A produce monitoring system and method

Abstract

A method of monitoring produce comprises: mounting a portable computing device to a mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the mounting bracket and initiating a second operation mode; capturing, after demounting the portable computing device, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

Description

[0001] A PRODUCE MONITORING SYSTEM AND METHOD

[0002] FIELD

[0003] The invention relates to systems and methods for monitoring produce, which may be agricultural produce such as potatoes, for example. Aspects of the invention also relate to a case or protective housing.

[0004] BACKGROUND

[0005] Produce, such as agricultural produce, can be monitored to determine one or more parameters associated with the produce. These parameters could include size, weight, colour, quality, or health, for example. The determined parameters can then be used to grade or categorise the produce. For example, potatoes can be graded or categorised by size.

[0006] Traditionally, a producer, such as a farmer, would determine any parameters associated with their produce manually. This could be by a manual or physical inspection of the produce, carried out by the producer, for example. This process could involve manually sizing, weighing, or inspecting the colour, quality, or health of the produce, for example.

[0007] These traditional processes are labour-intensive, time-consuming, and prone to human error.

[0008] The present invention seeks to alleviate problems associated with the prior art.

[0009] BRIEF DESCRIPTION OF THE INVENTION

[0010] Disclosed is a method of monitoring produce, comprising: mounting a portable computing device to a mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the mounting bracket and initiating a second operation mode; capturing, after demounting the portable computing device, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

[0011] The portable computing device may determine the at least one parameter using a machine learning model. The portable computing device may be communicatively connected to a mounted camera in the mounted operation mode.

[0012] The mounted camera may be mounted to a part of a harvester or grader.

[0013] The portable computing device may be configured to control the mounted camera, and the method may include the portable computing device capturing images automatically in the mounted operation mode.

[0014] The portable computing device may be communicatively connected to a second camera in the second operation mode.

[0015] The portable computing device may be configured to control the second camera, and the method may include capturing an image in response to a user input in the second operation mode.

[0016] The mounting bracket may be installed on a harvester or grader.

[0017] The portable computing device may be communicatively connected to a depth sensor and the method may include receiving depth data associated with the first and / or second image at the portable computing device.

[0018] The method may further include determining the at least one parameter for the first and / or second produce using the portable device, and verifying the determination using the remote device.

[0019] The remote device may be configured to determine the at least one parameter using a machine learning model.

[0020] The machine learning model used by the remote device may be different to the machine learning model used by the portable computing device.

[0021] The portable computing device may be configured to use a first machine learning model to determine the at least one parameter in the mounted operation mode and to use a second machine learning model to determine the at least one parameter in the second operation mode.

[0022] The second machine learning model may be trained to identify features that the first machine learning model is not trained to identify.

[0023] The features may include one or more parts of a user’s body. The second camera may be inbuilt into the portable computing device, and the mounted camera may be external to the portable computing device.

[0024] The mounted camera may be demountable such that it can be used in the second operation mode as the second camera.

[0025] Also disclosed is a system for monitoring produce, comprising: a portable computing device switchable between a mounted operation mode and a second operation mode, wherein the portable computing device is configured to be mounted to a mounting bracket in use in the mounted operation mode, and to be dismounted from the mounting bracket in use in the second operation mode, wherein the portable computing device is configured to receive an image of produce and to determine at least one parameter associated with the produce based on the received image, or to transmit the received image to a remote device and to receive a determination of the at least one parameter from the remote device.

[0026] The portable computing device may determine the at least one parameter using a machine learning model.

[0027] The system may further include a mounted camera communicatively connected to the portable computing device.

[0028] The mounted camera may be mounted to a part of a harvester or grader.

[0029] The portable computing device may be configured to control the mounted camera to capture images automatically in the mounted mode.

[0030] The system may further include a second camera, the portable computing device may be configured to control the second camera to capture an image in response to a user input in the second operation mode.

[0031] The mounting bracket may be installed on a harvester or grader.

[0032] The portable computing device may be communicatively connected to a depth sensor and the portable computing device may be further configured to receive depth data associated with the image.

[0033] The system may further include the remote device, the portable computing device may be configured to transmit the image from the portable computing device to the remote device, and the remote device may be configured to determine the at least one parameter. The remote device may be configured to determine the at least one parameter using a machine learning model.

[0034] The machine learning model used by the remote device may be different to the machine learning model used by the portable computing device.

[0035] The portable computing device may be configured to use a first machine learning model to determine the at least one parameter in the mounted operation mode and to use a second machine learning model to determine the at least one parameter in the second operation mode.

[0036] The second machine learning model may be trained to identify features that the first machine learning model is not trained to identify.

[0037] The features may include one or more parts of a user’s body.

[0038] The second camera may be inbuilt into the portable computing device, and the mounted camera may be external to the portable computing device.

[0039] The mounted camera may be demountable such that it can be used in the handheld operation mode as the second camera.

[0040] The method may further include mounting the second camera such that the second camera has at least part of a separate transporter within its field of view, such that produce transported by the separate transporter passes through the field of view of the second camera.

[0041] The second camera may be an inbuilt camera of the portable computing device.

[0042] The second camera may be external to the portable computing device, and may be connected to the portable computing device by a wired or wireless communications link.

[0043] The second operation mode may be a handheld operation mode in which the portable computing device is not mounted to a mounting bracket.

[0044] Also disclosed is a method of monitoring produce, comprising: mounting a portable computing device to a first mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the first mounting bracket and initiating a second operation mode; mounting the portable computing device to a second mounting bracket; capturing, after mounting the portable computing device to the second mounting bracket, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

[0045] Also disclosed is a method of monitoring produce, comprising: mounting a portable computing device to a mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the mounting bracket and initiating a second operation mode; capturing, using a handheld camera after demounting the portable computing device, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

[0046] Also disclosed is a system for monitoring produce, comprising: a portable computing device configured to be mounted to a mounting bracket of a harvester or grader and for communicating with a first camera mounted to the harvester or grader; and a second camera communicatively coupled to or part of the portable computing device and configured for use when the portable computing device is unmounted from the mounting bracket, wherein the first and second cameras are configured to capture respective images of produce for use in determining respective parameters associated with the produce.

[0047] The portable computing device may be configured to determine the respective parameters associated with the produce.

[0048] The second camera may be attachable to the portable computing device when the portable computing device is unmounted from the mounting bracket.

[0049] The system may further include the mounting bracket.

[0050] The system may further include the harvester or grader. The portable computing device may be a tablet computer.

[0051] The system may further include a transporter for produce separate from the harvester or grader, and the second camera may be configured to capture an image of produce on the separate transporter.

[0052] The second camera or portable computing device may include a depth sensor configured to generate depth data associated with the image captured by the second camera.

[0053] The portable computing device may be configured to be in a first mode of operation when mounted to the mounting bracket, and in a second mode of operation when unmounted from the mounting bracket, the first and second modes of operation being different modes of operation.

[0054] The portable computing device may be configured to determine the respective parameters and the different modes of operation determine the method used to determine the respective parameters.

[0055] Also disclosed is a case or protective housing for a portable computing device, the case or protective housing being configured to receive the portable computing device and including a first configuration for engagement by a mounting bracket to secure the case or protective housing to a harvester or grader, and a second configuration for engagement with a camera to secure the camera to the case or protective housing.

[0056] The first and second configurations may share one or more components.

[0057] Use of the first configuration to secure the case or protective housing to the harvester or grader may obstruct use of the second configuration to secure the camera to the case or protective housing.

[0058] The case or protective housing may be combined with one or more of the mounting bracket, the harvester or grader, and the camera.

[0059] The second camera may be mounted such that the second camera has at least part of a separate transporter within its field of view, such that produce transported by the separate transporter passes through the field of view of the second camera.

[0060] The second camera may be an inbuilt camera of the portable computing device.

[0061] The second camera may be external to the portable computing device, and may be connected to the portable computing device by a wired or wireless communications link. The second operation mode may be a handheld operation mode in which the portable computing device is not mounted to a mounting bracket.

