WO2025212480A1

WO2025212480A1 - Object tracking and tracing using mobile devices

Info

Publication number: WO2025212480A1
Application number: PCT/US2025/022257
Authority: WO
Inventors: Tao CUI; Lingyun Wang
Original assignee: Siemens AG; Siemens Corp
Current assignee: Siemens AG; Siemens Corp
Priority date: 2024-04-03
Filing date: 2025-03-31
Publication date: 2025-10-09
Anticipated expiration: 2026-10-03

Abstract

A method performed by a mobile device for tracking and tracing objects includes capturing an image of an object in an industrial environment via a camera of the mobile device, querying object detection applications with the image, querying a track and trace database to determine a match based on the detection results, outputting a suggestion identifying the object via a user interface, receiving user input confirming or modifying the suggestion, and updating the track and trace database based on the confirmed or modified identification. In some cases, the method may also include geo-tagging the captured image using location data from the mobile device and updating location information in the database. In various scenarios, the object detection applications may include object detection models from publicly available services and/or custom models trained on inventory data.

Description

OBJECT TRACKING AND TRACING USING MOBILE DEVICES

TECHNICAL FIELD

[0001] The present disclosure relates to object tracking and tracing, and more particularly to methods, systems and computer program products for tracking and tracing diverse types of objects in industrial settings utilizing mobile devices with sensors, such as smartphones, tablets, etc. leveraging one or more object detection models.

BACKGROUND

[0002] Object tracking and tracing systems play an important role in various industrial settings, from warehouses and manufacturing facilities to construction sites and tool depots. These systems help organizations manage inventory, monitor asset utilization, and streamline operations by providing real-time information about the location and status of objects.

[0003] In mass-production environments, such as parcel delivery services, tracking and tracing solutions have been largely automated. Barcodes or RFID tags are typically affixed to each item, allowing for seamless scanning and sorting throughout the logistics chain. However, in high-mix, low-volume industrial settings, implementing similar fully automated systems may present challenges. For example, it may not be possible to have barcode on every object to be tracked and/or have scanners placed in many different positions in the process. Furthermore, these environments often deal with a diverse array of objects having a variety of physical features, many of which may lack standardized identification markers, which can complicate automated recognition.

[0004] The complexity of tracking and tracing in high-mix, low-volume settings often necessitates human involvement in the process. Operators frequently need to manually identify, sort, and record the location of various tools, components, or equipment. This manual approach can be time-consuming, error-prone, and inefficient, particularly when dealing with custom- made items or those lacking clear identification markers. As industrial operations continue to evolve and diversify, there is a growing need for flexible and adaptable tracking solutions that can bridge the gap between fully automated systems and manual processes, while leveraging existing technologies and infrastructure. SUMMARY

[0005] The present disclosure can address the challenge of efficiently tracking and tracing diverse objects in industrial environments, particularly in high-mix, low-volume settings without relying on specialized hardware. Aspects of the present disclosure provide a flexible, mobile-based solution that can assist human operators in identifying, sorting, and managing a wide variety of objects in industrial settings.

[0006] According to an aspect of this disclosure, a method performed by a mobile device for tracking and tracing objects is provided. The method comprises receiving, via a camera of the mobile device, an image of an object in an industrial environment. The method further comprises querying one or more object detection applications with the received image. Based on a result from the one or more object detection applications, the method comprises querying a track and trace database to determine a match. Based on the determination, the method comprises outputting, via a user interface, a suggestion identifying the object. The method further comprises receiving, via the user interface, user input confirming or modifying the suggestion identifying the object. The method then comprises updating the track and trace database based on the confirmed or modified identification of the object.

[0007] According to another aspect of this disclosure, a computer program product is provided. The computer program product comprises a non-transitory computer-readable medium including instructions that, when executed by a processor, cause the processor to perform a method as described above

[0008] According to yet another aspect of the present disclosure, a system for tracking and tracing objects is provided. The system comprises a tracking database and a mobile device. The mobile device includes a camera, a user interface, a processor, and memory. The memory stores instructions executable by the processor, which configure the mobile device to perform a method as described above.

