On-device ai-based localization method and device based on image and fusion localization information

Abstract

The present invention relates to a real-time localization method and device using on-device AI on the basis of an image and fusion localization information. The present invention includes a method for collecting Wi-Fi RTT-based wireless localization information, CCTV image information, and sensor data in an FTM tag, and determining a user location by means of data fusion using an AI algorithm. The present invention applies weights by assessing data reliability, and corrects a location error by learning a movement pattern. Through this, the present invention can be used in various environments such as industry, security, logistics, or the like by lowering the dependency on a central server and enabling precise real-time positioning.

Description

On-device AI-based location determination method and device based on image and composite positioning information

[0001] The present invention relates to positioning, fusion localization, and AI-based data fusion technology. More specifically, it relates to an on-device AI-based location identification device and method that determines the 3D location of a user in real time by fusing wireless positioning information using Wi-Fi RTT (Fine Time Measurement), object location information based on CCTV video, sensor data of an FTM tag, and user personal information using AI.

[0002] Location tracking technology plays a crucial role in modern industrial and security environments, and various methods have been researched and put into practical use. Representative location tracking technologies include GPS (Global Positioning System), Wi-Fi-based positioning technology, RFID (Radio Frequency Identification), BLE (Bluetooth Low Energy)-based positioning, and UWB (Ultra-Wideband) positioning technology.

[0003] While GPS provides relatively high-accuracy location information over a wide area, it has the disadvantage of significantly degraded tracking accuracy in indoor environments due to signal blocking and multipath issues. Therefore, there is a need for alternative technologies to achieve precise location measurement in indoor settings.

[0004] To address these issues, Wi-Fi-based positioning technology has emerged. Generally, Wi-Fi positioning methods include those utilizing Received Signal Strength Indicator (RSSI) and Round Trip Time (RTT) methods based on Fine Time Measurement (FTM). While RSSI-based Wi-Fi positioning technology allows for relatively simple implementation, it has limitations in that location measurement errors occur depending on the signal environment. On the other hand, FTM-based Wi-Fi RTT technology enables more precise distance measurement by utilizing the signal round-trip time between an access point (AP) and a tag, thereby providing higher location accuracy in indoor environments.

[0005] In addition, CCTV-based video analysis technology is essential in security and industrial environments and is utilized to detect object movements and track specific targets. Common object detection and tracking techniques include YOLO (You Only Look Once), Faster R-CNN (Region-based Convolutional Neural Networks), SORT (Simple Online and Realtime Tracker), and DeepSORT (Deep Simple Online and Realtime Tracker).

[0006] While this video-based location tracking method can recognize objects with high accuracy within the camera's field of view, it has limitations such as difficulty in 3D location tracking due to insufficient depth (z-axis) information with a single camera, and inability to track objects when they are obscured by other objects. To address this, multi-view based location tracking techniques utilizing multiple cameras are being researched, and by employing triangulation techniques, 3D coordinates can be calculated by combining 2D coordinates obtained from multiple CCTVs. However, this method also has limitations in that it is difficult to track objects in areas where cameras are not installed.

[0007] Meanwhile, research is currently underway to develop more precise location tracking systems by integrating wireless positioning and video analysis. Combining Wi-Fi RTT-based distance information with CCTV-based object information can generate location data with higher reliability while compensating for the limitations of each technology. For example, after measuring a worker's approximate location using Wi-Fi RTT data, objects detected at that location can be accurately identified through CCTV video analysis.

[0008] However, existing studies typically processed sensor fusion on a central server, which presented the problem of requiring high network bandwidth and processing power to perform real-time computations. Additionally, the central server-dependent approach has the disadvantage of making real-time processing difficult in the event of network failures.

[0009] The present invention was devised to solve such problems and provides a method and apparatus capable of precisely determining real-time location based on on-device AI utilizing image and composite positioning information. To this end, it implements a technology that minimizes location errors by fusing Wi-Fi RTT-based wireless positioning information, CCTV video information, and sensor data within FTM tags, and applying an AI algorithm. Furthermore, it aims to enhance real-time performance and reliability by reducing dependence on a central server through on-device AI computation and enabling independent location tracking even in the event of network failures.

