3D article loss detection method and system based on monocular camera, and medium

Abstract

The invention relates to a 3D article loss detection method and system based on a monocular camera, and a medium. The method comprises the following steps: collecting image data in real time through the monocular camera; preprocessing the acquired image data to improve the quality of the image data; performing feature extraction and pixel-level depth value prediction on the preprocessed image data by adopting a monocular depth estimation algorithm based on deep learning to obtain depth information of the object, and constructing a 3D model of the object based on the depth information of the object; detecting an article in the monitoring area by adopting a pre-trained target detection model, and determining the category and the initial position of the article; tracking the detected article in real time through an SORT algorithm, and recording the position change track of the article; judging whether the article is lost or not based on the initial position of the article, the 3D model, the position change track and a preset judgment rule; when it is judged that the article is lost, an alarm is given in a sound-light alarm and short message notification mode, and article loss information is stored.

Description

Technical Field

[0001] This disclosure relates to the field of lost item detection, specifically to a 3D lost item detection method, system, and medium based on a monocular camera. Background Technology

[0002] In today's society, the problem of lost items is commonplace, causing numerous inconveniences and losses to people's lives and work. Whether in public places such as shopping malls, train stations, and airports, or in private places such as homes and offices, lost items can lead to the loss of important documents, valuables, and even trigger a series of security problems. For example, if a passenger accidentally loses their wallet, ID card, or other important items while waiting at a train station, it will not only affect their journey but may also expose them to the risk of identity information leakage; when shopping in a mall, consumers who lose their personal belongings may suffer financial losses and psychological distress. Therefore, timely and accurate detection of lost items is of great significance for protecting people's property and maintaining social order.

[0003] Traditional methods for detecting lost items primarily rely on manual patrols and monitoring. Manual patrols suffer from inefficiency, limited coverage, and susceptibility to human error. During patrols, staff cannot monitor the status of items in all areas in real time, leading to missed detections. Furthermore, prolonged patrols can cause staff fatigue, further reducing accuracy. Manual monitoring involves reviewing surveillance video to identify lost items, but this method is often reactive and cannot issue alerts immediately upon loss, increasing the difficulty of recovery. In addition, the sheer volume of surveillance video requires significant time and effort for manual review, and crucial information can easily be missed due to human error. Moreover, traditional monitoring systems lack depth information about items, making it difficult to accurately determine whether an item is truly lost, leading to misjudgments.

[0004] Although there are currently available technologies for missing item detection, such as computer vision methods and multi-view camera detection methods, they also have some drawbacks, as follows: (1). Computer vision methods: In item loss detection, this method often employs background subtraction and inter-frame subtraction. Background subtraction establishes a background model, compares the current frame with the background model, detects foreground targets, and thus determines whether an item is missing. However, this method requires high background stability. When the background changes, such as changes in lighting or dynamic background interference, the background model may not be updated in time, resulting in false positives or false negatives.

[0005] Inter-frame difference methods detect moving targets by comparing differences between consecutive frames, but they are not effective at detecting the loss of stationary objects because the differences between frames before and after the object is lost may not be significant. Furthermore, traditional computer vision methods lack utilization of object depth information, making it difficult to accurately determine whether an object has moved from its original position, resulting in low detection accuracy.

[0006] (2) Multi-view camera detection method: This method uses multiple cameras to acquire images from different angles and calculates the depth information of the object using the principle of stereo vision, thereby realizing the detection of the object's state. Compared with monocular cameras, multi-view cameras can acquire depth information more directly and accurately, and the detection accuracy is relatively high.

[0007] However, multi-view cameras suffer from high costs, complex installation, and difficult calibration. Deploying multiple cameras requires consideration of factors such as location and angle, the calibration process is cumbersome, and they are difficult to install in some space-constrained scenarios. Furthermore, multi-view cameras handle a large amount of data, placing high demands on hardware performance and increasing system complexity and cost. Summary of the Invention