[0009] Additional technical features and benefits may be realized through the techniques of the present disclosure. Embodiments and aspects of the disclosure are described in detail herein and are considered a part of the claimed subject matter. For a better understanding, refer to the detailed description and to the drawings. BRIEF DESCRIPTION OF FIGURES

[0010] The foregoing and other aspects of the present disclosure are best understood from the following detailed description when read in connection with the accompanying drawings. To easily identify the discussion of any element or act, the most significant digit or digits in a reference number refer to the figure number in which the element or act is first introduced.

[0011] FIG. 1 illustrates a use-case of sorting tools in a warehouse, wherein embodiments of this disclosure are applicable.

[0012] FIG. 2 schematically illustrates a system for tracking and tracing object using a mobile device, according to one or more embodiments.

[0013] FIG. 3 illustrates a process including offline training of an object detection model and online operational phase using the trained model, according to an example embodiment.

DETAILED DESCRIPTION

[0014] The following description sets forth exemplary aspects of the present disclosure. It should be recognized, however, that such description is not intended as a limitation on the scope of the present disclosure. Rather, the description also encompasses combinations and modifications to those exemplary aspects described herein.

[0015] A detailed description of systems, devices, and methods consistent with embodiments of the present disclosure is provided below. While several embodiments are described, it should be understood that disclosure is not limited to any one embodiment, but instead encompasses numerous alternatives, modifications, and equivalents. In addition, while numerous specific details are set forth in the following description in order to provide a thorough understanding of the embodiments disclosed herein, some embodiments can be practiced without some or all of these details. Moreover, for the purpose of clarity, certain technical material that is known in the related art has not been described in detail in order to avoid unnecessarily obscuring the disclosure.

[0016] In general, the methodology described herein provides a technique for tracking and tracing objects in industrial environments using mobile devices. The technique may be particularly useful in high-mix, low-volume settings where traditional automated tracking systems may be impractical or cost-prohibitive. The technique does not rely on specialized vision system hardware but can leverage mobile devices typically carried by human operators, such as smartphones or tablet computers (among others) equipped with standard features including a built-in camera and a user interface.

[0017] As used herein, the term "mobile device" generally refers to a portable electronic device capable of wireless communication, data processing, and image capture. Mobile devices may include smartphones, tablet computers, personal digital assistants (PDAs), handheld computers, or other similar devices equipped with a camera and user interface. A camera of a mobile device may be configured to capture high-definition color images, such as RGB images (containing information in red, green and blue color channels), that can be processed by object detection models. A "user interface" in this context refers to the means by which a user interacts with the mobile device and mobile app, typically including a touchscreen display for input and output.

[0018] The present methodology may be realized by a mobile application ("mobile app") including a software component designed to run on a mobile device providing specific functionality related to object tracking and identification. The mobile app may be configured to interface with one or more object detection applications and with a track and trace database.

[0019] As used herein, the term "object detection application" refers to a software component that incorporates a trained object detection model, typically including a deep neural network. The object detection model may be provided by a publicly available service or may be custom -trained using relevant objects specific to the industrial setting. The object detection application may process input images to identify, locate, and classify objects within those images. In some cases, the object detection application may include additional functionalities such as image preprocessing, post-processing of detection results, or integration with other systems.

[0020] A "track and trace database" refers to a data storage system that maintains information about objects, their identities, and locations within an industrial setting. For example, this database may include unique identifiers, serial numbers, model numbers, part descriptions, physical characteristics (such as dimensions, weight, and material composition), manufacturer information, and associated metadata for each object. It may also store current and historical location data, including both physical and logical locations within the industrial environment. In some cases, the database may contain timestamps for various events, user interactions (such as confirming or modifying object identifications), and historical tracking records. To efficiently manage and retrieve large volumes of tracking data across multiple industrial environments, the track and trace database can be implemented using various database technologies, including relational database structures, NoSQL databases, or distributed ledger technologies.