[0010] To achieve the above objective, a method for determining location based on on-device AI based on image and composite positioning information according to the present invention comprises: (a) a step of collecting one or more data for positioning; (b) a step of evaluating the reliability of the data collected in step (a); (c) a step of calculating a location by fusing the data evaluated in step (b) based on AI; and (d) a step of determining the location of a user based on the result calculated in step (c).

[0011] The data collected in step (a) above includes Wi-Fi RTT (Fine Time Measurement) based wireless positioning information, CCTV video information, and sensor information within the FTM tag.

[0012] The above Wi-Fi RTT-based wireless positioning information includes a method for calculating distance data by measuring the signal round-trip time between an FTM tag and multiple wireless access points (APs).

[0013] The above CCTV video information includes 2D coordinate and size information of an object, and includes a method for estimating the depth (z-axis) value of an object by applying a camera perspective projection technique.

[0014] The sensor information within the above FTM tag includes accelerometer and gyroscope data, and includes a method for performing position correction by analyzing the user's direction of movement and speed.

[0015] The method for evaluating data reliability in step (b) above includes Wi-Fi RTT signal quality analysis, pixel quality analysis of CCTV objects, and noise filtering of sensor data.

[0016] The above-mentioned Wi-Fi RTT signal quality analysis includes a method for calculating RSSI (Received Signal Strength Indicator) values, the variance and standard deviation of measurement data, the presence of multipath errors, and packet loss rates.

[0017] The reliability of the above Wi-Fi RTT measurement data is based on the Gaussian probability density function (PDF), and

[0018] The normalized probability density function (PDF) of the FTM measurement is,

[0019]

[0020] Here, v represents the variance of the FTM measurement, m represents the mean of the FTM measurement, and d represents the deviation between the FTM measurement and the reference value,

[0021] The FTM measurement reliability correction function is,

[0022]

[0023] Here, m rtt represents the RTT measurement value, and allows for the application of weights that reflect signal quality.

[0024] The pixel quality analysis of the above CCTV object includes a method of analyzing the object's contour recognition accuracy, pixel contrast, and difference value from the background (Feature Extraction).

[0025] The method of performing AI-based data fusion in step (c) above includes machine learning-based weight application, neural network-based data correction, and movement pattern learning using historical location data.

[0026] The above machine learning-based weight application includes a method of calculating the location by prioritizing the use of highly reliable data based on the results of a data reliability evaluation.

[0027] The above neural network-based data correction utilizes a Long Short-Term Memory (LSTM) model to compare continuous location data and predict movement paths to remove noise.

[0028] The movement pattern learning utilizing the aforementioned historical location data performs real-time location correction by analyzing patterns such as the user's average movement speed, direction of movement, and time spent within a specific area.

[0029] The final location information calculated in step (d) above is recorded in the on-device internal storage or transmitted to a central management server via a network.

[0030] When the above location data is transmitted to a central management server, the user's movement path is analyzed, and a warning is provided upon approaching a dangerous area.

[0031] In another aspect, the present invention for achieving such an objective is a computer program stored on a non-transient storage medium for on-device AI-based location determination based on image and composite positioning information, comprising: (a) a step of collecting one or more data for location determination; (b) a step of evaluating the reliability of the data collected in step (a); (c) a step of fusion of the data evaluated in step (b) based on AI to calculate a location; and (d) a command to execute a step of determining a user's location based on the result calculated in step (c).

[0032] Another aspect of the present invention for achieving such an objective is a device for determining location based on on-device AI based on image and composite positioning information, comprising: at least one processor; and at least one memory for storing computer-executable commands, wherein the computer-executable commands stored in the at least one memory enable the execution of the following steps by the at least one processor: (a) collecting one or more data for positioning; (b) evaluating the reliability of the data collected in step (a); (c) fusing the data evaluated in step (b) based on AI to calculate a location; and (d) determining the location of a user based on the result calculated in step (c).

