USB camera system enabled by AI and USB camera linkage method

Abstract

The invention discloses a USB (universal serial bus) camera system capable of being energized by AI (artificial intelligence), which comprises a USB camera for shooting videos in real time; the gateway obtains the video, extracts an image in the video, outputs a detection result, assembles the detection result into a JSON message, stores the JSON message in a memory, and covers a previous detection result with a current detection result; the detection result comprises a category label of the article and a corresponding confidence coefficient; the gateway is provided with a self-management system; the Internet of Things terminal creates an equipment linkage task in the gateway through a self-management system, and in the equipment linkage task, a USB camera, a category label of an article needing to be detected, a confidence coefficient condition and an equipment function of equipment triggered by meeting the confidence coefficient condition are set; and the gateway executes the task according to the equipment linkage task, and triggers the equipment function of the corresponding equipment when the confidence meets the confidence condition. The invention further discloses a USB camera linkage method. The invention is suitable for lightweight application and low in cost.

Description

Technical Field

[0001] The present invention relates to a monitoring device, and more specifically to an AI-enabled USB camera system and a USB camera linkage method. Background Art

[0002] The camera can record videos and store them locally or in the cloud. Users can view videos stored locally or in the cloud to understand what the camera has captured, which can be used for monitoring, anti-theft, etc. The camera is connected to a computer or other device via a USB interface, which is plug-and-play and does not require an acquisition card or power supply. It is easy to use and simple to operate. With the rapid development of AI technology, a large number of smart cameras have emerged on the market. These cameras integrate various AI algorithms for scenarios such as face recognition and object detection. However, existing AI cameras are relatively expensive and require support from specific hardware or cloud platforms, which limits their popularity in low-cost, lightweight scenarios.

[0003] Traditional USB cameras are cheap and widely used in various scenarios, but they have simple functions and can only shoot and store videos. If they are completely eliminated, the cost will be huge and cause great waste. Although traditional USB cameras can also be connected to the network through computers, network devices, and external devices such as mobile phones to achieve network interconnection, due to the lack of AI processing capabilities of USB cameras, even if the video shot by the USB camera is stored on the cloud platform through the network and then processed on the cloud platform, since the video data also needs to be stored on the cloud platform, there is not only the risk of privacy leakage, but also a large amount of storage space is required, which is costly, and it is also impossible to integrate traditional USB cameras into the Internet of Things system to achieve task linkage. Summary of the invention

[0004] In order to solve the technical problems that even if the prior art USB cameras realize data processing through a cloud platform, there are still privacy leaks, high costs, and cannot be well integrated into the Internet of Things system to realize task linkage, the present invention provides a USB camera system and a USB camera linkage method enabled by AI.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is to design a USB camera system enabled by AI, including: A USB camera that shoots video in real time, wherein the USB camera has a USB connection; A gateway is connected to the USB camera via the USB cable. The gateway acquires the video, extracts the image in the video, processes the image using an AI algorithm, and outputs a detection result. The detection result is assembled into a JSON message and stored in a memory, and the current detection result overwrites the previous detection result. The detection result includes a category label of an object and a corresponding confidence level. The gateway has a self-management system. An Internet of Things terminal connected to the gateway, wherein the Internet of Things terminal creates a device linkage task in the gateway through the self-management system, and in the device linkage task, a USB camera, a category label of an object to be detected, a confidence condition, and a device function of a device triggered by satisfying the confidence condition are set; The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

[0006] The AI-enabled USB camera system further includes: In the device linkage task, the frequency of task execution is set, and the gateway executes the task regularly according to the frequency.

[0007] The JSON message also includes a timestamp of when the detection result was generated.

[0008] The JSON message also includes the identity number of the image, the width of the image, the height of the image and the image format.

[0009] The JSON message also includes scene attribute information and location information of the object in the image.

[0010] The present invention also provides a USB camera linkage method, comprising: Connect the USB camera to the gateway via USB cable; USB camera to shoot video in real time; The gateway obtains the video, extracts the image in the video, processes it using an AI algorithm, and outputs a detection result, and assembles the detection result into a JSON message and saves it in the memory, and the current detection result overwrites the previous detection result; the detection result includes the category label of the object and the corresponding confidence level; the gateway has a self-management system; The IoT terminal is connected to the gateway through the IoT terminal, and the IoT terminal creates a device linkage task in the gateway through the self-management system. In the device linkage task, a USB camera, a category label of the object to be detected, a confidence condition, and a device function of the device triggered by satisfying the confidence condition are set; The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

[0011] The USB camera linkage method also includes: In the device linkage task, the frequency of task execution is set, and the gateway executes the task regularly according to the frequency.