[0021] In operation, an operator may capture an image of an object to be identified using their mobile device camera. An instance of the mobile app running on that mobile device may then receive the image from the mobile device camera, query one or more object detection applications with the received image, and based on a result thereof, query the track and trace database to determine a match. Based on the determination, the mobile app may output a suggestion identifying the object via the user interface, receive user input confirming or modifying the suggestion, and update the track and trace database based on the confirmed or modified identification. In some cases, the mobile app, may also geo-tag the captured image using location data from the mobile device and update location information in the track and trace database.

[0022] The present methodology can integrate identification methods based on barcodes and labels (where available), providing a unified solution to address diverse high-mix use cases in industrial environments. Utilizing a smartphone-based mobile app, operators can easily identify parts using existing camera (and additionally, depth sensors such as LiDAR sensors) without additional specialized hardware. The mobile app's connection to backend database systems can enable real-time streamlining of operations at the point of use. In some cases, the mobile app may allow the user to label the data for future training, essentially expanding the types of the objects that can be tracked and traced using the mobile app. Also, in some cases, CAD model-based training may be incorporated in backend operations, reducing time and effort required for object detection model development. By functioning as an assistant to human operators, the mobile app may provide suggestions, while maintaining human decisionmaking authority, potentially improving efficiency and mitigating detection errors.

[0023] Turning now to the figures, FIG. 1 illustrates a use-case of sorting tools in a warehouse, wherein embodiments of the present methodology are applicable. The figure depicts two images, namely a first image of a toolbox 100 returned from a job site and second image of a warehouse 150 where tools are to be stored.

[0024] As shown, toolbox 100 may contain a variety of items that need to be sorted and stored in the warehouse 150. The contents of the toolbox 100 may include standard tools 102, such as wrenches, screwdrivers, and drills. Standard tools 102, in this context, refer to tools that are widely known and may be recognized by object detection models from publicly available services. In addition to standard tools 102, the toolbox 100 may also contain custom- made tools 104. Custom-made tools 104 may be designed for specific tasks or operations particular to the job site. These tools may not be readily recognized by public object detection models. However, in many cases, computer-aided design (CAD) models may be available for such custom-made tools 104.

[0025] When tools are shipped from the warehouse 150 to a job site, they are typically organized. However, upon return, the tools are often unorganized, randomly placed in the toolbox 100 as shown in FIG. 1. Furthermore, in addition to tools 102, 104, the toolbox 100, as returned from the job site may contain items that do not belong to the warehouse 150, such as gloves 106 or other protective equipment.

[0026] The warehouse 150 represents the storage area where tools are organized and stored when not in use. The warehouse 150 may feature multiple storage cells 152, which can be compartments or shelves designed to house specific tools or categories of tools. The storage cells 152 may be addressable, for example, using aisle and bin numbers, among other methods. The storage cells 152 may provide a structured system for organizing the various items returned in toolboxes like the one shown.

[0027] In this use-case, one of the challenges for an operator is that, after the tools are returned from the job site, the tools need to be sorted and placed in the appropriate storage cells 152 in the warehouse 150. Some tools may have a label, logo, barcode or other identification. However, in many cases, tools may not have such identification. Furthermore, as described above, while some tools can be standard (i.e., commonly available in the marketplace), other tools may be custom-made (produced at the shopfloor). For the latter, there could be CAD models available, while for the former, no such models may be available. Also, some items may not belong to the warehouse. So, there is a high mix of items.

[0028] Sorting such a high mix of items presents challenges for automation, and when done manually, can be a time-consuming and error-prone process. Embodiments of the present methodology disclosed hereinafter may address these challenges by providing a system for tracking and tracing objects in such industrial environments.

[0029] FIG. 2 illustrates a block diagram of a system 200 for tracking and tracing objects in an industrial environment. The system 200 includes a mobile device 202 with a built-in camera 204 and a user interface 206. The system 200 may be used to address challenges such as those described in relation to FIG. 1, where tools and equipment returned from field use need to be sorted and organized in a warehouse.

