This continuous data stream is segmented into nonoverlapping windows, each representing a 24-hour period for a specific tuber. The segmentation ensures that temporal features are preserved and each window is treated as an independent sample for subsequent analysis. The 24 hour segmented data can then be processed independently.

[0129] Processing the first data may comprise generating, by decomposing the electrical signal, data indicative of a plurality of decomposed electrical signals. That is, to extract meaningful information from the raw electrical signals, the signals may be decomposed. With reference to Figure 7, the electrical signal 900 may be decomposed using discrete wavelet decomposition. Other forms of decomposition are envisaged. In discrete wavelet decomposition, an electrical signal may be decomposed according to a wavelet (e.g. Symlet, Haar, Gaussian, etc.). The inventors have found that applying a sym11 (i.e. a type of Symlet) wavelet is particularly advantageous for extracting features from the first data. The decomposed signal may then be processed using both a high pass (“HP”) and a low pass (“LP”) filter. Processing the electrical signal using the HP filter generates coefficients representing a first level 902 HP decomposition 912 of the electrical signal (typically known as detail coefficients). Processing the electrical signal using the LP filter generates coefficients representing a first level 902 LP decomposition 910 of the electrical signal (typically known as approximation coefficients). That is, the electrical signal is decomposed into two different sets of coefficients for each level (e.g. first level 902, second level 904, third level 906, and fourth level 908). To generate the coefficients of the second level 904, the first level 902 LP decomposition 910 of the electrical signal may be decomposed using the same process as for the first level 902 decomposition. This process may be repeated over multiple levels, e.g. 904, 906, 908. While it is appreciated that any number of levels are envisaged, an example number of levels used in implementations of this disclosure is sixteen levels. It will however be appreciated that fewer levels may be used, such as ten levels, and more levels may be used, such as twenty levels. By decomposing the electrical signal, for example using discrete wavelet decomposition, it is possible to extract meaningful information from the original electrical signal 900.

[0130] Processing the first data may further comprise, determining one or more features for each of the plurality of decomposed signals. That is, the data indicative of a plurality of decomposed electrical signals may be processed to extract said features. For example, the electrical signal 900 of Figure 7 is decomposed at four different levels 902, 904, 906, 908. Therefore, in this example, the data may be indicative of eight decomposed electrical signals (i.e. signals 910, 912, 914, 916, 918, 920, 922, 924), two for each level. For each of the decomposed electrical signals (910, 912, 914, 916, 918, 920, 922, 924), one or more features are extracted or computed. The one or more features may be a 5thpercentile value, a 25thpercentile value, a 75thpercentile value, a 95thpercentile value, a median value, a mean value, a standard deviation value, a variance value, a root mean square value, an entropy value, a number of zero crossings value, and / or a number of mean crossings value. That is, the coefficients for each decomposed electrical signal (910, 912, 914, 916, 918, 920, 922, 924) may be used to determine bounds of the distribution of coefficients (e.g. 5thpercentile). Likewise, a mean, median, mode, standard deviation, variance etc. may be computed based on the distribution of coefficients. Likewise, a measure of the magnitude of the signal (e.g. using root mean squared) and / or a measure of complexity / randomness of the data in the distribution (e.g. entropy) may be computed. Likewise, a number of zero crossings, i.e. the number of times the signal crosses the zero axis of the distribution, and / or a number of mean crossings, i.e. the number of times the signal crosses the mean value for that distribution, may be computed.