[0062] Also disclosed is a system for monitoring produce, comprising: a first mounting bracket for mounting a portable computing device; a second mounting bracket for mounting the portable computing device; and the portable computing device, wherein the portable computing device is configured to operate in a mounted operation mode when mounted to the first mounting bracket and to receive, in the mounted operation mode, a first image, the first image being of first produce, wherein the portable computing device is configured to determine at least one parameter associated with the first produce based on the first image, wherein the portable computing device is configured to operate in a second operation mode when mounted to the second mounting bracket, wherein the portable computing device is configured in use to capture, when mounted to the second mounting bracket, a second image, the second image being of second produce; wherein the portable computing device is configured to receive the second image in the second operation mode; and wherein the portable computing device is configured to determine at least one parameter associated with the second produce based on the second image.

[0063] Also disclosed is a system for monitoring produce, comprising: a mounting bracket for mounting a portable computing device; a handheld camera; and the portable computing device, wherein the portable computing device is configured to operate in a mounted operation mode when mounted to the mounting bracket and to receive, in the mounted operation mode, a first image, the first image being of first produce, wherein the portable computing device is configured to determine at least one parameter associated with the first produce based on the first image, wherein the portable computing device is configured to operate in a second operation mode when demounted from the mounting bracket, wherein the portable computing device is configured in use to capture, using the handheld camera, a second image, the second image being of second produce; wherein the portable computing device is configured to receive the second image in the second operation mode; and wherein the portable computing device is configured to determine at least one parameter associated with the second produce based on the second image.

[0064] Also disclosed is a defect detection system for detecting one or more defects associated with one or more produce items, the system comprising: a camera module configured to generate an image associated with the one or more produce items; and a processing unit configured to receive the image and to determine one or more defect parameters associated with the or each produce item using a machine learning model.

[0065] The one or more defect parameters may include at least one of a presence or absence of a defect, a number of defects, or a type of defect. The one or more defect parameters may include at least a type of defect, and the type of defect includes one or more of mechanical damage, growth crack, discolouration, infestation, rot, secondary growth, scab, scurf, disease, bruising, and / or sprouting.

[0066] The discolouration may be greening.

[0067] The processing unit may be further configured to identify individual produce items in the image before determining the one or more defect parameters by segmenting the image into at least one portion which represents a single produce item.

[0068] The processing unit may be configured to pass a portion of the image corresponding to a single produce item to the machine learning model to determine the one or more defect parameters for the single produce item.

[0069] The processing unit may be configured not to pass a portion of the image that does not correspond to the single produce item to the machine learning model when determining the one or more defect parameters for that produce item.

[0070] The processing unit may be further configured to use the determined one or more defect parameters to determine a value of the produce item or a group of produce items.

[0071] The defect detection system may be configured to transfer the determined value from a first location to a second location automatically or in response to a user input.

[0072] Also disclosed is a method for detecting defects in one or more produce items, the method comprising: training a machine learning model to determine one or more defect parameters associated with one or more produce items; generating an image of one or more produce items using a camera module; and analysing the image with a processing unit, using the machine learning model, to determine the one or more defect parameters for the one or more produce items

[0073] BRIEF DESCRIPTION OF THE FIGURES

[0074] In order that the present disclosure may be more readily understood, preferable embodiments thereof will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0075] Fig. 1 is a schematic illustration of produce;

[0076] Fig. 2 is a schematic illustration of the produce of Fig. 1 after harvesting;

[0077] Fig. 3 is a schematic illustration of produce; Fig. 4 is a schematic illustration of a harvester harvesting the produce of Fig. 3;

[0078] Fig. 5 is a schematic illustration of a part of a produce monitoring system in a mounted operation mode;

[0079] Fig. 6 is a schematic illustration of a part of a produce monitoring system in a handheld operation mode;

[0080] Fig. 7 is a flowchart outlining a method of produce monitoring;

[0081] Fig. 8 is a schematic illustration of a part of a produce monitoring system;

[0082] Fig. 9 is a schematic illustration of a part of a produce monitoring system;

[0083] Fig. 10 is a schematic illustration of a part of a produce monitoring system;

[0084] Fig. 11 is a schematic illustration of a part of a produce monitoring system in a handheld operation mode;

[0085] Figs. 12-15 show images used for training a machine learning model; and Figs. 16-24 show examples of potatoes having defects.

[0086] DETAILED DESCRIPTION OF THE DISCLOSURE

[0087] The disclosed technology includes a monitoring system 1 and / or a monitoring method. The monitoring system 1 and / or method may be for monitoring produce 2, such as agricultural produce 2, and may be for determining one or more parameters associated with the produce 2. The one or more parameters may, therefore, be determined by the monitoring system 1 and / or method.

[0088] The produce 2 may include one or more produce items 21. The produce 2 may be or include agricultural produce 2, such as foodstuffs, for example fruits, vegetables, meat, or eggs. The produce items 21 may, therefore, include foodstuff items, such as individual fruits, vegetables, eggs, or cuts of meat. Some examples of produce items 21 are potatoes, onions, and apples. Each individual potato, onion or apple (for example) may be a produce item 21 , and a group or crop of individual produce items 21 may be referred to as the produce 2. The produce 2 may, therefore, include a group of produce items 21 (such as a group of potatoes).

[0089] The one or more parameters may be determined for the produce 2 as a whole and / or separately for each produce item 21. For example, the one or more parameters may include an average value for a given group of produce items 21 , which may be referred to as a crop, for example a crop of potatoes. The average value could be a mean, median, or mode average. The one or more parameters may include individual values associated with each produce item 21. For example, an average size for the produce 2 may be determined, and / or the size of each produce item 21 may be determined individually.

[0090] The parameters may include one or more of: a size of a produce item 21 or each produce item 21 ; an average size for a group of produce items 21 ; a length of a produce item 21 or each produce item 21 ; an average length for a group of produce items 21 ; a width of a produce item 21 or each produce item 21 ; an average width for a group of produce items 21 ; an area (such as a surface area or a cross-sectional area) of a produce item 21 or each produce item 21 ; an average area (such as a surface area or a cross-sectional area) for a group of produce items 21 ; a volume of a produce item 21 or each produce item 21 ; an average volume for a group of produce items 21 ; a weight of a produce item 21 or each produce item 21 ; an average weight for a group of produce item 21 ; a crop type of a produce item 21 or each produce item 21 ; a variety of a crop type of a produce item 21 or each produce item 21 ; a colour of a produce item 21 or each produce item 21 ; and / or a defect associated with a produce item 21 or each produce item 21. Any average parameters determined may be a mean, median, or mode average, and may in particular be a mean average.

[0091] Fig. 1 provides a schematic illustration of produce items 21 in the form of apples 21 on an apple tree.

[0092] Fig. 2 provides a schematic illustration of produce 2 in the form of produce items 21 , in this example apples, in a container s following harvesting.

[0093] Fig. 3 provides a schematic illustration of produce 2 in the form of produce items 21 , in this example potatoes, before harvesting.

[0094] Fig. 4 provides a schematic illustration of produce 2, and produce items 21 , being harvested by a harvester 13. The system 1 may include the harvester 13.

[0095] The harvester 13 may include a cab 134 for a user. The cab 134 may include operating means for the user to operate the harvester 13.

[0096] The harvester 13 may include a transporter 132 for transporting the produce 2, and this may be in the form of a conveyor, such as a conveyor belt, for example. The transporter 132 (e.g. conveyor) may include slats or other spaced members to allow debris to be separated from the produce 2.

[0097] The system 1 may include a trailer 131. The produce 2 may be loaded into the trailer 131 after harvesting.

[0098] The system 1 may include a grader, such as a potato grader in examples relating to potatoes. The grader may include a transporter for transporting the produce 2, and this may be in the form of a conveyor, such as a conveyor belt, for example. The transporter (e.g. conveyor) may include slats or other spaced members to allow debris to be separated from the produce 2. The grader may provide a picking area to enable workers to pick unwanted items (such as debris) from the transporter. The grader may include one or more rollers or sieves for cleaning and / or grading produce items 21 (e.g. for sorting produce items 21 according to their size). For example, the rollers or sieves may include one or more gaps sized to allow produce items 21 falling within a first size range to pass through the gap or gaps and to prevent produce items 21 falling within a second size range from passing through the gap.

[0099] The system 1 may include a portable computing device 11 (see e.g. Fig. 8). The portable computing device 11 may include one or more of a processor (also referred to as a processing unit) 111 , a memory 112 (which may be a non-transitory computer-readable memory), an imaging unit 113, such as a camera 113, a depth sensor 114, a power source 115, a communications module 116, and / or a display 117.