[0012] The JSON message also includes a timestamp of when the detection result was generated.

[0013] The JSON message also includes the identity number of the image, the width of the image, the height of the image and the image format.

[0014] The JSON message also includes scene attribute information and location information of the object in the image.

[0015] The present invention sets a USB camera, a gateway and an Internet of Things terminal. The USB camera shoots a video, the gateway obtains the video, extracts an image in the video, processes it with an AI algorithm and outputs a detection result, assembles the detection result into a JSON message and stores it in a memory, and covers the previous detection result with the current detection result; the detection result includes a category label of an object and a corresponding confidence; the gateway has a self-management system; the Internet of Things terminal creates a device linkage task in the gateway through the self-management system, in which a USB camera, a category label of an object to be detected, a confidence condition and a device function of a device triggered by satisfying the confidence condition are set; the gateway executes a task according to the device linkage task, obtains the category label and confidence of the object of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition. Therefore, the detection results can be obtained by using the gateway, and only the latest detection results are retained in the memory to ensure the freshness of the storage and avoid excessive memory usage. At the same time, through the setting of the gateway, the gateway obtains the category label and confidence in the detection result, and triggers the device function of the corresponding device when the confidence meets the confidence condition, thereby integrating the USB camera into the Internet of Things. In addition, since the video data is stored locally, only the latest detection results are saved in the gateway, which not only occupies less memory, but also reduces the risk of privacy leakage, and can be better applied to lightweight scenarios. Finally, there is no need to make any changes to the existing USB camera, which realizes the full utilization of the existing USB camera and is low in cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention is described in detail below with reference to the embodiments and accompanying drawings, wherein: Figure 1 It is a structural principle diagram of the USB camera system enabled by AI in the present invention; Figure 2 It is a flow chart of the USB camera linkage method of the present invention. DETAILED DESCRIPTION

[0017] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings: See also Figure 1 The present invention utilizes an AI-enabled USB camera system comprising a USB camera 1, a gateway 2, and an IoT terminal 3. Among them: The USB camera 1 shoots video in real time, and the USB camera has a USB connection. The USB camera can be an existing USB camera, and no modification is required to the existing USB camera, so the cost is low.

[0018] The gateway 2 is connected to the USB camera via the USB cable. The gateway obtains the video, extracts the image in the video, processes it using an AI algorithm, and outputs the detection result. The detection result is assembled into a JSON message and saved in the memory, and the current detection result overwrites the previous detection result. The detection result includes the category label of the object and the corresponding confidence level. The gateway has a self-management system.

[0019] The gateway processes videos and extracts images from them using existing AI algorithms. The image processing module uses a lightweight model running on edge computing devices and supports multiple algorithms, such as YOLO for target detection and MobileNet for face recognition. It also supports parallel computing and model hot switching.

[0020] Image processing algorithms can support various existing algorithms according to actual needs. For face recognition, detect and identify the presence of a face for applications such as access control systems. For object detection, identify specific objects such as vehicles and packages for logistics or monitoring scenarios. For fire detection, detect flames, smoke, etc., and support automatic alarm systems. For gesture recognition, use it for interactive detection in scenarios such as remote control.

[0021] The detection result includes the category label of the object and the corresponding confidence level. The category label is the detected category, and the confidence level is the score of the object of the category in the image. The scoring criteria can be set according to clarity, completeness, matching, etc. The specific scoring criteria are set according to the actual scene and needs. The confidence level can be expressed in the form of 100 points or 1 point as the full score.

[0022] Here are a few category labels and confidence levels for illustration, for example: For the scene of detecting whether a person is wearing a mask, if it is detected that the person in the image is not wearing a mask, and the front image is captured, the integrity and clarity of the image are good, then the category label of the object is no mask: the confidence is 88 points. The detection result is stored in the memory in the form of {label (no mask): confidence (88)}.