[0131] In this way, a comprehensive set of features that encapsulate the characteristics of the original electrical signal 900 may be determined. The characteristics of the electrical signal may be determined based upon the one or more features. Simply, the characteristics of the electrical signal may be the plurality of features for the plurality of decomposed electrical signals. It will be appreciated that one or more features may also be determined for the original electrical signal 900 in the same way described above, and included, i.e. alongside the plurality of features for the plurality of decomposed electrical signals, as a characteristic of the electrical signal. Referring to Figure 7, the characteristics of the electrical signal may be a combination, i.e. concatenation, of the one or more features for each of the decomposed electrical signals 910, 912, 914, 916, 918, 920, 922, and 924, for example a concatenated feature vector. For a specific example of how the characteristics (e.g. plurality of features) are used to determine the condition indicative of sprouting, refer to Figure 8 as described below. The extracted plurality of features described above may be provided as input to a trained machine learning model to generate (i.e. output) output data 1300. Figure 8 depicts such a trained machine learning model 1200. Input data 1100 (e.g. a feature vector representing the characteristics, or plurality of features) may be provided as input to the trained machine learning model 1200. That is, the one or more features X 1110, the one or more features Y 1120, and the one or more features Z 1130 may, for example, be concatenated and provided as input to the trained machine learning model 1200. The one or more features X, Y, and Z may correspond to different decomposed electrical signals (e.g. decomposed signals 910, 912, and 914 in Figure 7). Only three features 1110, 1120, 1130 are shown for simplicity, but it will be appreciated that the input data 1100 may include features extracted from all of the decomposed signals. The output data 1300 may indicate the condition indicative of sprouting of the storage organ 100. For example, the output data 1300 may be indicative of a predicted number of days until sprouting, such as an integer. In another example, the output data 1300 may be a continuous value indicating a probability that the storage organ 100 will sprout in the next predetermined period of time. In some examples, the output data 1300 may be positive or negative. In this example, negative values may indicate that the electrical signal was recorded preceding the predicted day of sprouting. That is, output data 1300 being -6 may indicate that the electrical signal was measured 6 days before the predicted day of sprouting, i.e. a prediction that the storage organ 100 will sprout in 6 days. The trained machine learning model 1200, may have been trained using any suitable method (e.g. backpropagation with gradient descent) using labelled data as described herein, such as using supervised learning in which labelled training data comprising characteristics of the electrical signals (e.g. the extracted features for each of the decomposed electrical signals) and associated labels identifying the number of days prior to sprouting. During training, the output from the model (predicted number of days before sprouting) for a given training input can be compared to the label of that training input, the label indicating the actual number of days prior to sprouting. A suitable loss can be calculated and the model updated accordingly using an appropriate training algorithm, e.g. backpropagation for neural network type models, or an exact-greedy algorithm for the particular model described below.

[0132] The trained machine learning model 1200 may be any suitable type of machine learning model, such as an artificial neural network, support vector machine, linear regression model, etc. In implementations, the inventors have discovered that using an extreme gradient boosting regressor (“XGBoost”) as the machine learning model was particularly effective for the purpose of determining a condition indicative of sprouting in the storage organ 100.

[0133] In a specific implementation, the XGBoost model was generally configured with the default parameters in the XGBRegressor class found within the open-source Python package named “xgboost”, with the following modifications to default:

[0134] 'colsample_bylevel': 0.6960655132271771 ,

[0135] 'colsample_bynode': 1 ,

[0136] 'colsample_bytree': 0.8553811375465922,

[0137] 'gamma': 6,

[0138] 'learn i ng_rate' : 0.04914086531502215 ,

[0139] 'max_depth': 7,

[0140] 'min_child_weight': 7,

[0141] 'reg_alpha': 9,

[0142] 'regjambda': 1

[0143] Figure 9 depicts a plot 700 of data indicating a predicted time to sprouting 704 and an actual time to sprouting 702. The data was measured for a single storage organ 100 between November 2022 and January 2023. Each data point on the plot for the predicted time to sprouting 704 is the output data 1300 described with reference to Figure 8. For example, the trained machine learning model 1200 generated output data 1300 (i.e. data point 730) in November indicating -60 (i.e. a predicted 60 days until sprouting). As depicted, the predicted sprouting date 714 occurred approximately around the actual sprouting date 712, and data indicating the predicted time to sprouting 704 is highly correlated with the data indicating the actual time to sprouting 702. That is, by decomposing the electrical signal, extracting features for each decomposition, and using the trained machine learning model 1200, determining a condition indicative of sprouting of the storage organ 100 may be ascertained with a high degree of accuracy.