[0100] The portable computing device 11 may have a mounted operation mode and a handheld operation mode. The portable computing device 11 may be switchable between the mounted operation mode and the handheld operation mode, for example by interaction with a user interface of the portable computing device 11 (and / or the modes of operation may be switched automatically). In other words, the portable computing device 11 may be selectively operable in the mounted operation mode and the handheld operation mode. Automatic switching between modes may be achieved, for example, through the sensing by the portable computing device 11 of a connection to another system 1 component which is unique to the mounted operation mode or the handheld operation mode (such as a particular camera 14m / h, see below) and / or through the use of a motion sensor (such as an inertial measurement unit of the portable computing device 11) to determine a motion characteristic of the portable computing device 11 which is indicative of being mounted or handheld. The user interface may be displayed on the display 117. The display 117 may be a touch-sensitive display, and / or the portable computing device 11 may include input means such as a keyboard and / or mouse. The user may, therefore, interact with the portable computing device 11 using the user interface, and may control the operation of the portable computing device 11 using the user interface.

[0101] The system 1 may include a mounting bracket 12 (see e.g. Fig. 10). The mounting bracket 12 may be configured to releasably mount the portable computing device 11. The mounting bracket 12 may be installed on the harvester 13 or on the grader. For example, the mounting bracket 12 may be installed in the cab 134. The mounting bracket 12 may, therefore, mount the portable computing device 11 to a part of the harvester 13 or grader. The mounting bracket 12 may include one or more clamping members configured to secure the portable computing device 11 to the mounting bracket 12. For example, the mounting bracket 12 may include a pair of clamping members configured to clamp a part of the portable computing device 11. The system 1 may include a mounted camera 14m, which may be external to the portable computing device 11 . The mounted camera 14m may be a mounted external camera 14m, which may be mounted to a part of the harvester 13 or grader. The mounted camera 14m may be mounted to the harvester 13 or grader using a camera mount 133 (see e.g. Fig. 5). The camera mount 133 may be located such that, when mounted to the camera mount 133, the mounted camera 14m has a field of view that includes the transporter 132. The mounted camera 14m may, therefore, be configured to produce images of the produce 2 in use, and those images may be of the produce 2 in transit on the transporter 132. The mounted camera 14m may be communicatively connected to the portable computing device 11 , for example by a wired or wireless communications link. There may, therefore, be a wiring harness to communicatively connect the mounted camera 14m to the portable computing device 11 and connection to that wiring harness may be detectable by the portable computing device 11 for automatically determining the mode of operation, for example. Similarly, the identity of the camera 14m to which the portable computing device 11 is connected may be determined by the portable computing device 11 , and the identity of the camera 14m may be used for automatically determining the mode of operation. The identity of the camera 14m to which the portable computing device 11 is connected may be determined for either a wired or wireless connection, for example. The identity of the camera 14m may be determined through the use of an identification code for that camera 14m (which may be a unique identifier or an identifier common to a group of such cameras 14m). Images captured by the mounted camera 14m may, therefore, be sent to the portable computing device 11. The portable computing device 11 may be configured to receive images from the mounted camera 14m.

[0102] The portable computing device 11 may be communicatively connected to the mounted camera 14m in the mounted operation mode.

[0103] The system 1 may include a handheld camera 14h, which may be a handheld external camera 14h. In some versions the handheld camera 14h may be provided by an inbuilt camera 113 of the portable computing device 11. The handheld camera 14h may be releasably attached to the portable computing device 11 , for example by a camera mount attached to the portable computing device 11. The attachment of the handheld camera 14h to the portable computing device 11 may be direct or indirect (e.g. via attachment to a case of the portable computing device 11). The handheld camera 14h may be communicatively connected to the portable computing device 11 , for example by a wired or wireless communications link. In the case of a wired connection, this may be via a wiring harness which may be different to a wiring harness used to connect the portable computing device 11 to the mounted camera 14m (such that connection to either wiring harness may determine the mode of operation, for example). Similarly, as described above, the identity of the camera 14m, 14h , or 113 to which the portable computing device 11 is connected may be determined by the portable computing device 11 , and the identity of the camera 14m, 14h, 113 may be used for automatically determining the mode of operation.

[0104] In versions in which the portable computing device 11 is connected to more than one camera at the same time (for example in versions in which the portable computing device includes the inbuilt camera 113 for use in the handheld operation mode, and is connectable to the mounted camera 14m for use in the mounted mode), the portable computing device 11 may identify an active camera and the active camera may determine the mode of operation. The active camera may be the camera that is used to capture one or more images. For example, when the mounted camera 14m is determined to be the active camera, the portable computing device 11 may automatically enter the mounted operation mode. When the handheld camera 14h - which may be the inbuilt camera 113 - is determined to be the active camera, the portable computing device 11 may automatically enter the handheld operation mode.

[0105] In some versions a user may be able to select the mode of operation using the user interface and the user selection may override any automatic determination of the mode of operation. The user interface may also provide an option for the user to disable automatic determination of the mode of operation, for example.

[0106] The identity of the camera 14m, 14h , or 113 to which the portable computing device 11 is communicatively connected may be determined for either a wired or wireless connection, for example. The identity of the camera 14m, 14h, or 113 may be determined through the use of an identification code for that camera 14m, 14h , or 113 (which may be a unique identifier or an identifier common to a group of such cameras 14m, 14h , or 113). Images captured by the handheld camera 14h may, therefore, be sent to the portable computing device 11. The portable computing device 11 may be configured to receive images from the handheld camera 14h . The handheld camera 14h may be considered to be a handheld camera 14h if the camera is sized such that it can be carried by a user (e.g. in their hand or hands). The handheld camera 14h is an example of a portable camera and other portable cameras might be used, in some versions, in place of the handheld camera 14h . In some versions, the mounted camera 14m is a first camera and the handheld camera 14h is a second camera.

[0107] In some versions, the portable computing device 11 may be in the form of a tablet computer. In some versions, the portable computing device 11 may include a case or other protective housing. In some versions, the case or other protective housing may include mounting locations configured to be engaged by (or to engage) the camera mount to attach the handheld camera 14h to the portable computing device 11. In some versions, the camera mount includes one or more c- or L- shaped members configured to be rotated with respect to a part of the handheld camera 14h and into engagement with a part of the portable computing device 11 (e.g. a part of the case or other protective housing) to attach the handheld camera 14h to the portable computing device 11. With the portable computing device 11 and handheld camera 14h attached to each other, in some versions, a user may manipulate the portable computing device 11 (e.g. moving that device 11) to move the handheld camera 14h. In some versions, display 117 of the portable computing device 11 may be configured to face a user during operation with the handheld camera 14h attached thereto such that a field of view of the handheld camera 14h faces away from the user (e.g. in a direction opposite to the direction in which the display 117 is facing).

[0108] The portable computing device 11 may be communicatively connected to the mounted camera 14m in the mounted operation mode. The portable computing device 11 may be communicatively connected to the handheld camera 14h in the handheld operation mode. In some versions, the portable computing device may be connected to the mounted camera 14m and the handheld camera 14h simultaneously but only one such camera may be active (i.e. capturing images) at any one time. Accordingly, the mode of operation may be determined by the active camera (i.e. the mounted camera 14m may be active in the mounted operation mode and the handheld camera 14h may be active in the handheld operation mode). The user interface may enable the user to select a camera as the active camera and the user selection of the active camera may thus determine the mode of operation. For example the user interface may include an interactive element for switching between the mounted and handheld cameras 14m, 14h or may include a first interactive element associated with the mounted camera 14m and a second interactive element associated with the handheld camera 14h, and selection by the user of one of the interactive elements may thus activate a corresponding camera.

[0109] The mounted camera 14m may be demountable (e.g. from the camera mount 133) such that it is usable as the handheld camera 14h in the handheld mode. In such versions the user may select the mounted or handheld operation mode by interaction with the user interface. In some versions, separate cameras may be provided for the mounted and handheld modes. The portable computing device 11 may, therefore, switch between connections to the mounted camera 14m and the handheld camera 14h when the operation mode is switched between the mounted and handheld operation modes (e.g. by interaction of the user with the user interface).