[0008] This disclosure proposes a method, system, and medium for 3D object loss detection based on a monocular camera, to solve at least one technical problem in the background art. The technical solution adopted in this disclosure is as follows: The first aspect of this disclosure provides a 3D object loss detection method based on a monocular camera, comprising the following steps: Step S100: Acquire image data of the monitored area in real time using a monocular camera; Step S200: Preprocess the acquired image data to improve the image data quality; the preprocessing includes denoising, enhancement and normalization. Step S300: Using a monocular depth estimation algorithm based on deep learning, feature extraction and pixel-level depth value prediction are performed on the preprocessed image data to obtain the depth information of the object and construct a 3D model of the object based on the depth information of the object. Step S400: Use a pre-trained object detection model to detect items within the monitored area, determine the item category and initial location; use the SORT (Simple Online and Realtime Tracking) algorithm to track the detected items in real time and record the trajectory of the items' position changes; Step S500: Determine whether the item is lost based on its initial position, 3D model, position change trajectory, and preset judgment rules; Step S600: When it is determined that an item is lost, an alarm is issued through sound and light alarm and SMS notification, and the item loss information is stored; the item loss information includes relevant images and videos of the item and the time and location information of the item loss.

[0009] Preferably, in step S300, the deep learning-based monocular depth estimation algorithm is trained using an image dataset containing depth information to adapt to different scenes and objects.

[0010] Furthermore, the monocular depth estimation algorithm based on deep learning preferably employs the pre-trained DORN (Deeper Ordinal Regression Network) algorithm.

[0011] Preferably, in step S400, the target detection model is preferably the YOLOv5 model.

[0012] Preferably, in step S400, the SORT algorithm uses Kalman filtering to predict the motion state of the object, and combines it with the Hungarian algorithm to match the detection results with the tracking target, thereby ensuring the stability of continuous real-time tracking.

[0013] Preferably, the preset judgment rules in step S500 include: S501: If the trajectory of the object's position change within 30 seconds exceeds the preset warning range of the initial position, and the distance between the depth information displayed in the object's 3D model and the initial position exceeds a preset depth threshold, then it is determined that the object has been moved. S502: If the moved item does not return to the preset warning range of its initial position within 1 minute, and there are no traces of personnel activity related to the moved item in the vicinity, and minor movements caused by wind or vibration are excluded, the item is determined to be lost.

[0014] It's important to note that an item is considered moved if it moves beyond the preset warning range of its initial position within 30 seconds, and is considered lost if it fails to return to the preset warning range of its initial position within 1 minute. These two timeframes are just examples; in actual implementation, the timeframes for determining movement and loss will depend on specific needs. For instance, an item might be considered moved only if it moves beyond the preset warning range of its initial position within 5 minutes, and lost only if it fails to return to the preset warning range of its initial position within 10 minutes.

[0015] A second aspect of this disclosure provides a 3D object loss detection system based on a monocular camera, the system comprising: The image acquisition module is used to acquire image data of the monitored area in real time through a monocular camera. The preprocessing module is used to preprocess the acquired image data to improve the image data quality; the preprocessing includes noise reduction, enhancement and normalization. The depth estimation module is used to extract features and predict pixel-level depth values ​​from preprocessed image data using a deep learning-based monocular depth estimation algorithm, obtain the depth information of the object, and construct a 3D model of the object based on the depth information. The detection and tracking module is used to detect items within the monitored area using a pre-trained object detection model, determine the item category and initial location, and track the detected items in real time using the SORT (Simple Online and Realtime Tracking) algorithm, recording the trajectory of the item's position change. The pre-trained object detection model is trained using a dataset containing a variety of common items. The status determination module is used to determine whether an item is lost based on its initial position, 3D model, position change trajectory, and preset judgment rules. The alarm storage module is used to issue an alarm via sound and light alarm and SMS notification when an item is determined to be lost, and to store the item loss information; the item loss information includes relevant images and videos of the item and the time and location information of the item loss.

[0016] In a third aspect, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the 3D object loss detection method based on a monocular camera as described above.

[0017] In a fourth aspect, this disclosure provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the monocular camera-based 3D object loss detection method as described above.

[0018] The beneficial effects of this disclosure are as follows: This disclosure proposes a 3D object loss detection method, system, and medium based on a monocular camera. It employs depth detection based on a monocular camera, providing an efficient and accurate solution for object loss detection. This disclosure fully utilizes the advantages of monocular cameras, such as low cost, small size, easy installation, and high precision in depth detection. It can be flexibly deployed in various scenarios and can acquire images through a monocular camera, using algorithms such as deep learning to calculate the depth information of objects, thus accurately determining changes in the position of the objects.