[0144] However, the inventors have appreciated that the predictions (e.g. predicted time to sprouting 704) sometimes exhibit a degree of noise. That is, variance for the predictions over a given period of time may be high. To further improve the methods described herein, bagging or bootstrap aggregating methods may be used. For example, determining the condition indicative of sprouting of the storage organ 100 based on the characteristic may further comprise generating fifth data by providing the characteristic of the electrical signal as input to an additional trained machine learning model. That is, the characteristic may be provided to two (or more) trained machine learning models (i.e. the trained machine learning model 1200) to both generate a separate output (i.e. the output data 1300). In a specific implementation, ten different models trained on ten different subsets of the training dataset are used to generate ten separate outputs. As mentioned above, the trained machine learning models (i.e. both original and additional) may be trained using training data. In this example, both the original and additional trained machine learning models may be trained on different and / or random subsets of the training dataset. That is, the training dataset may be randomly sampled for subsets of the data in the training dataset, each model being trained on a different random sample of that training data. Any suitable sampling method may be used (e.g. simple random subsampling).

[0145] Once a prediction (i.e. output data 1300) is generated by each of the multiple trained machine learning models, the predictions may be used to determine the condition indicative of sprouting of the storage organ 100. For example, and average value (e.g. mean or mode) of the individual outputs from each model may be determined, and the average value may be provided as the prediction for a given 24 hour window of the electrical signal. An indication of the confidence may be provided as output to the user. For example, if the standard deviation of the predictions is greater than a predetermined threshold, it may be assumed that there is a relatively low confidence that the predicted date is correct, given the large variation in outputs from the models. This information may be provided to the user, along with the prediction. In other cases, a prediction that has a low confidence, such as the standard deviation of the predictions being greater than the predetermined threshold, may not be output to the user. Rather, an indication may be output to highlight that a prediction is not possible for the given 24 hour window of the electrical signal.