[0030] Functional aspects of the system 200 may be realized by way of hardware and programming of the mobile device 202. The programming may take the form of processorexecutable instructions stored on non-transitory machine-readable storage mediums and the hardware may include one or more processors to execute those instructions. As disclosed herein, the programming may be incorporated in a mobile application or mobile app 208.

[0031] As shown, the mobile device 202 may comprise a smartphone. The camera 204 of the mobile device 202 may be configured to capture high-definition color images, such as RGB images, of objects to be identified and tracked. The mobile app 208 may be installed on the mobile device 202 to provide specific functionality related to object tracking and identification as described herein. The mobile app 208 may be configured to access images captured by the camera 204. The mobile app 208 may be further configured to make API calls to respectively interface with one or more object detection applications 210, 212 and with a track and trace database 214. User interface 206 of the mobile device 202 may allow a user / operator to interact with the mobile app 208, including displaying suggestions about identified objects and receiving user input to confirm or modify the suggestions, input object labels, among others. In some embodiments, as shown, the user interface 206 may include a touch screen.

[0032] Consistent with disclosed embodiments, the system 200 may include a first object detection application 210 and a second object detection application 212. The first object detection application 210 may include an object detection model from a publicly available service, such as Google Lens or Amazon Catalog , among others. For example, in the usecase illustrated in FIG. 1, the first object detection application 210 can be useful in identifying the standard tools 102. The second object detection application 212 may include an object detection model custom -trained using labeled real -world images of objects of interest in the industrial environment or using computer-aided design (CAD) data of objects of interest. In the illustrated use-case, the second object detection application 212 may be particularly useful for identifying custom-made tools 104 that may not be recognized by publicly available models. In various embodiments, the second object detection application 212 may be hosted in an onsite server of the industrial facility or may be executed locally on the mobile device 202.

[0033] The object detection models may include various architectures and datasets. For example, convolutional neural network (CNN) architectures such as Mask R-CNN, YOLO (You Only Look Once), or SSD (Single Shot Detector) may be employed for their efficiency in real-time object detection tasks. The models may be pre-trained on large-scale datasets like ImageNet or Common Objects in Context (COCO) and then fine-tuned on domain-specific datasets.

[0034] In some cases, specialized object detection models such as PoseCNN may be used that are designed for pose estimation and 3D object detection. These models may be particularly useful when dealing with complex industrial objects or when precise spatial/geometrical information is required.

[0035] For custom-trained models, datasets specific to the industrial setting may be created. This may include collections of labeled images of tools, parts, and equipment commonly found in the particular industrial environment. In some implementations, synthetic datasets generated from CAD models may be used to create or augment the training data, especially for custom- made tools or rare items.

[0036] In some cases, domain-specific datasets such as the Digital Twin Tracking Dataset (DTTD) or DTTD-Mobile may be used, which are designed for object tracking in industrial settings. These datasets may provide a foundation for training models that can handle the unique challenges of industrial environments, such as varying lighting conditions, occlusions, and diverse object types. [0037] The track and trace database 214 may serve as a central repository for storing and managing information about objects within the industrial environment. The track and trace database 214 may store information about objects, their identities, and locations within the industrial setting. For example, the database 214 may contain detailed records for each object, including unique identifiers, serial numbers, model numbers, part descriptions, physical characteristics (such as dimensions, weight, and material composition), manufacturer information, and associated metadata. Additionally, the database 214 may store both current and historical location data, encompassing physical locations within the facility (e.g., where an object is on the shop floor) as well as logical locations (e.g., which bin or shelf the object is stored in).

[0038] The track and trace database 214 may be queried and updated in real time by the mobile app 208 based on the results of object detection and user input, ensuring that the stored data remains current and accurate, which allows for efficient inventory management and asset tracking across the industrial setting. In some embodiments, the track and database 214 may maintain a history of object movements including location data and associated timestamps. This can be used to synchronize an object's physical and logical location with its digital representation, for example, in a Manufacturing Execution System (MES), or a Product Lifecycle Management (PLM) system, or an inventory management system, among others. The digital information can be effectively utilized in the business logic, such as arranging and triggering work orders, ensuring product quality, preserving digital traces for warranty, etc.