[0033] According to the present invention, by providing a method and device for determining location using on-device AI based on images and composite positioning information, it is possible to overcome various limitations of existing location tracking technologies and provide improved location measurement accuracy. In the present invention, a method is applied to determine the user's location in real time by fusing Wi-Fi RTT (Fine Time Measurement)-based wireless positioning information, CCTV video information, and sensor data within an FTM tag using AI. Through this, it provides the advantage of being able to track location with high precision even in indoor environments compared to existing GPS or single wireless positioning technologies.

[0034] In particular, this invention adopts a method of directly processing and computing data using on-device AI. Conventional location tracking technologies typically processed data on a central server, which resulted in issues such as increased network load, data transmission delays, and limitations in real-time processing. However, this invention incorporates a location determination AI module within the FTM tag to directly calculate the user's location on the device and enables integration with a central server when necessary. This reduces network dependency and improves the system to allow for independent location tracking even in the event of network failures.

[0035] Furthermore, the present invention is designed to provide more accurate location information through data reliability evaluation and the application of AI-based weights. While existing location tracking technologies often simply utilize collected data as is, the present invention evaluates the reliability of each data point by applying techniques such as Wi-Fi signal quality evaluation (RSSI, multipath error analysis), CCTV object recognition reliability analysis (pixel contrast, analysis of difference values ​​from the background), and sensor data noise filtering. This enables AI computations to be performed based on data with high reliability. Through this, location errors are minimized, and the user's location can be determined with high precision.

[0036] In addition, the present invention includes a function for learning and predicting a user's movement patterns. Rather than simply performing static location tracking, it utilizes machine learning-based movement path learning models (RNN, LSTM, etc.) to analyze patterns such as the user's movement speed, direction, and time spent within a specific area, and enables real-time location correction. This allows for stable location tracking even if temporary errors in sensor data or signal interference occur, and enables consideration of future location changes through a movement path-based prediction function.

[0037] Furthermore, the present invention is designed to enable long-term location data analysis and tracking by including a function for storing and managing location data. User location information can be stored on the device and, when necessary, transmitted to a central management server to be integrated with a real-time monitoring system. This enables various applications, such as monitoring worker locations in industrial sites, controlling access within secure zones, and analyzing movement paths in logistics environments. Additionally, a function to automatically provide warnings upon approaching hazardous areas can be added, thereby enhancing connectivity with safety management systems.

[0038] In conclusion, by applying on-device AI-based location identification technology, the present invention provides effects such as higher accuracy, real-time capability, network independence, and improved data reliability compared to existing technologies, and can provide an innovative location tracking solution that can be efficiently utilized in various industrial and security environments.

[0039] FIG. 1 is a diagram illustrating a system for on-device AI-based location identification based on image and composite positioning information according to the present invention.

[0040] FIG. 2 is a diagram showing a sequence diagram for on-device AI-based location determination based on image and composite positioning information according to the present invention.

[0041] FIG. 3 is a flowchart illustrating an on-device AI-based location determination method based on image and composite positioning information according to the present invention.

[0042] FIG. 4 is a diagram showing a computer with an embedded on-device AI-based location identification module based on image and composite positioning information according to the present invention.

[0043] Hereinafter, embodiments according to the present invention will be described in detail with reference to the attached drawings. Identical or similar components are assigned identical or similar reference numerals, and redundant descriptions thereof are omitted. In describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art may obscure the essence of the embodiments disclosed in this specification, such detailed description is omitted. The attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification, and the technical concept disclosed in this specification is not limited by the attached drawings; it should be understood that they include all modifications, equivalents, or substitutions that fall within the spirit and technical scope of the present invention.

[0044] Terms containing ordinal numbers, such as first, second, etc., may be used to describe various components; however, these terms are used solely for the purpose of distinguishing one component from another, and the corresponding components are not limited by these terms. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0045] Terms such as “comprising,” “comprising,” or “having” as used herein should be understood as limiting the existence of the features, steps, components, or combinations thereof described herein, and are not intended to exclude the possibility that one or more other features, steps, components, or combinations thereof may exist or be added.