[0110] In the handheld mode, the portable computing device 11 may be handheld by a user 4 (see e.g. Fig. 6 and 11). The handheld camera 14h , 113 may also be handheld in use in the handheld mode. This may allow the user 4 freedom to remove the portable computing device 11 from its mounting bracket 12, for example, and to capture images by directing the handheld camera 14h at any produce 2 to be imaged. This may allow greater freedom to image produce 2 in different locations, for example to capture images of samples of produce 2 as shown in Fig. 6 or to capture images of produce 2 that has not yet been harvested as shown in Fig. 11. In some versions, handheld use does not preclude the use of one or more straps to aid the user in carrying the portable computing device 11 and / or the handheld camera 14,113.

[0111] In some versions, the mounting bracket 12 may be configured to receive the portable computing device 11 including the case or other protective housing - such that this case or other protective housing does not need to be removed from the rest of the portable computing device 11 for the device 11 to be mounted to the mounting bracket 12.

[0112] In some versions, the case or other protective housing may include a battery or other portable power supply to assist in powering the portable computing device 11. In some versions, the portable computing device 11 may receive electrical power from the harvester 13 or grader when mounted in the mounting bracket 12. In some versions, the electrical power for the operation of the portable computing device 11 may be provided at least in part by the battery or other portable power supply when the portable computing device 11 is not mounted to the mounting bracket 12. The portable computing device 11 may include one or more further batteries or other portable power supplies (in addition to any provided as part of the case or other protective housing).

[0113] In some versions, the case or protective housing includes a first configuration for use in securing the case or protective housing to the mounting bracket 12 and a second configuration for use in securing the handheld camera 14h to the case or protective housing. The first and second configurations may include one or more respective arms, clamps, protrusions, recesses, or other members used in the securing of the case or protective housing to the mounting bracket 12 or handheld camera 14h . In some versions, the securing of the case or protective housing to the mounting bracket 12 may cover or otherwise obstruct access to at least part of the second configuration, such that - for example - the handheld camera 14h cannot be secured to the case or protective housing when the case or protective housing is secured to the mounting bracket 12. In some cases, the first and second configurations are the same configurations.

[0114] In summary, therefore, the portable computing device 11 may be switchable between the mounted operation mode, in which it is mounted to the mounting bracket 12, and for example mounted to a part of a harvester 13 or grader, and the handheld operation mode, in which the portable computing device 11 is released (and removed) from the mounting bracket 12 such that it can be carried by the user in a handheld manner.

[0115] In some versions the mounted camera 14m may be the inbuilt camera 113 of the portable computing device 11. In such versions, therefore, the mounting bracket 12 and the camera mount 133 may be the same entity. For example, the entire portable computing device 11 may be mounted to the camera mount 133, which may also function as the mounting bracket 12, such that the inbuilt camera 113 of the portable computing device 11 is configured to capture images of the produce 2 from its mounted position. As described, this may involve the transporter 132 being visible within a field of view of the camera 113.

[0116] In some versions, a mounting bracket (which may be similar to or identical to the mounting bracket 12 or 133) may be provided which is positioned relative to a separate transporter for transporting the produce 2 (e.g. a conveyor, such as a conveyor belt) which may be a standalone transporter or part of another system such as a processing system for the produce 2 (e.g. to clean and / or prepare the produce). The positioning may be such that the handheld camera 14h, 113 has at least part of the separate transporter within its field of view, such that produce transported by the separate transporter passes through the field of view. In some versions, the handheld camera 14h is separate from the portable computing device 11 but communicatively connected thereto through a wired or wireless connection, with the handheld camera 14h mounted to a further camera mount positioned with the separate transporter within its field of view. Thus, the portable computing device 11 may have a further mode of operation in which the portable computing device 11 has been demounted from the harvester 13 or grader but is, instead, mounted to other equipment (such as the separate transporter) or otherwise used in relation to that other equipment (e.g. with the handheld camera 14h mounted relative to the separate transporter). This further mode of operation may be a variation of the handheld mode of operation or may be a separate mode of operation (e.g. an other equipment mode of operation). In such a mode of operation, the portable computing device 11 may be handheld by the user 4, but the handheld camera 14h may be mounted relative to the separate transporter such that produce 2 transported by the separate transporter passes through its field of view. In versions where the handheld camera 14h is the inbuilt camera 113, the portable computing device 11 , including the inbuilt camera 113, may be mounted relative to the separate transporter such that produce 2 transported by the separate transporter passes through its field of view.

[0117] The mounting bracket 12 may include one or more arms or other members configured to clamp or lock the portable computing device 11 in place - the one or more arms or other members being releasable to allow removal of the portable computing device 11.

[0118] A method of monitoring produce may, therefore, include mounting the portable computing device 11 to the mounting bracket 12 and initiating the mounted operation mode. The mounted operation mode may be activated by the user 4 interacting with the user interface, which may be displayed on the display 117 (or may be automatically activated).

[0119] In the mounted operation mode, the portable computing device 11 may be configured to receive images, which may be images of produce 2, from the mounted camera 14m (which images may have been captured by the mounted camera 14m). Produce items 21 may, therefore, be visible in the images. The transporter 132 may also be visible in such images. The portable computing device 11 may be configured to control the mounted camera 14m in the mounted operation mode. The portable computing device 11 may be configured to control the mounted camera 14m to capture images automatically in the mounted mode. The automatic image capture may be triggered by variations in light levels, for example, or activated by motion (e.g. detected by use of the mounted camera 14m). The portable computing device 11 may control the mounted camera 14m to continue capturing images at a predetermined frequency until a deactivation criterion is met, such as when no motion is detected, for example, or when a light level criterion is met.

[0120] The portable computing device 11 may be configured to determine at least one parameter associated with the produce 2. The at least one parameter may be determined based on the image(s) of the produce 2. The image(s) may, for example, be received using the communications module 116, and may be stored (temporarily or permanently) in the memory 112, which may be a non-transitory computer-readable memory.

[0121] The at least one parameter may include one or more of size, weight, colour, quality, or health, for example. The parameter may be associated with the produce 2 as a whole, for example as an average value, or may be determined individually for individual produce items 21.

[0122] The determination of the at least one parameter may be performed using the processor 111.

[0123] The parameter may be determined using a machine learning model. The machine learning model may be trained, in a training phase, using training data. The training data may be or include ground truth data, which may be annotated with labels or values provided by humans, for example. For example, the ground truth data may include images of produce 2 with associated parameters identified for each image. The associated parameters may form metadata, for example. The ground truth data may, therefore, include a collection of images of produce 2 for which the parameter to be determined has been established manually. The machine learning model may then learn from the training data so that it can determine the one or more parameters for produce 2 in an operative phase.

[0124] In the operative phase, the machine learning model may determine at least one parameter associated with the produce 2 based on the image or images received at the portable computing device 11 , which may be images from the inbuilt camera 113, the mounted camera 14m, or the handheld camera 14h.

[0125] The parameter determined by the portable computing device 11 may be displayed on the display 117 or transmitted to a remote device using the communications module 116. The portable computing device 11 may be demounted from the mounting bracket 12. The portable computing device 11 may be switched from the mounted operation mode to the handheld operation mode.

[0126] In the handheld operation mode (which may include the other equipment mode), the portable computing device 11 may be configured to control the handheld camera 14h, and optionally to capture an image in response to a user input. The user input may be via the user interface, for example, such as by interacting with a “capture” button.

[0127] The image capture in the handheld mode may, therefore, be triggered manually, whereas the image capture in the mounted mode may be automatic. In some versions, the image capture may be automatic in the handheld mode, as described for the mounted mode.

[0128] The user 4 may, therefore, use the portable computing device 11 and handheld camera 14h to capture an image of produce 2 in the handheld mode, such as is illustrated in Figs. 6 and 11. The produce 2 may be harvested produce, such as in Fig. 6, or may be unharvested produce, such as in Fig. 11. The produce 2 may be a sample of a larger crop, for example. The parameters determined by the portable computing device 11 for the sample can then be used to estimate corresponding parameters for the larger crop. For example, a sample of a crop taken from one part of a field could be used to estimate corresponding parameters for the crop for the whole field.

[0129] The image captured by the handheld camera 14h may be sent to the portable computing device 11. Accordingly, the image may be received by the portable computing device 11.

[0130] The portable computing device 11 may then determine the at least one parameter associated with the produce 2 based on the captured image.

[0131] As described, this may involve the use of a machine learning model.