[0146] It will be appreciated that the multiple predictions may be used in any suitable way to determine the condition indicative of sprouting of the storage organ 100. By utilizing multiple trained machine learning models trained in such a way (i.e. with random subsamples), the accuracy of predictions may be improved. In yet another implementation, the data output (i.e. the output data 1300) by the trained machine learning model 1200 may be post-processed prior to output, or prior to determining an error during training of the model. For example, during inference the machine learning model 1200 may output the predicted number of days to sprouting. The predicted number of days may be converted into a date (i.e. the date of predicted sprouting), e.g. -6 may indicate 6 days to sprouting and may be converted to a predicted date for sprouting, such as 10thMarch 2024. During inference, processing the data indicative of an electrical signal (e.g. the segmented electrical signal described above) to determine the characteristic of the electrical signal and determining the condition indicative of sprouting of the storage organ may be repeated over a plurality of periods of time (e.g. for numerous dates). That is, for a given date, features extracted from a corresponding electrical signal may be provided as input to the trained machine learning model 1200 to output the predicted number of days to sprouting for that date. This may be repeated over a plurality of periods of time (e.g. 1 March 2024, 2 March 2024, 3 March 2024, etc.). A count of the number of predictions for a given predicted date may be tracked, e.g. by using a histogram. An average value may be calculated from the histogram, e.g. based upon the plurality of predicted dates. For example, the mean, median, or mode of all of the predicted dates collected so far may be used to predict the date of sprouting. For example, there may be 265 predictions for 10 March 2024, 397 predictions for 11 March 2024, and 112 predictions for 12 March 2024. In this example, the mean of all of the values (e.g. Unix timestamps) representing the predicted dates may be taken to determine the average predicted date of sprouting, e.g. as a Unix timestamp. That is, a sum of all of the Unix timestamps for all of the predicted dates divided by 774 (i.e. 265+397+112) may be the average value. In another example, if the number of predictions (i.e. number of occurrences) for 10 March 2024 is 145, the number of predictions for 11 March 2024 is 289, and the number of predictions for 12 March 2024 is 212, the average value may be a value (e.g. a Unix timestamp) representing the 11 March 2024 determined as the average using mode as the average metric. It will be readily appreciated that the average value for the predicted dates may be determined in any suitable way. In this way, the predicted output from the trained machine learning model 1200 is combined with previous outputs from previous days to obtain an average (such as mean, mode or median), of a date of predicted sprouting. As time progresses, the predicted output becomes more accurate as the prediction is based on outputs generated from features of the electrical signal over multiple time periods. During training, an untrained (or trained, if undergoing additional training) machine learning model may be trained based on an error value. As mentioned, the trained machine learning model 1200 may have been trained using training data. That is, the training data may comprise characteristics of electrical signals obtained from a training storage organ and a ground truth value indicative of the date in which the training storage organ sprouts, e.g. a Unix timestamp representing the date 12 March 2024. The training storage organ may be the storage organ 100 (i.e. the same storage organ 100 used during inference), or a different storage organ altogether. It will be appreciated that a plurality of training storage organs may be used during training. Multiple sets of such data may be available, e.g. multiple sets of electrical signals obtained from multiple training storage organs. The characteristics of the electrical signals in each training example may be processed by the untrained machine learning model to output a predicted date for sprouting. These predicted dates may be averaged to determine an average value, e.g. a Unix timestamp, as described above with respect to model inference. It will be understood that such a process may occur in batches, for example batches of 16, 32, 64, etc. training examples. An error value may subsequently be determined by comparing the average value for these training examples, i.e. the average value based upon the plurality of predicted dates, which may be the result of processing the training examples with the untrained machine learning model 1200, and the ground truth value from the training data. For example, the error value may be determined based upon a comparison between “1727991021” being the average value and “1728250221” being the ground truth value, such as 1728250221 - 1727991021 = 259200, where 259200 is the error value. In some examples, such an error value as described is used fortraining only when the error value is below a predetermined threshold. It will be readily appreciated that an error value may be determined based upon any suitable form of comparison, and the average and ground truth values may be any suitable values. The model may subsequently be trained based upon the error using any suitable method as previously described. It will be readily appreciated that any of the machine learning models (e.g. the additional trained machine learning models) described herein may have been trained in such a way using the error value as described.

[0147] It will be appreciated that embodiments disclosed herein can be implemented in any convenient form. For example, embodiments disclosed herein may be implemented by appropriate computer programs which may be carried on appropriate carrier media which may be tangible carrier media (e.g. disks) or intangible carrier media (e.g. communications signals). Embodiments disclosed herein may also be implemented using suitable apparatus which may take the form of programmable computers running computer programs arranged to implement the embodiments disclosed herein.

[0148] Embodiments of the subject matter and the operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions, encoded on computer storage medium for execution by, or to control the operation of, data processing apparatus. Alternatively or in addition, the program instructions can be encoded on an artificially-generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus. A computer storage medium can be, or be included in, a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more of them. Moreover, while a computer storage medium is not a propagated signal, a computer storage medium can be a source or destination of computer program instructions encoded in an artificially-generated propagated signal. The computer storage medium can also be, or be included in, one or more separate physical components or media (e.g., multiple CDs, disks, or other storage devices).

[0149] The operations described in this specification can be implemented as operations performed by a processor on data stored on one or more computer-readable storage devices or received from other sources.

[0150] The term “processor” encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, a system on a chip, or multiple ones, or combinations, of the foregoing. The apparatus can include special purpose reprogrammable logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit). The apparatus can also include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or a combination of one or more of them. The apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.

[0151] Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, a processor will receive instructions and data from a read-only memory or a random access memory or both. The essential elements of a computer are a processor for performing actions in accordance with instructions and one or more memory devices for storing instructions and data. Devices suitable for storing computer program instructions and data include all forms of computer-readable media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD- ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry and fiber-optic platform for faster data transfer remotely.