[0023] For the face detection scenario, if the face in the image is only captured from the side and the image clarity is not high, then the category label of the object is face; the confidence level is 30 points; the detection result is stored in the memory in the form of {label (face): confidence level (30)}.

[0024] For the flame detection scenario, if the flame feature in the image is very obvious and the image is also very clear, then the category label of the object is flame, and the confidence is 100. The detection result is stored in the memory in the form of {label (flame): confidence (100)}.

[0025] The gateway only retains the most recent detection result. For previous detection results, no matter what the confidence level is, for the same category label, only the last detection result is retained to ensure the freshness of the storage and avoid excessive memory usage.

[0026] In this specific embodiment, in order to facilitate understanding of the time of the current result, the JSON message also includes a timestamp when the test result was generated. Through the timestamp, the time of the test result can be quickly known.

[0027] In this specific embodiment, in order to facilitate the rapid understanding of the image information of the detection result, the JSON message also includes the image identification number, image width, image height and image format. The image identification number can be generated according to the hardware address, year, month, day and specific time of the USB camera, or it can be generated by using the USB camera number and digital sequence coding. The image width and image height can be represented by existing parameters describing the width and height of the graphic, and the image format can be described in formats such as JPG and BMP.

[0028] In this specific embodiment, in order to facilitate understanding of the scene of the detection result and the position of the object in the image, the JSON message also includes scene attribute information and the position information of the object in the image. The current scene can be quickly known through the scene attribute information, and the specific position of the category label in the image can be quickly known through the position information of the object in the image, thereby quickly knowing the position of the object in the scene.

[0029] A specific design example of a JSON message is as follows: { "timestamp": "2024-11-01T12:34:56Z", / / timestamp of result generation "image_id": "img_123456", / / Unique ID to identify the image "image_metadata": { / / Image metadata "width": 1920, "height": 1080, "format": "jpeg" }, "object_detections": [ / / Object detection results { "label": "person", "confidence": 0.98, "bounding_box": { "x": 100, "y": 150, "width": 50, "height": 100 }, "attributes": { "age": "adult", "gender": "female", "clothing_color": "blue" } }, { "label": "car", "confidence": 0.95, "bounding_box": { "x": 400, "y": 200, "width": 100, "height": 50 }, "attributes": { "make": "Toyota", "color": "red" } } ], "image_classification": { / / Image detection results "predicted_label": "urban scene", / / predicted category label "confidence": 0.93 / / confidence }, "semantic_segmentation": { / / semantic segmentation result "segmented_image": "segmented_img_123456.png", / / segmented image file name "classes": [ / / All segmentation categories { "label": "road", "color": [128, 64, 128] / / RGB color }, { "label": "person", "color": [220, 20, 60] } ] }, "behavior_recognition": [ / / behavior recognition results { "label": "walking", "confidence": 0.92, "associated_objects": ["person"], "start_time": "2024-11-01T12:30:00Z", "end_time": "2024-11-01T12:35:00Z" }, { "label": "driving", "confidence": 0.89, "associated_objects": ["car"], "start_time": "2024-11-01T12:20:00Z", "end_time": "2024-11-01T12:35:00Z" } ], "scene_attributes": { / / Scene attributes "lighting_conditions": "outdoor", "time_of_day": "afternoon", "weather": "sunny" } } The IoT terminal 3 is connected to the gateway, and the IoT terminal creates a device linkage task in the gateway through the self-management system. In the device linkage task, the USB camera, the category label of the object to be detected, the confidence condition, and the device function of the device 4 triggered by meeting the confidence condition are set. The IoT terminal can be a computer, a mobile phone, etc., and it only needs to be able to connect to the gateway, access the gateway's self-management system, and set the self-management system.

[0030] The self-management system has a setting menu, through which the IoT terminal can conveniently set various parameters, thereby quickly creating linkage tasks.

[0031] The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

[0032] By creating a device linkage task in the gateway, the gateway can then execute tasks according to the created device linkage task, obtain the detection results of the corresponding USB camera, thereby obtaining the category label and confidence of the item, and judging whether the confidence meets the confidence condition of the device linkage task, and triggering the device function of the corresponding device when the confidence condition is met. For example, when used for flame detection, the category label of the item in the created device linkage task is flame, and the flame alarm is triggered when the confidence is greater than 50. If the category label of the item in the obtained detection result is flame and the confidence is 90 points, the gateway will send an alarm signal to the flame alarm, and the flame alarm will alarm.