[0132] However, in some versions, the machine learning model used in the mounted operation mode may differ from the machine learning model used in the handheld operation mode. For example, the handheld mode machine learning model may be trained to identify features that the mounted mode machine learning model is not trained to identify. Those features may include a part of a user’s body, such as a user’s feet (which may include identification of shoes, for example). The features that the handheld mode machine learning model is additionally trained to identify may be features that are not expected to be visible in the mounted mode (such as a user’s feet or thumb / finger). These features may be excluded from the determining of at least one parameter, as they are not produce items 21. In some versions, when in the handheld operation mode a part of the captured image(s) may be excluded from analysis to identify produce items 21 , this part of the image(s) representing the region closest to the user for example, to reduce the risk of parts of the user (e.g. their feet) from being incorrectly identified as produce items 21 .

[0133] The portable computing device 11 may, therefore, be used to determine at least one parameter for first produce 2 in the mounted operation mode, and to determine at least one parameter for second produce 2 in the handheld operation mode. The first produce 2 may include one or more produce items 21 visible in an image captured by the mounted camera 14m, for example, and the second produce 2 may include one or more produce items 21 visible in an image captured by the handheld camera 14h (which may be different produce items 21 to those visible in the image captured by the mounted camera 14m).

[0134] In this manner, the portable computing device 11 may be used to determine the at least one parameter for produce 2 both during harvesting, for example when mounted to the mounting bracket 12, and before harvesting, for example using the handheld mode. The produce monitoring system 1 is, therefore, more versatile than previous systems, which can be confined to operation only when mounted to a harvester 13 or grader, and cannot be demounted for operation in a handheld mode.

[0135] The mounted camera 14m and / or handheld camera 14h may include a depth sensor, which may produce depth data associated with the captured images. The depth data may be used by the portable computing device 11 to determine the parameter associated with the produce 2, in addition to the received image or images. In some versions the depth sensor may be a depth sensor 114 that is inbuilt into the portable computing device 11 . Each image received by the portable computing device 11 may, therefore, be received in conjunction with corresponding depth data, which may be used by the portable computing device 11 (e.g. by the machine learning model) to determine the at least one parameter. The machine learning model may, therefore, be trained using ground truth depth data associated with training images.

[0136] The depth sensor may include, for example, a sensor such as a LiDAR sensor. In some versions, the depth sensor is part of the mounted camera 14m and / or handheld camera 14h which may each include a respective pair of spaced apart camera devices to provide parallax information from which depth data can be determined. In some versions, the mounted camera 14m and / or handheld camera 14h may include a projector which is configured to project a speckle pattern of infrared light onto the produce 2, from which the camera(s) 14m, h may determine the depth data.

[0137] The mounted camera 14m and the handheld camera 14h may use the same technology as each other or may use different technologies. The mounted camera 14m and the handheld camera 14h may be the same type of camera, for example. In some instances, the handheld camera 14h may have different imaging characteristics (e.g. resolution) compared to the mounted camera 14m.

[0138] The system 1 may include a remote device 5 (see Fig. 9), which may be cloud-based. The portable computing device 11 may be configured to send data to the remote device 5. The portable computing device 11 may, therefore, be configured to send data to the cloud. The remote device 5 may be considered to be a distributed remote device 5 (i.e. the physical components of the remote device 5 may be distributed across a plurality of locations). This may be the case for a cloud-based remote device 5, for example. Examples of components of the remote device 5 that may be distributed across a plurality of locations include non-transitory computer readable storage media (e.g. a first storage medium may be located in a first location and a second storage medium may be located in a second location) and / or processing units 51 (e.g. a first processing unit may be located in a first location and a second processing unit may be located in a second location). Particular components of the remote device 5 to be used at any particular time may be allocated (e.g. by a control module of the remote device 5) based on available capacity. Accordingly, the remote device 5 may include one or more of a processing unit 51 , non-transitory computer- readable storage medium 52, and communications module 53 (which may be configured to send and / or receive data, such as to or from the portable computing device 11).

[0139] The portable computing device 11 may be configured to send images received from the cameras 14m, 14h, 113 and / or associated depth data to the remote device 5. The portable computing device 11 may be configured to send the determined parameters to the remote device 5.

[0140] The remote device 5 may be configured to determine one or more parameters associated with the produce 2 based on the data received from the portable computing device 11 , such as the image or images and / or associated depth data. The parameter or parameters determined by the remote device 5 may be the same as those determined by the portable computing device 11 , or may be different. The remote device 5 may be configured to determine the same parameters as the portable computing device 11 , and to determine additional parameters as well.

[0141] The parameters determined by the remote device 5 for the produce 2 may be sent back to the portable computing device 11. The parameter or parameters determined by the remote device 5 may be displayed on the display 117.

[0142] The remote device 5 may be configured to determine the one or more parameters using a machine learning model, which may be trained as described previously.

[0143] In some versions, the machine learning model used by the remote device 5 to determine the one or more parameters may be different to the machine learning model used by the portable computing device 11. For example, the machine learning model used by the remote device 5 may require more computing power, and may be more sophisticated than that used by the portable computing device 11. The portable computing device 11 may, therefore, use a relatively simple machine learning model to determine the one or more parameters, whereas the remote device 5 may use a relatively complex machine learning model to determine the one or more parameters.

[0144] The machine learning model used by the remote device 5 may be trained using a larger dataset than the model used by the portable computing device 11. The machine learning model used by the remote device 5 may be updated more frequently than the model used by the portable computing device 11. In some versions, the machine learning model used by the portable computing device 11 may not be updated, and the remote device 5 may therefore use a newer machine learning model to determine the one or more parameters, which may be more effective than the model used by the portable computing device 11.

[0145] The portable computing device 11 may, therefore, be used to determine an initial estimate of the one or more parameters associated with the produce 2, and the remote device 5 may verify and / or improve upon the estimate provided by the portable computing device 11 .

[0146] In some versions the portable computing device 11 may not determine the one or more parameters. The one or more parameters may be determined by the remote device 5 as described (e.g. the one or more images received by the portable computing device 11 may be sent to the remote device 5, which may be configured to determine the one or more parameters (e.g. using a machine learning model) as described previously). The one or more parameters determined by the remote device 5 may be sent to the portable computing device 11 or to another device (e.g. to a user device such as a smartphone).

[0147] The determination of the parameters may, therefore, be performed solely by the remote device 5, and the determined parameters may then be sent to a user device such as the portable computing device 11 for display to the user. In such versions, therefore, a portable computing device 11 having relatively weak computing power may be used, and relatively computationally intensive tasks may be performed by the remote device 5.

[0148] In some versions the one or more determined parameters (whether determined by the portable computing device 11 , the remote device 5, or both) may be stored in a remotely accessible storage medium, and the one or more determined parameters may be uploaded from the portable computing device 11 or remote device 5 to the remotely accessible storage medium. The remotely accessible storage medium may be accessible using a web portal, for example, which may require a user login or other identifying user characteristics in order to access the one or more determined parameters. In such a manner, the determined one or more parameters may be accessible by multiple users in multiple locations, for example.

[0149] In some versions the one or more images of the produce 2 (or produce item or items 21) may be processed before being input to the machine learning model (regardless of whether the machine learning model is implemented using the portable computing device 11 or the remote device 5), and this may occur in the training and / or operation of the machine learning model.

[0150] The processing of the or each image may be performed using the processor or processing unit 111 of the portable computing device 11 , a processing unit of the remote device 5, or a processing unit of another device to which the images are sent.

[0151] In particular, the or each image may be segmented into at least one portion determined to represent a single produce item 21. The segmentation may include, for example, thresholding, clustering, histogram-based methods, edge detection, region growing, and / or the like. In some versions the segmentation may be performed by a machine learning model trained to identify produce items 21. The image may therefore be segmented into a number of portions at least corresponding to the number of produce items 21 (and optionally a further portion that does not correspond to any produce items 21).

[0152] As illustrated in Fig. 12, an image captured by a camera of the system 1 (e.g. the camera 14m, 14h, or 113) may include one or more produce items 21. The image may be an RGB image, for example. In some versions the image may be a 2D image captured by a single camera. In other versions the image may include depth data as described previously, or may be a composite image generated by combining a plurality of images from a plurality of cameras, for example.