[0046] When a component is described as being "connected" or "joined" to another component, it should be understood that it may be directly connected or joined to that other component, or that there may be another component in between. On the other hand, when a component is referred to as being "directly connected" or "directly joined" to another component, it should be understood that no other component exists in between.

[0047] FIG. 1 is a diagram showing the schematic configuration of a system (10) for on-device AI-based location identification based on image and composite positioning information according to the present invention.

[0048] Referring to FIG. 1, the system (10) of the present invention includes an FTM tag (200), a plurality of CCTV devices (100-1, 100-2), and a plurality of wireless access points (AP, 300-1, 300-2), and has a structure in which the FTM tag (200) calculates the location in real time based on on-device AI.

[0049] The FTM tag (200), which plays a key role in the system (10) of the present invention, is a device that can be attached to a user's safety helmet or work clothes and includes the function of collecting wireless positioning information based on Wi-Fi RTT (Fine Time Measurement) and determining the location in real time by performing AI calculations independently. Unlike conventional location tracking systems that process data on a central server, the present invention is designed to determine the location by fusing wireless positioning information, image-based object location information, and sensor data within the tag directly within the FTM tag. This reduces network load and enables faster location identification even in environments requiring real-time performance.

[0050] The system (10) of the present invention includes a plurality of CCTV devices (100-1, 100-2) linked with an FTM tag (200), and these devices detect the movement of a user in video and track the location of an object. Each CCTV device can extract 2D coordinate and size information of an object from the captured video, and based on this, can estimate the depth (z-coordinate) of the object by applying perspective. When multiple CCTV devices are utilized, more precise 3D location tracking is possible using triangulation techniques. However, since there is a high possibility of location errors occurring with only video analysis, the present invention fuses wireless positioning information along with video information to derive a more reliable location.

[0051] Additionally, the system (10) of the present invention includes a plurality of wireless access points (AP, 300-1, 300-2) that communicate with the FTM tag (200). The access points can measure the signal round-trip time using the FTM tag and Wi-Fi RTT and calculate distance information based on this. When multiple APs are deployed, the initial 3D location of the FTM tag can be roughly estimated by applying a trilateration technique. However, Wi-Fi RTT-based positioning may be affected by the signal environment and may lack accurate direction information (z-axis information). Therefore, the present invention is designed to perform more accurate location tracking by fusing wireless positioning information and CCTV-based video information using AI. Additionally, GPS information may be used.

[0052] The system (10) of the present invention operates in the following manner. First, the FTM tag (200) communicates with a plurality of APs (300-1, 300-2) to obtain distance information based on Wi-Fi RTT. At the same time, the CCTV device (100-1, 100-2) collects video data of the user and extracts 2D location and size information of the object. Subsequently, the FTM tag (200) calculates an initial location based on Wi-Fi RTT data and performs a process of matching the user's ID and location by comparing it with CCTV object data detected at that location.

[0053] The FTM tag (200) contains on-device AI, which is used to fuse wireless positioning information, CCTV-based distance information, and user personal information collected by the FTM tag using AI to ultimately determine the user's 3D location. During the AI ​​computation process, the reliability of each data is analyzed to apply weights, and a function is performed to optimize the location by learning movement patterns. For example, Wi-Fi RTT signal quality (RSSI, multipath error, etc.) is evaluated, and the pixel quality of the CCTV video and whether objects are obscured are analyzed to perform location calculation based on data with high reliability.

[0054] Finally, the AI-based location information can be recorded in the internal storage of the FTM tag or transmitted via a network to a central management server or other devices. During this process, features such as analyzing the user's movement path and providing warnings when approaching specific risk zones can also be added.