[0019] Compared to traditional methods, this disclosure enables real-time monitoring of changes in the state of objects. When an object is moved or taken, it can quickly identify the situation and issue an alarm, significantly improving the timeliness of lost object detection. Furthermore, by introducing depth information, this disclosure can effectively distinguish the true state of objects, reducing the occurrence of false positives. In addition, the monocular camera depth detection technology enables automated detection without manual intervention, reducing labor costs and improving detection efficiency.

[0020] Compared to traditional computer vision methods, this disclosure improves detection accuracy and robustness by introducing depth information, effectively addressing interference factors such as background changes. Compared to multi-camera detection methods, monocular cameras are lower in cost, simpler to install and maintain, and applicable to more scenarios. Furthermore, with the continuous development of deep learning algorithms, the accuracy of monocular camera depth detection is constantly improving, meeting the practical needs of lost item detection. Attached Figure Description

[0021] The accompanying drawings, which form part of this application, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0022] Figure 1 The flowchart of the 3D object loss detection method based on a monocular camera as described in Embodiment 1 of this disclosure is as follows. Figure 1 .

[0023] Figure 2 The flowchart of the 3D object loss detection method based on a monocular camera as described in Embodiment 1 of this disclosure is as follows. Figure 2 .

[0024] Figure 3 This is the architecture of the 3D object loss detection system based on a monocular camera as described in Embodiment 2 of this disclosure. Figure 3 .

[0025] Figure 4 This is the architecture of the 3D object loss detection system based on a monocular camera as described in Embodiment 2 of this disclosure. Figure 4 . Detailed Implementation

[0026] The present disclosure will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0027] The following detailed descriptions are exemplary and intended to provide further detailed explanation of this disclosure. Unless otherwise specified, all technical terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure.

[0028] Example 1: like Figure 1 As shown, this disclosure provides a 3D object loss detection method based on a monocular camera, the method including steps S100 to S600.

[0029] Step S100: Collect image data of the monitored area in real time using a monocular camera.

[0030] Furthermore, in step S100, the resolution of the monocular camera is not less than 1080P and the frame rate is 25-30FPS.

[0031] Furthermore, in step S100, the installation location of the monocular camera is selected based on the characteristics of the monitored area to ensure that the monocular camera can fully cover the monitored area and avoid blind spots. For indoor scenes, such as offices and homes, the camera can be installed on the ceiling or wall for a top-down view; for outdoor scenes, such as shopping mall entrances and station squares, it should be installed at a higher location to reduce the impact of obstructions. During installation, the camera's angle and focus also need to be adjusted to ensure that objects within the monitored area are clearly imaged.

[0032] Step S200: Preprocess the acquired image data to improve the image data quality.

[0033] Furthermore, the preprocessing includes denoising, enhancement, and normalization.

[0034] Step S300: Using a monocular depth estimation algorithm based on deep learning, feature extraction and pixel-level depth value prediction are performed on the preprocessed image data to obtain the depth information of the object and construct a 3D model of the object based on the depth information of the object.

[0035] Preferably, in step S300, the deep learning-based monocular depth estimation algorithm is trained using an image dataset containing depth information to adapt to different scenes and objects.

[0036] Furthermore, the monocular depth estimation algorithm based on deep learning preferably employs the pre-trained DORN (Deeper Ordinal Regression Network) algorithm.

[0037] Furthermore, the training process of the DORN algorithm includes: using a large image dataset containing depth information of different scenes and objects to train a deep convolutional neural network, optimizing the parameters of the ordinal regression model, thereby improving the accuracy of depth value prediction.

[0038] Furthermore, the DORN algorithm achieves pixel-level depth prediction through a deep convolutional neural network, and the training dataset contains depth images of different scenes and different types of objects.

[0039] In implementation, the DORN algorithm extracts features from image data using a deep convolutional neural network and predicts the depth value of each pixel in the image data using ordinal regression, thus achieving high-precision depth prediction and estimation. In practical applications, the DORN algorithm is first trained using a large dataset of images containing depth information to obtain a pre-trained DORN model, improving the algorithm's depth estimation capabilities for different scenes and objects. Then, the pre-trained DORN model is applied to actual image depth estimation, providing accurate depth data for determining the location of objects.