[0153] As illustrated in Fig. 13, the image may be segmented (e.g. by the processing unit 111 , the processing unit 51 of the remote device 5, or another processing unit) into one or more portions, each representing a single produce item 21. In some versions more than one produce item 21 may be present in an image segment.

[0154] As illustrated in Fig. 14, the segmented image portions corresponding to produce items 21 may be isolated from the original image to generate an isolated image including one or more produce items 21. The isolated image may include only the portions of the original image determined to correspond to produce items 21 , and may not include any portions of the original image determined not to correspond to a produce item 21 , as illustrated in Fig. 14.

[0155] As illustrated in Fig. 15, the isolated image may be divided into one or more portions each corresponding to a single produce item 21 , to generate one or more divided images. The number of divided images may correspond to the number of produce items 21 determined to be present in the original image.

[0156] The machine learning model configured to determine the one or more parameters may receive as an input the divided image or images. Accordingly, the image input to the machine learning model may include a single produce item 21. The processing unit 111 or 51 may therefore be configured to pass a portion of the image corresponding to a single produce item 21 to the machine learning model to determine the one or more defect parameters for the single produce item 21. Similarly, the processing unit 111 , 51 may be configured not to pass a portion of the image that does not correspond to the single produce item 21 to the machine learning model when determining the one or more defect parameters for that produce item 21. This may improve the accuracy of the determination of the one or more parameters for the produce item(s) 21.

[0157] Accordingly, the segmentation process described herein may be used to train the machine learning model using training data comprising a group of images each depicting a single produce item 21. Similarly, the segmentation process described herein may be used in operation to provide as an input to the machine learning model an image depicting a single produce item 21 for the determination of the one or more parameters.

[0158] In some versions the machine learning model may be trained to determine the one or more parameters for each produce item 21 based on an image including a plurality of produce items 21.

[0159] As described previously, the one or more parameters determined for the or each produce item 21 may include one or more of: a size of a produce item 21 or each produce item 21 ; an average size for a group of produce items 21 ; a length of a produce item 21 or each produce item 21 ; an average length for a group of produce items 21 ; a width of a produce item 21 or each produce item 21 ; an average width for a group of produce items 21 ; an area (such as a surface area or a cross- sectional area) of a produce item 21 or each produce item 21 ; an average area (such as a surface area or a cross-sectional area) for a group of produce items 21 ; a volume of a produce item 21 or each produce item 21 ; an average volume for a group of produce items 21 ; a weight of a produce item 21 or each produce item 21 ; an average weight for a group of produce item 21 ; a crop type of a produce item 21 or each produce item 21 ; a variety of a crop type of a produce item 21 or each produce item 21 ; a colour of a produce item 21 or each produce item 21 ; and / or a defect associated with a produce item 21 or each produce item 21 .

[0160] The machine learning model may be trained, as described previously, to determine the one or more parameters. The accuracy of the model may be determined by n-fold cross-validation, such as 5-fold or 10-fold cross validation, for example. The size (or average size) parameter may be determined by categorising the or each produce item 21 into a size category or size band, for example, and the output of the size parameter may include a determined category or band for the or each produce item 21. Additionally or alternatively, the size parameter may be provided as a measurement, for example in mm.

[0161] The length (or average length) parameter may be determined as a longest axis of the produce item 21 , for example as a longest axis passing through a central point of the produce item 21 , which may be a longitudinal axis. The length parameter may be provided as a measurement in mm for example.

[0162] The width (or average width) parameter may be determined as a dimension of the produce item 21 measured perpendicular to the length of the produce item 21 (e.g. perpendicular to the longitudinal axis of the produce item 21 ). The width parameter may be provided as a measurement in mm for example.

[0163] The area (or average area) parameter may be determined as a surface area or a cross-sectional area of the produce item 21 , and may be provided as a measurement in square millimetres for example.

[0164] The volume (or average volume) parameter may be determined as the volume occupied by the produce item 21 , and may be provided as a measurement in cubic millimetres for example.

[0165] The weight (or average weight) parameter may be determined as the expected weight of the produce item 21 , and may be provided as a measurement in grams, for example. A predetermined density for the produce item 21 may therefore be used to predict the weight of the produce item 21 using one or more of its dimensions or measurements (e.g. its volume). It will be appreciated that the machine learning model may be trained to predict the weight of the produce item 21 from image data using ground truth data and so a density or volume calculation may not be needed.

[0166] The crop type parameter of the produce item 21 may include a species of the crop or a common name by which the crop is known in the relevant technical field, for example “potato”, “apple”, or onion .

[0167] The variety of the crop type parameter may include a variety of the produce item 21 , which may be a subset of the crop type, and may be a more specific name for the produce item 21 compared to the crop type parameter. Some examples of varieties include “Maris Piper” in the case of potatoes and “Braeburn” in the case of apples. The colour parameter may represent an average colour of the produce item 21 , for example provided as an RGB or HSV value or by categorisation into a qualitative category, such as “green”, “red”, and so on.

[0168] The one or more defect parameters may include at least one of a presence or absence of a defect, a number of defects, or a type of defect.

[0169] The presence or absence of a defect may be a simple binary determination as to whether any defects are present in the produce item 21 (which may be regardless of the total number of defects identified). The presence or absence of a defect may be provided by categorisation of the produce item 21 into a first or second category, such as a “healthy” or “defective” category.

[0170] The number of defects may be provided as a count of the total number of defects identified for the produce item 21 , which may be a count of the number of different types of defect identified for the produce item 21.

[0171] The type of defect may include one or more of mechanical damage, growth crack, discolouration, infestation, rot, secondary growth, scab, scurf, disease, bruising, and / or sprouting. The produce item 21 may therefore be categorised into one or more categories corresponding to the identified defect types. In some versions more than one defect type may be identified for a single produce item 21 , and the produce item 21 may therefore be categorised into a corresponding number of categories (which may be equal to the number of different types of defect identified).

[0172] Mechanical damage may be identified as a defect corresponding to damage of the produce item 21 by mechanical impact (such as by a spade during harvesting, for example). Examples of mechanical damage in potatoes are illustrated in Fig. 16.

[0173] A growth crack may be identified as a crack, split or cleft in the produce item 21 formed naturally during the growth of the produce item 21 . Examples of growth cracks in potatoes are illustrated in Fig. 17.

[0174] Discolouration may be identified as the presence of an unwanted colour on the produce item 21 (and the unwanted colour may be set by a user, for example it may be identified in ground truth training data for the machine learning model). For example, the discolouration may be greening (i.e. the presence of a green colour on a part of the produce item 21 where this colour is undesirable). For example, green potatoes are generally unwanted. Examples of potatoes with greening are shown in Fig. 18. Infestation may be identified as the presence of one or more pests (such as insects, worms, etc.).

[0175] Examples of potatoes suffering from infestation are shown in Fig. 19.

[0176] Rot may be identified as an indication that the produce item 21 includes a rotten portion. Examples of rotten potatoes are shown in Fig. 20.

[0177] Secondary growth may be identified as a misshapen produce item 21 having a main body and a secondary portion extending from the main body. Examples of potatoes with secondary growth are shown in Fig. 21.

[0178] Scab may be identified as the presence of one or more scabs or scabby patches on the surface of the produce item 21 . Examples of potatoes with scab are shown in Fig. 22.

[0179] Scurf may be identified as the presence of a scurf disease on the produce item 21 (for example, potato black scurf).

[0180] Disease may be identified as the presence of one or more diseases in the produce item 21 as indicated by the presence of one or more symptoms in the image.

[0181] Bruising may be identified as the presence of one or more bruises on the surface of the produce item 21 . Examples of bruised potatoes are shown in Fig. 23.

[0182] Sprouting may be identified as the presence of one or more sprouts emerging from the produce item 21 . Examples of sprouting potatoes are shown in Fig. 24.

[0183] Training data exhibiting the defects to be identified may be used to train the machine learning model to identify the relevant defects (e.g. as ground truth data). In particular, the training data may include images of produce items 21 in which a single defect is present (and appropriately identified in the training data). In some versions the training data may include images of produce items 21 having a plurality of defects in which each defect is correspondingly identified.

[0184] The machine learning model may be trained to output the one or more defect parameters either as a qualitative output (e.g. yes / no categorisation of each defect being present or absent) or as a quantitative output (e.g. providing a confidence level, for example as a percentage, that a given defect is present in the produce item 21).