[0039] The integration between the mobile app 208 and the track and trace database 214 can enable a seamless flow of information, allowing operators to access up-to-date object data in the field and make informed decisions about asset management and resource allocation. This coupling between mobile data collection and centralized data storage may thus form the backbone of the tracking and tracing system 200.

[0040] In operation, the camera 204 of the mobile device 202 may be used capture an image of an object in the industrial environment. For example, in the use-case illustrated in FIG. 1, the image may be captured by an operator using their mobile device at the point of sorting tools in the toolbox 100 returned from the jobsite for storage in the warehouse 150. The mobile app 208 may receive this image and query one or more of the object detection applications 210, 212 with the image and receive a result. [0041] In some embodiments, the result from the object detection applications 210, 212 may be generated by fusing inference outputs of multiple object detection applications. The fusion process can enhance the accuracy and reliability of object detection by leveraging the strengths of different models or approaches. Fusing the inference outputs may comprise ranking the outputs based on a confidence score associated with the inference outputs. The fusion process may also involve selecting one or more of the outputs based on confidence scores associated with the inference outputs. For example, if one model produces a high- confidence inference output while another produces a low-confidence inference output, the high-confidence inference output may be prioritized.

[0042] Based on the results from the object detection applications 210, 212, the mobile app 208 may query the track and trace database 214 to determine a match. The matching process may involve comparing various attributes of the detected object, such as its identified label, dimensions, etc. with the records stored in the database. The track and trace database 214 may contain detailed information about known objects in the industrial environment, including their identifiers, physical characteristics, and current or historical locations. In some cases, the matching process may employ algorithms to calculate similarity scores or use rule-based systems to identify the most likely match for the detected object. In cases where multiple potential matches are found, the mobile app 208 may rank them based on confidence levels. If an exact match is not found, the mobile app 208 may still provide the closest matches as suggestions for the user to review. The matching step thereby matches the real -world object captured in the image to its digital representation in the tracking system, enabling inventory management and asset tracking.

[0043] Based on a result of the matching process, the mobile app 208 may output, via the user interface 206, one or more suggestions identifying the object. For example, as shown in FIG. 2, the suggestion may identify the detected object and its logical location in the track and trace database 214. The mobile app 208 may then receive, via the user interface 206, user input confirming or modifying the suggestion identifying the object. Based on this input, the mobile app 208 may update the track and trace database 214. In some embodiments, such as in the illustrated use-case, the mobile app 208 may update the track and trace database 214 by adding an instance of the identified object to a logical location in the track and trace database, the logical location corresponding to a physical storage location 152 in the warehouse 150. [0044] In some embodiments, if no match is determined by the mobile app 208 or confirmed by the user, the mobile app 208 may receive, via the user interface 206, user input to register the object as a new object in the track and trace database 214. The mobile app 208 may also receive, via the user interface 206, a label for the image identifying the object, as shown in FIG. 2. This labeled image may be used as part of a dataset for re-training an object detection model of at least one of the object detection applications 210, 212. To that end, the system 200 may also include a training module configured to retrain an object detection model based on an image label received via the user interface 206. This may allow the system 200 to continuously improve its object detection capabilities based on user feedback.

[0045] The training module may comprise one or more software components for retraining an object detection model using image labels of new objects input by the user through an iterative process that leverages the newly acquired data. The training module may be hosted, for example, in an onsite server, or in the mobile device itself. When a user inputs a label for a previously unrecognized object, this labeled image may be added to a training dataset. In some embodiments, the re-training process may involve fine-tuning the existing neural network weights using transfer learning techniques. In this approach, the pre-trained layers of the neural network may be frozen or updated with a low learning rate, while the final layers may be retrained more aggressively using the new labeled data. In some embodiments, the training module may employ data augmentation techniques to expand the dataset created from user- labeled images. This may include applying transformations such as rotations, flips, color jittering, or adding noise to the original images, thereby creating multiple training samples from a single labeled image. In some embodiments, the training module may utilize techniques like few-shot or one-shot learning to adapt the model to new object classes with limited examples. This may allow the system to quickly incorporate new objects into its detection capabilities without requiring large numbers of labeled examples. The re-training process may be triggered periodically, after a certain number of new labeled images have been collected, or on-demand based on system requirements. In some cases, the training may occur in the background on the mobile device itself, leveraging on-device machine learning capabilities. After re-training, the updated model may undergo a validation process using a held-out dataset to ensure that the changes have improved overall performance without introducing regressions. If the validation is successful, the updated model may be deployed for use in the object detection applications.