[0033] In this specific embodiment, the USB camera system enabled by AI further includes: in the device linkage task, setting the frequency of task execution, and the gateway executing the task regularly according to the frequency. For example, in the device linkage task, the frequency of task execution is set to be executed once every 1 minute, then the gateway obtains the detection result of the corresponding USB camera every 1 minute, and determines whether the trigger condition is met according to the category label and confidence of the item, and triggers the device function of the corresponding device if it is met.

[0034] When configuring the device linkage task, select one or more attributes of device-function-return parameter, and configure the detection frequency of device linkage. At the same time, configure the trigger for a specific return parameter attribute, configure the category label, confidence, and other device functions called after the conditions are met. When the device linkage task is initiated, the gateway starts to detect the detection results, and obtains the corresponding detection results from the gateway's JSON message according to the different types of return parameter configuration.

[0035] Device functions are also designed in JSON format. An example of a specific design of device function design JSON is as follows: { "id": "18d26c09fc1d4d1989db91dc445f6690", "create_time": "2024-10-28 16:55:28", "name": "Image Recognition", "describe": "Get the most recent USB camera image recognition content", "mode_type": 0, "protocol_name": "OTHER", "protocol_req_describe": "{\"function_code\": \"SYSTEM_USB_CAMERA\", \"register_write\": []}", / / Description of the input parameters required by the device "protocol_resp_describe": [{ / / Return parameter description required by the device "lable": "person", "name": "person", "type": "1" / / Algorithm type 1 object detection, 2 image classification, 3 semantic segmentation, 4 behavior recognition }, { "key": "urban scene", "name": "City Scenery", "type": "2" }, { "key": "person", "name": "person", "type": "3" } { "key": "walking", "name": "Go", "type": "4" } ], "s_device_id": "", "type": 1, "code": "SYSTEM_USB_CAMERA" } The present invention sets a USB camera, a gateway and an Internet of Things terminal. The USB camera shoots a video, the gateway obtains the video, extracts an image in the video, processes it with an AI algorithm and outputs a detection result, assembles the detection result into a JSON message and stores it in a memory, and covers the previous detection result with the current detection result; the detection result includes a category label of an object and a corresponding confidence; the gateway has a self-management system; the Internet of Things terminal creates a device linkage task in the gateway through the self-management system, in which a USB camera, a category label of an object to be detected, a confidence condition and a device function of a device triggered by satisfying the confidence condition are set; the gateway executes a task according to the device linkage task, obtains the category label and confidence of the object of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition. Therefore, the detection results can be obtained by using the gateway, and only the latest detection results are retained in the memory to ensure the freshness of the storage and avoid excessive memory usage. At the same time, through the setting of the gateway, the gateway obtains the category label and confidence in the detection result, and triggers the device function of the corresponding device when the confidence meets the confidence condition, thereby integrating the USB camera into the Internet of Things. In addition, since the video data is stored locally, only the latest detection results are saved in the gateway, which not only occupies less memory, but also reduces the risk of privacy leakage, and can be better applied to lightweight scenarios. Finally, there is no need to make any changes to the existing USB camera, which realizes the full utilization of the existing USB camera and is low in cost.

[0036] Please also see Figure 2 The USB camera linkage method of the present invention comprises: Step 1: Connect the USB camera to the gateway via USB cable.

[0037] Step 2: Use the USB camera to shoot video in real time.

[0038] Step 3: The gateway obtains the video, extracts the image in the video, processes it using an AI algorithm, and outputs the detection result. The detection result is assembled into a JSON message and saved in the memory, and the current detection result overwrites the previous detection result. The detection result includes the category label of the object and the corresponding confidence level. The gateway has a self-management system.

[0039] The gateway processes videos and extracts images from them using existing AI algorithms. The image processing module uses a lightweight model running on edge computing devices and supports multiple algorithms, such as YOLO for target detection and MobileNet for face recognition. It also supports parallel computing and model hot switching.