[0185] The machine learning model may be trained to identify a plurality of defects for a single produce item 21 . In other words, for a single image, the machine learning model may be trained to determine a plurality of defect parameters (such as the presence or absence of a plurality of the defects described herein).

[0186] The system 1 may include a first machine learning model configured to determine a first of the one or more parameters disclosed herein, and a second machine learning model configured to determine a second of the one or more parameters disclosed herein. The second machine learning model may in particular be configured to identify secondary growth. The second machine learning model may be trained using training data depicting secondary growth on produce items 21. In some versions, the first machine learning model may use as an input the segmented images described herein, and the second machine learning model may use as an input the original image (unsegmented) image received from the camera 14m, 14h, 113. The second machine learning model may therefore provide a separate identification of the second parameter (such as the secondary growth).

[0187] The machine learning model may be configured to categorise, for each type of defect identified, an extent of the defect. For example, the machine learning model may be configured to categorise the produce item 21 into a category depending on the extent of the defect identified. In other words, for a single identified defect (e.g. greening), the extent of the defect may be determined and the produce item 21 may be categorised depending on the determined extent. For example, the categories may be low, medium, or high.

[0188] The determination of the one or more defect parameters may be performed by the remote device 5 (e.g. by the processing unit 51 thereof), which, as described, may be a cloud-based device. The image or images of the produce item or items 21 may therefore be sent from the portable device 11 to the remote device 5 (e.g. to the cloud) for determination of the one or more defect parameters (and thus relatively computationally intensive tasks may be carried out by the remote device 5).

[0189] The determined parameter or parameters may be displayed to the user.

[0190] In some versions the portable computing device 11 and / or remote device 5 may be configured to determine a value (such as a monetary value) for a produce item 21 or group of produce items 21 , and the determined value may be based on the determined parameters, such as the determined defect parameters and / or the determined size. For example, the portable computing device 11 and / or remote device 5 may be configured to assign a value to the produce item 21 or group of produce items 21 based on the number and / or type of defects determined. In some cases the value may be determined for a single produce item 21 based on the determined parameters, while in other cases the value may be determined for a group of produce items 21 . For example, a sample or subset of produce items 21 may be used to determine a value for a larger group or set of produce items 21 represented by the sample (such as for a field of produce items 21 represented by a sample of that field). The value of the produce items 21 determined based on the determined parameters may therefore be extrapolated to determine a value for a larger group of produce items 21 for which images of each produce item 21 are not available.

[0191] The portable computing device 11 and / or remote device 5 may be configured to transfer the determined value (e.g. as a monetary amount) from a first location to a second location (e.g. from a first account to a second account automatically or in response to a user input. For example, the determined value may be displayed to a user, for example via the display 117, and the user may be presented with an option to accept or reject the determined value (e.g. via the user interface of the portable device 11). If the user accepts the determined value, the determined amount may be transferred from the first location to the second location (e.g. to the user’s bank account). In some cases the value may be determined and transferred automatically, for example without requiring a user input to accept the determined value.

[0192] In some versions the parameter determination may be performed using a computing device or system that is not switchable between a mounted and handheld mode of operation. For example, the parameter determination (such as the defect determination) may be carried out using a mounted system including a mounted computing device and associated mounted camera (e.g. mounted to a harvester or grader) that has only a single mode of operation. In other words, the disclosed parameter determination processes (e.g. the use of the disclosed machine learning models, for example for defect detection) are not related or inextricably linked to the provision of a computing device or system that is switchable between a first and second mode of operation (such as a mounted and handheld mode of operation).

[0193] As will be appreciated, whilst the term “handheld” has been used herein, other terms may be used in relation to some versions in which the feature or mode is not necessarily “handheld” by a user. In some versions, the handheld features or modes may be referred to as second features or second modes, for example. Likewise, where “handheld” has been used, “mobile” may be used instead.

[0194] When used in this specification and claims, the terms "comprises" and "comprising" and variations thereof mean that the specified features, steps or integers are included. The terms are not to be interpreted to exclude the presence of other features, steps or components.

[0195] The invention may also broadly consist in the parts, elements, steps, examples and / or features referred to or indicated in the specification individually or collectively in any and all combinations of two or more said parts, elements, steps, examples and / or features. In particular, one or more features in any of the embodiments described herein may be combined with one or more features from any other embodiment(s) described herein. Protection may be sought for any features disclosed in any one or more published documents referenced herein in combination with the present disclosure. Although certain example embodiments of the invention have been described, the scope of the appended claims is not intended to be limited solely to these embodiments. The claims are to be construed literally, purposively, and / or to encompass equivalents.

Claims

CLAIMS1. A method of monitoring produce, comprising: mounting a portable computing device to a mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the mounting bracket and initiating a second operation mode; capturing, after demounting the portable computing device, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

2. A method according to claim 1 , wherein the portable computing device determines the at least one parameter using a machine learning model.

3. A method according to any preceding claim, wherein the portable computing device is communicatively connected to a mounted camera in the mounted operation mode.

4. A method according to claim 3, wherein the mounted camera is mounted to a part of a harvester or grader.

5. A method according to claim 3 or 4, wherein the portable computing device is configured to control the mounted camera, and wherein the method includes the portable computing device capturing images automatically in the mounted operation mode.

6. A method according to any preceding claim, wherein the portable computing device is communicatively connected to a second camera in the second operation mode.

7. A method according to claim 6, wherein the portable computing device is configured to control the second camera, and wherein the method includes capturing an image in response to a user input in the second operation mode.

8. A method according to any preceding claim, wherein the mounting bracket is installed on a harvester or grader.

9. A method according to any preceding claim, wherein the portable computing device is communicatively connected to a depth sensor and the method includes receiving depth data associated with the first and / or second image at the portable computing device.

10. A method according to any preceding claim, including determining the at least one parameter for the first and / or second produce using the portable device, and verifying the determination using the remote device.

11. A method according to any preceding claim, wherein the remote device is configured to determine the at least one parameter using a machine learning model.

12. A method according to claim 11 when dependent on claim 2, wherein the machine learning model used by the remote device is different to the machine learning model used by the portable computing device.

13. A method according to any preceding claim when dependent on claim 2, wherein the portable computing device is configured to use a first machine learning model to determine the at least one parameter in the mounted operation mode and to use a second machine learning model to determine the at least one parameter in the second operation mode.

14. A method according to claim 13, wherein the second machine learning model is trained to identify features that the first machine learning model is not trained to identify.

15. A method according to claim 14, wherein the features include one or more parts of a user’s body.

16. A method according to claim 3 or any of claims 4-15 when dependent on both claims 3 and 6, wherein the second camera is inbuilt into the portable computing device, and the mounted camera is external to the portable computing device.

17. A method according to claim 3 or any of claims 4-15 when dependent on both claims 3 and 6, wherein the mounted camera is demountable such that it can be used in the second operation mode as the second camera.

18. A system for monitoring produce, comprising:a portable computing device switchable between a mounted operation mode and a second operation mode, wherein the portable computing device is configured to be mounted to a mounting bracket in use in the mounted operation mode, and to be dismounted from the mounting bracket in use in the second operation mode, wherein the portable computing device is configured to receive an image of produce and to determine at least one parameter associated with the produce based on the received image, or to transmit the received image to a remote device and to receive a determination of the at least one parameter from the remote device.

19. A system according to claim 18, wherein the portable computing device determines the at least one parameter using a machine learning model.

20. A system according to claim 18 or 19, further including a mounted camera communicatively connected to the portable computing device.21 . A system according to claim 20, wherein the mounted camera is mounted to a part of a harvester or grader.

22. A system according to claim 20 or 21 , wherein the portable computing device is configured to control the mounted camera to capture images automatically in the mounted mode.

23. A system according to any of claims 18-22, further including a second camera, wherein the portable computing device is configured to control the second camera to capture an image in response to a user input in the second operation mode.

24. A system according to any of claims 18-23, wherein the mounting bracket is installed on a harvester or grader.

25. A system according to any of claims 18-24, wherein the portable computing device is communicatively connected to a depth sensor and the portable computing device is further configured to receive depth data associated with the image.

26. A system according to any of claims 18-25, further including the remote device, wherein the portable computing device is configured to transmit the image from the portable computing device to the remote device, and the remote device is configured to determine the at least one parameter.

27. A system according to claim 26, wherein the remote device is configured to determine the at least one parameter using a machine learning model.