[0055] FIG. 2 is a diagram showing a sequence diagram for on-device AI-based location determination based on image and composite positioning information according to the present invention. The system of the present invention includes a CCTV (100), a plurality of FTM tags (200-1, 200-2), and a plurality of wireless access points (300-1, 300-2), and operates in such a way that the FTM tags calculate the location in real time based on on-device AI. FIG. 2 illustrates the operation flow of such a system step by step and explains the information exchange and calculation process between each device.

[0056] First, in step S100, the FTM tag (200-2) collects personal information of the user. The personal information of the user (A) may include data such as a user ID, role, and access rights pre-registered in the FTM tag, thereby enabling accurate identification of the location of a specific user. This information is utilized in a subsequent location determination process.

[0057] Subsequently, in step S110, the CCTV (100) broadcasts and transmits the detected object information to an FTM tag. In this process, the CCTV detects an object in the captured video and extracts and transmits the 2D coordinates (u, v) and size information of the object. The FTM tag can analyze the video-based location information based on the received CCTV data and estimate the direction and speed of movement of the object.

[0058] In step S120, the FTM tag (200-2) collects video information. At this time, the FTM tag stores object information (B) received from the CCTV and enables real-time analysis of the object through on-device AI computation. The video information is used to estimate the distance (z value) between the object and the camera by applying a perspective projection technique.

[0059] In step S130, wireless positioning information is exchanged between the FTM tag (200-2) and multiple wireless access points (300-1, 300-2). The FTM tag measures the signal round-trip time with each AP using the Wi-Fi RTT (Fine Time Measurement) protocol and calculates distance data based on this. The APs (300-1, 300-2) transmit this data back to the FTM tag, and in this process, the FTM tag can calculate a more precise location by fusing information collected from multiple APs.

[0060] In step S140, the FTM tag (200-2) collects wireless positioning information. The wireless positioning information (C) collected here may include not only distance information to each AP, but also data such as signal quality (RSSI) and whether there are multipath errors. Through this, the FTM tag can analyze the wireless signal environment and determine the location based on the optimal signal data.

[0061] In step S150, a process of collecting sensor data inside the FTM tag (200-2) is performed. The FTM tag can detect the user's movement and analyze the direction and speed of movement by utilizing built-in sensors such as an accelerometer and a gyroscope. This sensor data (D) is fused with existing wireless positioning information and image information in the on-device AI to enable more precise location tracking.

[0062] Finally, in step S160, the on-device AI of the FTM tag performs a process of determining the user's location. The AI ​​computation is performed by fusing collected information (A: user personal information, B: CCTV video information, C: wireless positioning information, D: sensor information within the FTM tag). The AI ​​model determines the optimal 3D location by analyzing the reliability of each data point, applying weights, and learning movement patterns using machine learning. In this process, location tracking with minimized error is possible by comprehensively considering Wi-Fi RTT signal quality, the reliability of video objects, and sensor data.

[0063] Unlike conventional location tracking methods, the S160 process of the present invention is differentiated in that it enables real-time location determination without relying on a central server by directly processing data and performing calculations based on on-device AI. Furthermore, by integrating Wi-Fi RTT, CCTV object recognition, and sensor data, it can compensate for the limitations of each technology and provide a location tracking system with higher precision and reliability.

[0064] In conclusion, the S160 process is a core step that determines the user's location in real time by fusing A (user personal information), B (video information), C (wireless positioning information), and D (sensor information within the FTM tag) based on AI and calculating the optimal location. By applying this AI-based fusion technique, the present invention is designed to overcome the limitations of existing location tracking technologies and provide a precise location tracking solution that can be utilized in various industrial and security environments.

[0065] FIG. 3 is a flowchart illustrating an on-device AI-based location determination method based on image and composite positioning information according to the present invention, showing the on-device AI-based location determination process for process S160 of FIG. 2.

[0066] In FIG. 3, various data collected by the FTM tag (200) is fused based on AI, and a calculation is performed to derive an optimized result after precisely estimating the user's location. Through this, the user's 3D location can be identified in real time, and it is designed to provide high precision and reliability compared to existing technologies.