[0152] To provide for interaction with a user, embodiments of the subject matter described in this specification can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor including audio, for displaying information (e.g. an indication and / or alert) to the user. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback.

[0153] Although the disclosure has been described in terms of preferred embodiments as set forth above, it should be understood that these embodiments are illustrative only and that the claims are not limited to those embodiments. For example, while the example with respect to Figure 1 relates to a potato 100, it will be readily appreciated that the potato 100 may be replaced with any other storage organ, such as a carrot or onion. The skilled person practiced in the art will be able to make modifications and alternatives in view of the disclosure which are contemplated as falling within the scope of the appended claims. Each feature disclosed or illustrated in the present specification may be incorporated in the disclosure, whether alone or in any appropriate combination with any other feature disclosed or illustrated herein.

Claims

CLAIMS:

1. A computer implemented method of determining a condition indicative of sprouting of a storage organ, the method comprising: receiving first data indicative of an electrical signal from the storage organ; processing the first data to determine a characteristic of the electrical signal; and determining a condition indicative of sprouting of the storage organ based on the characteristic.

2. The computer implemented method of claim 1 , wherein the characteristic of the electrical signal comprises an event in the electrical signal.

3. The computer implemented method of claim 2, wherein the event in the electrical signal is a spike in the electrical signal.

4. The computer implemented method of claim 2 or 3, wherein the characteristic of the electrical signal comprises a burst of events in the electrical signal.

5. The computer implemented method of claim 4, wherein the burst of events comprises a predetermined number of events within a predetermined time period.

6. The computer implemented method of claims 2 to 5, wherein the event indicates an action potential event generated within the storage organ.

7. The computer implemented method of any preceding claim, wherein the electrical signal is indicative of an electric potential difference.

8. The computer implemented method of claim 7, wherein the electric potential difference is measured between a first electrode coupled to a first portion of the storage organ and a second electrode coupled to a second portion of the storage organ.

9. The computer implemented method of claim 8, wherein the first portion is at or near a surface of the storage organ.

10. The computer implemented method of claim 8 or 9, wherein the second portion is at or near a centre portion of the storage organ.11 . The computer implemented method of any preceding claim, wherein the storage organ belongs to a first set of storage organs; and further comprising: determining, based on the determination of the condition indicative of sprouting of the storage organ, a condition indicative of sprouting of storage organs belonging to a second set of storage organs, wherein the second set of storage organs are associated with the first set of storage organs.

12. The computer implemented method of any preceding claim, further comprising outputting second data indicating the condition of the storage organ.

13. The computer implemented method of claim 12, wherein outputting the second data indicating the condition of the storage organ comprises outputting a warning.

14. The computer implemented method of claims 12 or 13, when dependent on claim 11 , wherein outputting second data indicating the condition of the storage organ comprises outputting a control signal to an environmental control apparatus, the environmental control apparatus configured to automatically change an environment in which the second set of storage organs are stored.

15. The computer implemented method of any preceding claim, wherein the first data is indicative of an electrical signal from each of a plurality of storage organs; and processing the first data comprises processing the first data to determine the characteristic of at least one of the electrical signals.

16. The computer implemented method of any preceding claim, wherein the storage organ is a root vegetable.

17. The computer implemented method of claim 16, wherein the root vegetable is any one of a potato, carrot or onion.

18. The computer implemented method of any preceding claim, further comprising: generating, by segmenting the first data, data indicative of a segmented electrical signal corresponding to a predetermined period of time; and wherein processing the first data to determine the characteristic of the electrical signal comprises processing data indicative of a segmented electrical signal.