[0040] Image processing algorithms can support various existing algorithms according to actual needs. For face recognition, detect and identify the presence of a face for applications such as access control systems. For object detection, identify specific objects such as vehicles and packages for logistics or monitoring scenarios. For fire detection, detect flames, smoke, etc., and support automatic alarm systems. For gesture recognition, use it for interactive detection in scenarios such as remote control.

[0041] The detection result includes the category label of the object and the corresponding confidence level. The category label is the detected category, and the confidence level is the score of the object of the category in the image. The scoring criteria can be set according to clarity, completeness, matching, etc. The specific scoring criteria are set according to the actual scene and needs. The confidence level can be expressed in the form of 100 points or 1 point as the full score.

[0042] Here are a few category labels and confidence levels for illustration, for example: For the scene of detecting whether a person is wearing a mask, if it is detected that the person in the image is not wearing a mask, and the front image is captured, the integrity and clarity of the image are good, then the category label of the object is no mask: the confidence is 88 points. The detection result is stored in the memory in the form of {label (no mask): confidence (88)}.

[0043] For the face detection scenario, if the face in the image is only captured from the side and the image clarity is not high, then the category label of the object is face; the confidence level is 30 points; the detection result is stored in the memory in the form of {label (face): confidence level (30)}.

[0044] For the flame detection scenario, if the flame feature in the image is very obvious and the image is also very clear, then the category label of the object is flame, and the confidence is 100. The detection result is stored in the memory in the form of {label (flame): confidence (100)}.

[0045] The gateway only retains the most recent detection result. For previous detection results, only the last detection result is retained to ensure the freshness of storage and avoid excessive memory usage.

[0046] In this specific embodiment, in order to facilitate understanding of the time of the current result, the JSON message also includes a timestamp when the test result was generated. Through the timestamp, the time of the test result can be quickly known.

[0047] In this specific embodiment, in order to facilitate the rapid understanding of the image information of the detection result, the JSON message also includes the image identification number, image width, image height and image format. The image identification number can be generated according to the hardware address, year, month, day and specific time of the USB camera, or it can be generated by using the USB camera number and digital sequence coding. The image width and image height can be represented by existing parameters describing the width and height of the graphic, and the image format can be described in formats such as JPG and BMP.

[0048] In this specific embodiment, in order to facilitate understanding of the scene of the detection result and the position of the object in the image, the JSON message also includes scene attribute information and the position information of the object in the image. The current scene can be quickly known through the scene attribute information, and the specific position of the category label in the image can be quickly known through the position information of the object in the image, thereby quickly knowing the position of the object in the scene.

[0049] Step 4: Connect to the gateway through the IoT terminal, and the IoT terminal creates a device linkage task in the gateway through the self-management system. In the device linkage task, set the USB camera, the category label of the item to be detected, the confidence condition, and the device function of the device triggered by meeting the confidence condition.

[0050] The self-management system has a setting menu, through which the IoT terminal can conveniently set various parameters, thereby quickly creating linkage tasks.

[0051] Step 5: The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

[0052] By creating a device linkage task in the gateway, the gateway can then execute tasks according to the created device linkage task, obtain the detection results of the corresponding USB camera, thereby obtaining the category label and confidence of the item, and judging whether the confidence meets the confidence condition of the device linkage task, and triggering the device function of the corresponding device when the confidence condition is met. For example, when used for flame detection, the category label of the item in the created device linkage task is flame, and the flame alarm is triggered when the confidence is greater than 50. If the category label of the item in the obtained detection result is flame and the confidence is 90 points, the gateway will send an alarm signal to the flame alarm, and the flame alarm will alarm.

[0053] In this specific embodiment, the USB camera linkage method further includes: In the device linkage task, the frequency of task execution is set, and the gateway executes the task regularly according to the frequency.

[0054] For example, in a device linkage task, if the frequency of task execution is set to once every 1 minute, the gateway will obtain the detection results of the corresponding USB camera every 1 minute, and determine whether the trigger conditions are met based on the category label and confidence level of the item. If so, the device function of the corresponding device will be triggered.

[0055] When configuring the device linkage task, select one or more attributes of device-function-return parameter, and configure the detection frequency of device linkage. At the same time, configure the trigger for a specific return parameter attribute, configure the category label, confidence, and other device functions called after the conditions are met. When the device linkage task is initiated, the gateway starts to detect the detection results, and obtains the corresponding detection results from the gateway's JSON message according to the different types of return parameter configuration.