[0040] Step S400: Use a pre-trained object detection model to detect items within the monitoring area, determine the item category and initial location; use the SORT (Simple Online and Realtime Tracking) algorithm to track the detected items in real time and record the trajectory of the item's position change.

[0041] Furthermore, in step S400, the target detection model is preferably the YOLOv5 model.

[0042] The YOLOv5 algorithm is characterized by its fast detection speed and high accuracy, enabling it to quickly detect items within a monitored area and determine their category and location. During training, a dataset containing various common items is used to train the YOLOv5 model, improving its ability to detect different items.

[0043] Furthermore, in step S400, the SORT algorithm uses Kalman filtering to predict the motion state of the object, and combines it with the Hungarian algorithm to match the detection results with the tracking target, thereby ensuring the stability of continuous real-time tracking.

[0044] In step S400, the item tracking uses the SORT (Simple Online and Realtime Tracking) algorithm, which combines Kalman filtering and the Hungarian algorithm, so as to more accurately track the detected items in real time and record the precise movement trajectory of the items.

[0045] In practice, the pre-trained object detection model is trained using a dataset containing a variety of common items. Step S400 combines the obtained detection results (item category and initial position) with the tracking results (item position change trajectory) to achieve continuous monitoring of the items.

[0046] Step S500: Determine whether the item is lost based on the item's initial position, 3D model, position change trajectory, and preset judgment rules.

[0047] Furthermore, the preset judgment rules in step S500 include: S501: If the trajectory of the object's position change within 30 seconds exceeds the preset warning range of the initial position, and the distance between the depth information displayed in the object's 3D model and the initial position exceeds a preset depth threshold, then it is determined that the object has been moved. S502: If the moved item does not return to the preset warning range of its initial position within 1 minute, and there are no traces of personnel activity related to the moved item in the vicinity, and minor movements caused by wind or vibration are excluded, the item is determined to be lost.

[0048] In implementation, the preset warning range and preset depth threshold in step S500 can be dynamically adjusted according to the monitoring scenario and item category. In this disclosure, item status judgment is mainly based on the item's position change and depth information. First, a preset warning range and preset depth threshold are set for the item. When the item's position change trajectory exceeds the preset warning range of the initial position within a certain period (e.g., 30 seconds), and the depth information shows that the distance between the item and the initial position exceeds the preset depth threshold, it is determined that the item may have been moved. Then, the item's position change trajectory is continued to be tracked. If the item does not return to the preset warning range of the initial position for a long time (e.g., 1 minute) after being moved, and there are no traces of personnel activity related to the item in the vicinity, it is determined that the item is lost. At the same time, in order to reduce false judgments, the influence of environmental factors also needs to be considered, such as excluding slight movement of the item caused by wind, vibration, etc. The preset warning range and preset depth threshold can be determined according to actual needs, or can be implemented by existing technology, which will not be elaborated here.

[0049] Step S600: When it is determined that an item is lost, an alarm is issued through sound and light alarm and SMS notification, and the item loss information is stored; the item loss information includes relevant images and videos of the item and the time and location information of the item loss.

[0050] In one feasible implementation, such as Figure 2 As shown, the method may further include: Step S700: Read the stored item loss information and generate a loss event report based on the item loss information; the loss event report includes item category, loss time, location and related image screenshots.

[0051] Example 2: like Figure 3 As shown, in a second aspect, this disclosure provides a 3D object loss detection system based on a monocular camera, the system comprising: Image acquisition module 100 is used to acquire image data of the monitored area in real time through a monocular camera; The preprocessing module 200 is used to preprocess the acquired image data to improve the image data quality; the preprocessing includes denoising, enhancement and normalization. The depth estimation module 300 is used to perform feature extraction and pixel-level depth value prediction on the preprocessed image data using a monocular depth estimation algorithm based on deep learning, to obtain the depth information of the object and to construct a 3D model of the object based on the depth information of the object. The detection and tracking module 400 is used to detect items within the monitored area using a pre-trained object detection model, determine the item category and initial position, and track the detected items in real time using the SORT (Simple Online and Realtime Tracking) algorithm, recording the position change trajectory of the items. The pre-trained object detection model is trained using a dataset containing a variety of common items. The status judgment module 500 is used to determine whether an item is lost based on its initial position, 3D model, position change trajectory, and preset judgment rules. The alarm storage module 600 is used to issue an alarm via sound and light alarm and SMS notification when an item is determined to be lost, and to store the item loss information; the item loss information includes relevant images and videos of the item and the time and location information of the item loss.