[0067] First, the FTM tag (200) collects data from various sources (S161). Here, the user's personal information (A) includes data such as user ID, job information, and access rights, and is used to identify the user's identity. The video information (B) includes 2D coordinate and size information of an object provided by the CCTV (100), and is used to estimate the location in conjunction with the wireless positioning information collected by the FTM tag. The wireless positioning information (C) includes Wi-Fi RTT (Fine Time Measurement) based distance data and is calculated based on the signal round-trip time between the FTM tag and the wireless AP (300-1, 300-2). Finally, the sensor information (D) within the FTM tag includes data such as an accelerometer and gyroscope, and is used to determine the user's direction of movement and speed.

[0068] When data is collected, the on-device AI inside the FTM tag performs a process of evaluating data reliability (S162). In the present invention, rather than simply combining multiple data, the reliability of each data is analyzed and appropriate weights are assigned to enable more accurate location calculation. For example, the reliability of the Wi-Fi RTT signal quality analysis can be evaluated based on RSSI (signal strength), the variance and standard deviation of the measurement data, the presence of multipath errors, and the method of calculating the packet loss rate. Here, the reliability of the Wi-Fi RTT measurement data is evaluated using a Gaussian probability density function (PDF), and the normalized probability density function (PDF) of the FTM measurement value is as follows [Equation 1].

[0069] [Formula 1]

[0070]

[0071] Here, v represents the variance of the FTM measurement, m represents the mean of the FTM measurement, and d represents the deviation between the FTM measurement and the reference value. The reliability correction function for the FTM measurement is as follows [Equation 2].

[0072] [Equation 2]

[0073]

[0074] Here, m rtt This refers to the RTT measurement value, which allows for the application of weights reflecting signal quality. Furthermore, the reliability of CCTV object data is analyzed by considering factors such as image quality and whether the object is occluded. Additionally, sensor reliability is evaluated by comparing internal FTM tag sensor data with the user's movement patterns. This reliability analysis process is differentiated from existing technologies in that it is designed to minimize data errors and enable more precise position calculations.

[0075] Next, the on-device AI of the FTM tag performs a process (S163) of fusion of reliability-evaluated data to calculate the optimal user location. In the present invention, a machine learning-based data fusion technique is utilized to analyze the characteristics of each data and is designed to estimate the 3D location based on the most reliable data. In this process, a triangulation technique can be applied by combining distance information calculated based on Wi-Fi RTT and 2D coordinate information of an object acquired from a CCTV, and an AI model is used to learn movement patterns by comparing the location data of the previous frame with the current data. Through this, more precise user location calculation is possible than a simple positioning-based location tracking method, and optimized results can be derived based on data reliability.

[0076] Finally, once the AI-based location calculation is completed, a process (S164) is performed to determine the user's 3D location and, if necessary, store it or transmit it to an external system. The FTM tag can internally store location data and utilize it to analyze the user's movement path and detect behavioral patterns in specific areas. Additionally, the system of the present invention can transmit location data to a central management server via a network, thereby providing a function to monitor and analyze the user's location in real time.

[0077] The AI-based computation performed during the S160 process may go beyond simple location calculation and include the function of learning and predicting the user's movement path. For example, the AI ​​model of the present invention may apply movement pattern analysis techniques based on LSTM (Long Short-Term Memory) or RNN (Recurrent Neural Network), and through this, it can provide the function of analyzing the user's movement speed, direction, and past movement data to perform movement prediction in real time and correct the location. This movement pattern prediction function is a differentiating factor from existing simple positioning methods and enables more precise location management in environments such as industrial sites, logistics centers, and smart buildings.

[0078] FIG. 4 is a diagram showing the schematic hardware configuration of a computer (200) with an on-device AI-based location identification module embedded based on image and composite positioning information according to the present invention. In the present invention, the FTM tag (200) is a core device that performs real-time location calculations based on on-device AI, and FIG. 4 describes the main components inside the FTM tag that perform such calculations.

[0079] The FTM tag (200) is a computer system including a position determination AI module (230) and has a hardware architecture including a processor (110), a storage unit (120), a memory (130), a communication unit (140), and an OS (operating system). These components are interconnected via an internal bus and perform real-time position calculation and data processing.