[0046] In some embodiments, the mobile app 208 may also geo-tag the received image using a location service running on the mobile device 202. Examples of location services include Global Positioning Service (GPS), Assisted Global Position Service (A-GPS) or indoor localization features such as Wi-fi, Bluetooth, etc. available on the mobile device 202. When updating the track and trace database 214, the mobile app 208 may update location information associated with the identified object based on this geo-tagging. The mobile app 208 may also update the track and trace database 214 with a timestamp associated with the captured image, that correlates with the location information, thereby enabling tracking of object movement.

[0047] In some embodiments, the mobile device 202 may further comprise a depth sensor. For example, many state-of-the-art smartphones are provided with a built-in Light Detection and Ranging (LiDAR) sensor, which may be utilized as a depth sensor in this case. Utilizing a depth map captured by a depth sensor (such as LiDAR sensor) in combination with an RGB image captured by the camera 204, a 3D image of the object may be generated, which may be received as input by the mobile app 208. An example of a 3D image is an RGB-D image, which includes information in red, green and blue color channels as well as a depth channel. In this case, one or more of the object detection applications 210, 212 may include an object detection model (e.g., PoseCNN) capable of processing 3D images to identify objects. Furthermore, in some embodiments, the mobile app 208 may query the track and trace database 214 by providing object dimensions derived from such a 3D image. For example, a given type of tool (e.g., screwdriver) may come in different sizes. A combination of camera and depth sensor can be used to provide results that are more specific, enabling the user to find the right match in the track and trace database 214.

[0048] FIG. 3 illustrates a process 300 including offline training of an object detection model 302 and online operational phase using the trained model, according to an example embodiment.

[0049] The process 300 utilizes 3D CAD models 304 as part of its training phase, which occurs offline. The 3D CAD models 304 may be used to generate synthetic data, for example by randomizing the background, environment, pose, etc. (e.g., by randomizing background surface textures and/or colors, ambient lighting, object orientation, etc.). The synthetic data generated using the 3D CAD models 304 may be used to train the object detection model 302. In some cases, the object detection model 302 may comprise a convolutional neural network architecture, such as Mask R-CNN, PoseCNN, etc. [0050] The trained object detection model 302 may be used to perform inference of object detection on real RGB-D images. In some cases, the inference may be executed on the mobile device 202 in the field, using the RGB-D sensor (combination of camera and LiDAR sensor) of the mobile device 202.

[0051] For inference, the model input 306 comprises two components: a depth image 306a and an RGB image 306b. The depth image 306a may provide three-dimensional spatial information, while the RGB image 306b may supply color and texture data. The model input 306 may be processed by the object detection model 302 during the online operating phase.

[0052] The object detection model 302 may analyze the input data and generate a model output (inference) 308. The output 308 may include the identification and localization of objects within the input images. The model output 308 may be visualized through bounding boxes overlaid on the original image. Three such bounding boxes are shown: a bounding box 308a, a bounding box 308b, and abounding box 308c, each corresponding to a detected object in the scene. Each bounding box 308a, 308b, 308c may be associated with an identification label, which may then be used to query a track and trace database to determine a match.

[0053] In another aspect, embodiments of this disclosure may be included in an article of manufacture (e.g., one or more computer program products) having, for example, a non- transitory computer-readable storage medium. The computer readable storage medium has embodied therein, for instance, computer readable program instructions for providing and facilitating the mechanisms of the embodiments of the present disclosure. The article of manufacture can be included as part of a computer system or sold separately.