[0052] Furthermore, the state determination module 500 includes: The movement determination unit is used to determine that the item has been moved if the trajectory of the item's position change within 30 seconds exceeds the preset warning range of the initial position, and the distance between the depth information displayed in the item's 3D model and the initial position exceeds a preset depth threshold. The loss determination unit is used to determine that an item is lost if the moved item does not return to its initial position within a preset warning range within 1 minute, and there are no traces of personnel activity related to the moved item in the vicinity, while excluding minor movements caused by wind or vibration.

[0053] In one feasible implementation, such as Figure 4As shown, the system may further include: The remote monitoring platform 700 is used to read the stored information on the lost items and generate a loss event report based on the information on the lost items; the loss event report includes the item category, the time of loss, the location, and related image screenshots.

[0054] It should be noted that the image acquisition module 100, preprocessing module 200, depth estimation module 300, detection and tracking module 400, status judgment module 500, alarm storage module 600, and remote monitoring platform 700 can correspond to steps S100, S200, S300, S400, S500, S600, and S700, respectively. However, the system described in Embodiment 2 is only one system architecture of the method described in Embodiment 1, and does not mean that the method described in Embodiment 1 necessarily depends on the system described in Embodiment 2.

[0055] Example 3: Embodiment 3 of this disclosure provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the 3D object loss detection method based on a monocular camera as described in Embodiment 1. Alternatively, a 3D object loss detection system based on a monocular camera, as described in Example 2, can be implemented.

[0056] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules, or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other memory technologies, CD-ROM (Compact Disc Read-Only Memory), DVD or other optical disc storage, cartridges, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer.

[0057] Example 4: Embodiment 4 of this disclosure provides an electronic device, which includes a processor and a memory. The processor is used to execute a computer program stored in the memory to implement the 3D object loss detection method based on a monocular camera described in Embodiment 1.

[0058] Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0059] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0064] In summary, the 3D object loss detection method, system, and medium based on a monocular camera provided in embodiments 1-4 of this disclosure offer an efficient and accurate solution for object loss detection by employing depth detection based on a monocular camera. This disclosure fully utilizes the advantages of monocular cameras, such as low cost, small size, easy installation, and high precision in depth detection. It can be flexibly deployed in various scenarios and can acquire images through a monocular camera, using algorithms such as deep learning to calculate the depth information of objects, accurately determining changes in their position. Compared to traditional methods, this disclosure can monitor changes in the state of objects in real time. When an object is moved or taken away, it can quickly identify the situation and issue an alarm, greatly improving the timeliness of object loss detection. Simultaneously, by introducing depth information, this disclosure can effectively distinguish the true state of objects, reducing the occurrence of misjudgments. Furthermore, monocular camera depth detection technology can achieve automated detection without manual intervention, reducing labor costs and improving detection efficiency.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit them. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of this disclosure. Any modifications or equivalent substitutions that do not depart from the spirit and scope of this disclosure should be covered within the protection scope of the claims of this disclosure.

Claims

1. A 3D object loss detection method based on a monocular camera, characterized in that, Includes the following steps: Step S100: Acquire image data of the monitored area in real time using a monocular camera; Step S200: Preprocess the acquired image data to improve the image data quality; The preprocessing includes noise reduction, enhancement, and normalization. Step S300: Using a monocular depth estimation algorithm based on deep learning, feature extraction and pixel-level depth value prediction are performed on the preprocessed image data to obtain the depth information of the object and construct a 3D model of the object based on the depth information of the object. Step S400: Use a pre-trained object detection model to detect items within the monitoring area, determine the item category and initial position; use the SORT algorithm to track the detected items in real time and record the trajectory of the item's position change; Step S500: Determine whether the item is lost based on its initial position, 3D model, position change trajectory, and preset judgment rules; Step S600: When an item is determined to be lost, an alarm is triggered via audible and visual alarm and SMS notification, and the item loss information is stored; The lost item information includes images and videos related to the item, as well as the time and location information of the item's loss.