[0080] First, the processor (110) serves as the central computing unit of the system and is responsible for performing operations such as AI-based location calculation, data fusion, and reliability evaluation. In the present invention, the functions of a conventional simple FTM tag are expanded to enable real-time analysis of Wi-Fi RTT-based positioning information, image information, and sensor data through on-device AI, and to optimize the location using machine learning algorithms. Accordingly, the processor (110) may include a high-performance embedded processor or an AI accelerator (NPU, Neural Processing Unit) to perform such complex operations.

[0081] The storage unit (120) performs the function of storing location data, user information, AI training data, etc. The storage unit (120) may be composed of flash memory, SSD, or other non-volatile storage devices and may store model parameters and past location data required for AI computations performed inside the FTM tag. In particular, since it is necessary to refer to past data to analyze the user's movement path in the present invention, the storage unit (120) may be configured to store movement trajectory data, signal quality data, AI-based location correction results, etc.

[0082] The memory (130) serves as the execution memory of the system and provides a temporary storage space for AI computation and data processing. Since the FTM tag of the present invention must fuse and analyze multiple data in real time, it requires a memory capacity of a certain level or higher to provide high processing speed. The memory (130) includes an operating system (OS), an AI-based positioning module (230), a data buffer, etc., thereby supporting real-time computation.

[0083] In particular, the location determination AI module (230) is a software and hardware module that performs core operations of the present invention and plays a role in determining the optimal location of a user by fusing various data. The module analyzes Wi-Fi RTT signals, CCTV-based object information, FTM tag internal sensor data, etc., based on AI, and calculates the optimal location value after evaluating the reliability of the data by applying a machine learning model. The location determination AI module (230) can learn the user's movement patterns by utilizing deep learning models such as LSTM (Long Short-Term Memory) and RNN (Recurrent Neural Network), and through this, it can be configured to enable prediction-based location correction beyond simple location tracking.

[0084] The communication unit (140) is responsible for exchanging data between the FTM tag and an external system and can support various wireless communication methods such as Wi-Fi, BLE (Bluetooth Low Energy), and UWB (Ultra-Wideband). The FTM tag of the present invention must not only measure distance by performing Wi-Fi RTT-based communication with a wireless access point (AP), but also be able to share data with a central server or other devices as needed. To this end, the communication unit (140) may include a wireless module that supports multiple protocols, and in certain situations, may share location data by communicating directly with other FTM tags or CCTVs.

[0085] All components of the system are connected via an internal bus and are controlled by an operating system (OS). The OS performs management of the entire system and resource allocation, and is designed to maintain an efficient data flow between the processor (110), memory (130), storage unit (120), and communication unit (140). Additionally, since the OS of the present invention must perform real-time data processing, a lightweight RTOS (Real-Time Operating System) or an embedded OS supporting Edge AI can be utilized.

[0086] In conclusion, the FTM tag (200) illustrated in FIG. 4 operates as an independent computing device capable of performing real-time location calculations based on on-device AI, rather than a conventional simple wireless positioning device. The present invention is distinguished from existing technologies in that it performs high-precision location tracking by fusing Wi-Fi RTT, video analysis, and sensor data, and is designed to enable AI-based location calculations on the FTM tag itself without a central server. Through this, the system of the present invention can be utilized in various environments such as industrial sites, smart factories, and security zones, and can provide high-precision real-time location tracking.

[0087] As described above, although the present invention has been explained by limited embodiments and drawings, the present invention is not limited thereto, and it is obvious that various modifications and variations are possible within the scope of the technical spirit of the present invention and the equivalent scope of the claims described below by those skilled in the art to which the present invention belongs.