[0056] The present invention has the following characteristics: 1. Use AI to empower ordinary USB cameras The present invention combines an ordinary USB camera with an existing AI algorithm to enable it to have an intelligent detection function, which can meet various application requirements without replacing hardware devices, thereby reducing equipment costs.

[0057] 2. Tag management mechanism of memory cache The system has designed a lightweight label cache mechanism that only retains the most recent detection result, avoiding redundant storage and complex database management, and improving processing speed and resource utilization efficiency.

[0058] 3. Combining edge computing with task linkage All data processing is completed in the local gateway without relying on cloud computing, which not only protects data privacy but also improves response speed. At the same time, by creating device linkage tasks, automatic task triggering is achieved, and device connection is achieved by accessing JSON data in memory.

[0059] 4. Flexible and scalable algorithm It supports users to dynamically load and switch different AI algorithms according to scene requirements, making ordinary USB cameras suitable for a variety of application scenarios, such as monitoring, access control, logistics, etc., achieving high scalability of the algorithm.

[0060] 5. Low-cost and efficient solution Compared with traditional AI cameras, it reduces hardware investment by utilizing off-the-shelf ordinary USB cameras and gateways to achieve intelligence, making it suitable for lightweight and large-scale deployment scenarios.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A USB camera system powered by AI, characterized in that include: A USB camera that shoots video in real time, wherein the USB camera has a USB connection; A gateway, which is connected to the USB camera via the USB cable, the gateway acquires the video, extracts the image in the video, outputs the detection result after processing using the AI ​​algorithm, and assembles the detection result into a JSON message and saves it in the memory, and the current detection result overwrites the previous detection result; the detection result includes the category label of the object and the corresponding confidence level; The gateway has a self-management system; An Internet of Things terminal connected to the gateway, wherein the Internet of Things terminal creates a device linkage task in the gateway through the self-management system, and in the device linkage task, a USB camera, a category label of an object to be detected, a confidence condition, and a device function of a device triggered by satisfying the confidence condition are set; The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

2. The AI-enabled USB camera system according to claim 1, characterized in that: The AI-enabled USB camera system further includes: In the device linkage task, the frequency of task execution is set, and the gateway executes the task regularly according to the frequency.

3. The AI-enabled USB camera system according to claim 1, characterized in that: The JSON message also includes a timestamp of when the detection result was generated.

4. The AI-enabled USB camera system according to claim 3, characterized in that: The JSON message also includes the identity number of the image, the width of the image, the height of the image and the image format.

5. The AI-enabled USB camera system according to claim 4, characterized in that: The JSON message also includes scene attribute information and location information of the object in the image.

6. A USB camera linkage method, characterized in that include: Connect the USB camera to the gateway via USB cable; USB camera to shoot video in real time; The gateway obtains the video, extracts the image in the video, processes it using an AI algorithm, and outputs a detection result, and assembles the detection result into a JSON message and saves it in a memory, and overwrites the previous detection result with the current detection result; the detection result includes the category label of the object and the corresponding confidence level; The gateway has a self-management system; The IoT terminal is connected to the gateway through the IoT terminal, and the IoT terminal creates a device linkage task in the gateway through the self-management system. In the device linkage task, a USB camera, a category label of the object to be detected, a confidence condition, and a device function of the device triggered by satisfying the confidence condition are set; The gateway executes the task according to the device linkage task, obtains the category label and confidence of the item of the corresponding USB camera in the memory, and triggers the device function of the corresponding device when the confidence meets the confidence condition.

7. The USB camera linkage method according to claim 6, characterized in that: The USB camera linkage method also includes: In the device linkage task, the frequency of task execution is set, and the gateway executes the task regularly according to the frequency.

8. The USB camera linkage method according to claim 6, characterized in that: The JSON message also includes a timestamp of when the detection result was generated.

9. The USB camera linkage method according to claim 8, characterized in that: The JSON message also includes the identity number of the image, the width of the image, the height of the image and the image format.

10. The USB camera linkage method according to claim 9, characterized in that: The JSON message also includes scene attribute information and location information of the object in the image.

Images