2. The 3D object loss detection method based on a monocular camera as described in claim 1, characterized in that, In step S100, the resolution of the monocular camera is not less than 1080P and the frame rate is 25-30FPS.

3. The 3D object loss detection method based on a monocular camera as described in claim 1, characterized in that, In step S300, the monocular depth estimation algorithm based on deep learning is trained using an image dataset containing depth information to adapt to different scenes and objects. The monocular depth estimation algorithm based on deep learning preferably adopts the pre-trained DORN algorithm; The training process of the DORN algorithm includes: using a large image dataset containing depth information of different scenes and objects to train a deep convolutional neural network, optimizing the parameters of the ordinal regression model, thereby improving the accuracy of depth value prediction; The DORN algorithm achieves pixel-level depth prediction through a deep convolutional neural network, and the training dataset contains depth images of different scenes and different types of objects.

4. The 3D object loss detection method based on a monocular camera as described in claim 1, characterized in that, In step S400, the target detection model is the YOLOv5 model; In step S400, the SORT algorithm uses Kalman filtering to predict the motion state of the object, and combines it with the Hungarian algorithm to match the detection results with the tracking target, ensuring the stability of continuous real-time tracking.

5. The 3D object loss detection method based on a monocular camera as described in claim 1, characterized in that, The preset judgment rules in step S500 include: If the trajectory of the object's position change exceeds the preset warning range of the initial position within 30 seconds, and the distance between the depth information displayed in the object's 3D model and the initial position exceeds a preset depth threshold, then it is determined that the object has been moved. If the moved item does not return to its initial position within the preset warning range within 1 minute, and there are no traces of personnel activity related to the moved item in the vicinity, and minor movements caused by wind or vibration are excluded, the item is determined to be lost.

6. A 3D object loss detection system based on a monocular camera, characterized in that, The system includes: The image acquisition module is used to acquire image data of the monitored area in real time through a monocular camera. The preprocessing module is used to preprocess the acquired image data to improve the image data quality; the preprocessing includes noise reduction, enhancement and normalization. The depth estimation module is used to extract features and predict pixel-level depth values ​​from preprocessed image data using a deep learning-based monocular depth estimation algorithm, obtain the depth information of the object, and construct a 3D model of the object based on the depth information. The detection and tracking module is used to detect items within the monitored area using a pre-trained object detection model, determine the item category and initial position, and track the detected items in real time using the SORT algorithm, recording the trajectory of the item's position change. The pre-trained object detection model is trained using a dataset containing a variety of common items. The status determination module is used to determine whether an item is lost based on its initial position, 3D model, position change trajectory, and preset judgment rules. The alarm storage module is used to issue an alarm via sound and light alarm and SMS notification when an item is determined to be lost, and to store the item loss information; the item loss information includes relevant images and videos of the item and the time and location information of the item loss.

7. The 3D object loss detection system based on a monocular camera as described in claim 6, characterized in that, The status determination module includes: The movement determination unit is used to determine that the item has been moved if the trajectory of the item's position change within 30 seconds exceeds the preset warning range of the initial position, and the distance between the depth information displayed in the item's 3D model and the initial position exceeds a preset depth threshold. The loss determination unit is used to determine that an item is lost if the moved item does not return to its initial position within a preset warning range within 1 minute, and there are no traces of personnel activity related to the moved item in the vicinity, while excluding minor movements caused by wind or vibration.

8. The 3D object loss detection system based on a monocular camera as described in claim 6, characterized in that, The system also includes: A remote monitoring platform is used to read the stored information on the lost items and generate a loss event report based on the information on the lost items; the loss event report includes the item category, the time of loss, the location, and related image screenshots.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the 3D object loss detection method based on a monocular camera as described in any one of claims 1 to 5, or the 3D object loss detection system based on a monocular camera as described in any one of claims 6 to 8.

10. An electronic device comprising a processor and a memory, characterized in that, The processor is used to execute a computer program stored in the memory to implement the 3D object loss detection method based on a monocular camera as described in any one of claims 1 to 5, or to implement the 3D object loss detection system based on a monocular camera as described in any one of claims 6 to 8.

Images