28. A system according to claim 27 when dependent on claim 19, wherein the machine learning model used by the remote device is different to the machine learning model used by the portable computing device.

29. A system according to any of claims 18-28 when dependent on claim 19, wherein the portable computing device is configured to use a first machine learning model to determine the at least one parameter in the mounted operation mode and to use a second machine learning model to determine the at least one parameter in the second operation mode.

30. A system according to claim 29, wherein the second machine learning model is trained to identify features that the first machine learning model is not trained to identify.31 . A system according to claim 30, wherein the features include one or more parts of a user’s body.

32. A system according to claim 23 when dependent on claim 20, wherein the second camera is inbuilt into the portable computing device, and the mounted camera is external to the portable computing device.

33. A system according to claim 23 when dependent on claim 20, wherein the mounted camera is demountable such that it can be used in the handheld operation mode as the second camera.

34. A method according to claim 6 or any of claims 7-17 when dependent on claim 6, further including mounting the second camera such that the second camera has at least part of a separate transporter within its field of view, such that produce transported by the separate transporter passes through the field of view of the second camera.

35. A method according to claim 34, wherein the second camera is an inbuilt camera of the portable computing device.

36. A method according to claim 34, wherein the second camera is external to the portable computing device, and is connected to the portable computing device by a wired or wireless communications link.

37. A method according to any of claims 1-17, wherein the second operation mode is a handheld operation mode in which the portable computing device is not mounted to a mounting bracket.

38. A method of monitoring produce, comprising:mounting a portable computing device to a first mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the first mounting bracket and initiating a second operation mode; mounting the portable computing device to a second mounting bracket; capturing, after mounting the portable computing device to the second mounting bracket, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

39. A method of monitoring produce, comprising: mounting a portable computing device to a mounting bracket and initiating a mounted operation mode; receiving, in the mounted operation mode, a first image, the first image being of first produce, at the portable computing device; determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the first produce based on the first image; demounting the portable computing device from the mounting bracket and initiating a second operation mode; capturing, using a handheld camera after demounting the portable computing device, a second image, the second image being of second produce; receiving, in the second operation mode, the second image at the portable computing device; and determining, with the portable computing device or a remote device communicatively coupled to the portable computing device, at least one parameter associated with the second produce based on the second image.

40. A system for monitoring produce, comprising:a portable computing device configured to be mounted to a mounting bracket of a harvester or grader and for communicating with a first camera mounted to the harvester or grader; and a second camera communicatively coupled to or part of the portable computing device and configured for use when the portable computing device is unmounted from the mounting bracket, wherein the first and second cameras are configured to capture respective images of produce for use in determining respective parameters associated with the produce.41 . A system according to claim 40, wherein the portable computing device is configured to determine the respective parameters associated with the produce.

42. A system according to claim 40 or 41 , wherein the second camera is attachable to the portable computing device when the portable computing device is unmounted from the mounting bracket.

43. A system according to any of claims 40-42, further including the mounting bracket.

44. A system according to claim 43, further including the harvester or grader.

45. A system according to any of claims 40-44, wherein the portable computing device is a tablet computer.

46. A system according to any of claims 40-45, further including a transporter for produce separate from the harvester or grader, wherein the second camera is configured to capture an image of produce on the separate transporter.

47. A system according to any of claims 40-46, wherein the second camera or portable computing device includes a depth sensor configured to generate depth data associated with the image captured by the second camera.

48. A system according to an of claims 40-47, wherein the portable computing device is configured to be in a first mode of operation when mounted to the mounting bracket, and in a second mode of operation when unmounted from the mounting bracket, the first and second modes of operation being different modes of operation.

49. A system according to claim 48, wherein the portable computing device is configured to determine the respective parameters and the different modes of operation determine the method used to determine the respective parameters.

50. A case or protective housing for a portable computing device, the case or protective housing being configured to receive the portable computing device and including a first configuration forengagement by a mounting bracket to secure the case or protective housing to a harvester or grader, and a second configuration for engagement with a camera to secure the camera to the case or protective housing.51 . The case or protective housing according to claim 50, wherein the first and second configurations share one or more components.

52. The case or protective housing according to any of claims 50-51 , wherein use of the first configuration to secure the case or protective housing to the harvester or grader obstructs use of the second configuration to secure the camera to the case or protective housing.

53. The case or protective housing according to any of claims 50-52 in combination with one or more of the mounting bracket, the harvester or grader, and the camera.

54. A system according to claim 23 or any of claims 24-33 when dependent on claim 23, wherein the second camera is mounted such that the second camera has at least part of a separate transporter within its field of view, such that produce transported by the separate transporter passes through the field of view of the second camera.

55. A system according to claim 54, wherein the second camera is an inbuilt camera of the portable computing device.

56. A system according to claim 54, wherein the second camera is external to the portable computing device, and is connected to the portable computing device by a wired or wireless communications link.

57. A system according to any of claims 18-33 or 54-56, wherein the second operation mode is a handheld operation mode in which the portable computing device is not mounted to a mounting bracket.

58. A system for monitoring produce, comprising: a first mounting bracket for mounting a portable computing device; a second mounting bracket for mounting the portable computing device; and the portable computing device, wherein the portable computing device is configured to operate in a mounted operation mode when mounted to the first mounting bracket and to receive, in the mounted operation mode, a first image, the first image being of first produce, wherein the portable computing device is configured to determine at least one parameter associated with the first produce based on the first image,wherein the portable computing device is configured to operate in a second operation mode when mounted to the second mounting bracket, wherein the portable computing device is configured in use to capture, when mounted to the second mounting bracket, a second image, the second image being of second produce; wherein the portable computing device is configured to receive the second image in the second operation mode; and wherein the portable computing device is configured to determine at least one parameter associated with the second produce based on the second image.

59. A system for monitoring produce, comprising: a mounting bracket for mounting a portable computing device; a handheld camera; and the portable computing device, wherein the portable computing device is configured to operate in a mounted operation mode when mounted to the mounting bracket and to receive, in the mounted operation mode, a first image, the first image being of first produce, wherein the portable computing device is configured to determine at least one parameter associated with the first produce based on the first image, wherein the portable computing device is configured to operate in a second operation mode when demounted from the mounting bracket, wherein the portable computing device is configured in use to capture, using the handheld camera, a second image, the second image being of second produce; wherein the portable computing device is configured to receive the second image in the second operation mode; and wherein the portable computing device is configured to determine at least one parameter associated with the second produce based on the second image.

60. A defect detection system for detecting one or more defects associated with one or more produce items, the system comprising: a camera module configured to generate an image associated with the one or more produce items; and a processing unit configured to receive the image and to determine one or more defect parameters associated with the or each produce item using a machine learning model.

61. A defect detection system according to claim 60, wherein the one or more defect parameters include at least one of a presence or absence of a defect, a number of defects, or a type of defect.

62. A defect detection system according to claim 60, wherein the one or more defect parameters include at least a type of defect, and the type of defect includes one or more of mechanical damage,growth crack, discolouration, infestation, rot, secondary growth, scab, scurf, disease, bruising, and / or sprouting.

63. A defect detection system according to claim 62, wherein the discolouration is greening.

64. A defect detection system according to any of claims 60-63, wherein the processing unit is further configured to identify individual produce items in the image before determining the one or more defect parameters by segmenting the image into at least one portion which represents a single produce item.

65. A defect detection system according to claim 64, wherein the processing unit is configured to pass a portion of the image corresponding to a single produce item to the machine learning model to determine the one or more defect parameters for the single produce item.

66. A defect detection system according to claim 65, wherein the processing unit is configured not to pass a portion of the image that does not correspond to the single produce item to the machine learning model when determining the one or more defect parameters for that produce item.

67. A defect detection system according to any of claims 60-65, wherein the processing unit is further configured to use the determined one or more defect parameters to determine a value of the produce item or a group of produce items.

68. A defect detection system according to claim 67, wherein the defect detection system is configured to transfer the determined value from a first location to a second location automatically or in response to a user input.

69. A method for detecting defects in one or more produce items, the method comprising: training a machine learning model to determine one or more defect parameters associated with one or more produce items; generating an image of one or more produce items using a camera module; and analysing the image with a processing unit, using the machine learning model, to determine the one or more defect parameters for the one or more produce items.

Images