[0054] The computer readable storage medium can include a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. Computer readable program instructions described herein can be downloaded to respective computing/processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and/or a wireless network.

Claims

1. A method performed by a mobile device for tracking and tracing objects, the method comprising: receiving, via a camera of the mobile device, an image of an object in an industrial environment, querying one or more object detection applications with the received image, based on a result from the one or more object detection applications, querying a track and trace database to determine a match, based on the determination, outputting, via a user interface, a suggestion identifying the object, receiving, via the user interface, user input confirming or modifying the suggestion identifying the object, and updating the track and trace database based on the confirmed or modified identification of the object.

2. The method according to claim 1, further comprising geo-tagging received image using a location service running on the mobile device, wherein updating the track and trace database comprises updating location information associated with the identified object in the track and trace database based on the geo-tagging.

3. The method according to claim 1, wherein updating the track and trace database includes adding an instance of the identified object to a logical location in the track and trace database, the logical location corresponding to a physical storage location in a warehouse.

4. The method according to any of claims 1 to 3, wherein the one or more object detection applications include an object detection model from a publicly available service.

5. The method according to any of claims 1 to 4, wherein the one or more object detection applications include an object detection model custom -trained using labeled real- world images of objects of interest in the industrial environment.

6. The method according to any of claims 1 to 5, wherein the one or more object detection applications include an object detection model custom -trained using computer- aided design (CAD) data of objects of interest.

7. The method according to any of claims 1 to 6, wherein the result from the one or more object detection applications is generated by fusing inference outputs of multiple object detection applications.

8. The method according to claim 7, wherein fusing the inference outputs of multiple object detection applications comprises ranking the inference outputs, and/or selecting one or more of the inference outputs, based on confidence scores associated with the inference outputs.

9. The method according to any of claims 1 to 8, comprising, if no match is determined or confirmed: receiving, via the user interface, user input to register the object as a new object in the track and trace database.

10. The method according to claim 9, further comprising: receiving, via the user interface, a label for the image identifying the object, wherein labeled image is used as part of a dataset for re-training an object detection model of at least one of the one or more object detection applications.

11. The method according to any of claims 1 to 10, wherein the mobile device further comprises a depth sensor, and wherein the image of the object comprises a three- dimensional (3D) image.

12. The method according to claim 11, wherein the one or more object detection applications include an object detection model capable of processing 3D images to identify objects.

13. The method according to any of claims 11 and 12, wherein querying the track and trace database includes providing object dimensions derived from the 3D image.

14. A computer program product comprising a non-transitory computer-readable medium storing instructions that, when executed by a processor of a mobile device, cause the mobile device to perform a method according to any of claims 1 to 13.

15. A system for tracking and tracing objects, the comprising: a track and trace database, a mobile device comprising a camera, a user interface, a processor and memory, the memory storing instructions executable by the processor, which configure the mobile device to: receive, via the camera, an image of an object in an industrial environment, query one or more object detection applications with the received image, based on a result from the one or more object detection applications, query the track and trace database to determine a match, based on the determination, output, via the user interface, a suggestion identifying the object, receive, via the user interface, user input confirming or modifying the suggestion identifying the object, and update the track and trace database based on the confirmed or modified identification of the object.

16. The system according to claim 15, further comprising a server configured to host at least one of the one or more object detection applications.

17. The system according to any of claims 15 and 16, wherein the mobile device is configured to execute at least one of the one or more object detection applications locally on the mobile device.

18. The system according to any of claims 15 to 17, wherein the instructions further configure the mobile device to: geo-tag the received image using a location service running on the mobile device, and update the track and trace database by updating location information associated with the identified object in the track and trace database based on the geo-tagging.

19. The system according to any of claims 15 to 17, wherein the instructions configure the mobile device to: update the track and trace database by adding an instance of the identified object to a logical location in the track and trace database, the logical location corresponding to a physical storage location in a warehouse.

20. The system according to any of claims 15 to 17, further comprising a model training module configured to retrain at an object detection model of at least one of the one or more object detection applications based on an image label received via the user interface identifying the object.