19. The computer implemented method of claim 18, wherein the predetermined period of time is a 24-hour period of time.

20. The computer implemented method of any preceding claim, wherein processing the first data to determine the characteristic of the electrical signal comprises: generating, by decomposing the electrical signal, third data indicative of a plurality of decomposed electrical signals; determining, based upon the third data, one or more features for one or more of the plurality of decomposed electrical signals; and determining, based upon the one or more features, the characteristic of the electrical signal.21 . The computer implemented method of claim 20, wherein the one or more features comprise one or more values extracted from the respective decomposed electrical signal, the one or more values comprising: a 5th percentile value, a 25th percentile value, a 75th percentile value, a 95th percentile value, a median value, a mean value, a standard deviation value, a variance value, a root mean square value, an entropy value, a number of zero crossings value, and / or a number of mean crossings value.

22. The computer implemented method of any preceding claim, wherein determining the condition indicative of sprouting of the storage organ based on the characteristic comprises: generating fourth data by providing the characteristic of the electrical signal as input to a trained machine learning model; and determining the condition based upon the fourth data.

23. The computer implemented method of claim 22, wherein the trained machine learning model is an extreme gradient boosting regressor (XGBoost) model.

24. The computer implemented method of claims 22 or 23, wherein determining the condition indicative of sprouting of the storage organ based on the characteristic further comprises: generating fifth data by providing the characteristic of the electrical signal as input to an additional trained machine learning model, the trained machine learning model having been trained on a random subset of a training dataset, and the additional trained machine learning model having been trained on a different random subset of the training dataset; and wherein determining the condition is based upon the fourth data and fifth data.

25. The computer implemented method of any of claims 18 to 24, wherein: processing the data indicative of a segmented electrical signal to determine the characteristic of the electrical signal and determining the condition indicative of sprouting of the storage organ are repeated over a plurality of periods of time to determine a plurality of conditions indicative of sprouting of the storage organ; each condition corresponding to a respective one of the plurality of periods of time; and each condition indicating a predicted date in which the storage organ is predicted to sprout.

26. The computer implemented method of any preceding claim when dependent upon claims 22 or 24, wherein the trained machine learning model and / or the additional training machine learning model were trained by: obtaining training data indicative of a plurality of characteristics of a plurality of electrical signals obtained from a training storage organ; processing the training data with the respective machine learning model to obtain a plurality of outputs, the plurality of outputs indicating a plurality of predicted dates, each predicted date being a date in which the training storage organ is predicted to sprout; determining an average value based upon the plurality of predicted dates; comparing the average value to a ground truth value indicative of the date in which the training storage organ sprouts to determine an error value; and updating the machine learning model based on the error value.

27. The computer implemented method of claim 20 or any of claims 21 to 26 when dependent upon claim 20, wherein the characteristic of the electrical signal comprises the one or more features of the plurality of decomposed electrical signals.

28. The computer implemented method of any preceding claim, when dependent on claim 20, wherein decomposing the electrical signal comprises performing a discrete wavelet decomposition.

29. The computer implemented method of claim 28, wherein performing the discrete wavelet decomposition comprises applying a sym11 wavelet.

30. A computer implemented method of determining a condition indicative of sprouting of a potato, the method comprising: receiving, at a processor, first data comprising an electrical signal obtained from the potato, the electrical signal obtained via first and second electrodes attached to the potato, the electrical signal indicating an electric potential difference between two parts of the potato; determining, by the processor, a plurality of spikes in the electric potential difference; determining, by the processor, that there are a predetermined number of the plurality of spikes in the electric potential difference within a predetermined time period, the predetermined number of the plurality of spikes in the electric potential difference within the predetermined time period indicating a condition indicative of sprouting of the potato; and outputting, by the processor, second data indicating the condition indicative of sprouting of the potato.

31. An apparatus comprising: a processor a memory comprising computer readable instructions, that when executed by the processor, cause the processor to carry out the computer implemented method of any preceding claim.

32. A computer readable storage medium comprising computer readable instructions, the computer readable instructions, when executed by a processor, cause the processor to carry out the computer implemented method of any of claims 1 to 30.

Images