Claims

1. A method for determining location based on on-device AI based on image and composite positioning information, (a) A step of collecting one or more data for positioning; (b) a step of evaluating the reliability of the data collected in step (a) above; (c) a step of calculating the location by fusing the data evaluated in step (b) based on AI; and, (d) A step of determining the user's location based on the result calculated in step (c) above. On-device AI-based location determination method based on images and composite positioning information including 2. In Claim 1, The data collected in step (a) above includes Wi-Fi RTT (Fine Time Measurement) based wireless positioning information, CCTV video information, and sensor information within the FTM tag. An on-device AI-based location determination method based on images and composite positioning information characterized by 3. In Claim 2, The above Wi-Fi RTT-based wireless positioning information calculates distance data by measuring the signal round-trip time between an FTM tag and multiple wireless access points (APs). An on-device AI-based location determination method based on images and composite positioning information characterized by 4. In Claim 2, The above CCTV video information includes 2D coordinate and size information of an object, and estimates the depth (z-axis) value of the object by applying a camera perspective projection technique. An on-device AI-based location determination method based on images and composite positioning information characterized by 5. In Claim 2, The sensor information within the above FTM tag includes accelerometer and gyroscope data, and performs position correction by analyzing the user's direction of movement and speed. An on-device AI-based location determination method based on images and composite positioning information characterized by 6. In Claim 1, The method for evaluating data reliability in step (b) above includes Wi-Fi RTT signal quality analysis, pixel quality analysis of CCTV objects, and noise filtering of sensor data. An on-device AI-based location determination method based on images and composite positioning information characterized by 7. In Claim 6, The above Wi-Fi RTT signal quality analysis includes a method for calculating RSSI (Received Signal Strength Indicator) values, the variance and standard deviation of measurement data, the presence of multipath errors, and packet loss rates, and The reliability of the above Wi-Fi RTT measurement data is based on the Gaussian probability density function (PDF), and The normalized probability density function (PDF) of the FTM measurement is, Here, v represents the variance of the FTM measurement, m represents the mean of the FTM measurement, and d represents the deviation between the FTM measurement and the reference value, The FTM measurement reliability correction function is, Here, m rtt refers to the RTT measurement value, to which weights reflecting signal quality can be applied. An on-device AI-based location determination method based on images and composite positioning information characterized by 8. In Claim 6, The pixel quality analysis of the above CCTV object involves analyzing the object's contour recognition accuracy, pixel contrast, and the difference value from the background (Feature Extraction). An on-device AI-based location determination method based on images and composite positioning information characterized by 9. In Claim 1, The method of performing AI-based data fusion in step (c) above includes machine learning-based weight application, neural network-based data correction, and movement pattern learning using historical location data. An on-device AI-based location determination method based on images and composite positioning information characterized by 10. In Claim 9, The above machine learning-based weight application calculates the location by prioritizing the use of highly reliable data based on the results of the data reliability evaluation. An on-device AI-based location determination method based on images and composite positioning information characterized by 11. In Claim 9, The above neural network-based data correction utilizes a Long Short-Term Memory (LSTM) model to compare continuous location data and predict movement paths to remove noise. An on-device AI-based location determination method based on images and composite positioning information characterized by 12. In Claim 9, The movement pattern learning utilizing the aforementioned historical location data analyzes patterns such as the user's average movement speed, direction of movement, and time spent within a specific area to perform real-time location correction. An on-device AI-based location determination method based on images and composite positioning information characterized by 13. In Claim 1, The final location information calculated in step (d) above includes a method in which it can be recorded in an on-device internal storage or transmitted to a central management server via a network. An on-device AI-based location determination method based on images and composite positioning information characterized by 14. In Claim 13, When the above location data is transmitted to a central management server, the user's movement path is analyzed, and a warning is provided upon access to a danger zone. An on-device AI-based location determination method based on images and composite positioning information characterized by 15. A program that determines location based on on-device AI based on image and composite positioning information, (a) A step of collecting one or more data for positioning; (b) a step of evaluating the reliability of the data collected in step (a) above; (c) a step of calculating the location by fusing the data evaluated in step (b) based on AI; and, (d) A step of determining the user's location based on the result calculated in step (c) above. A computer program stored on a non-transient storage medium for on-device AI-based location determination based on image and composite positioning information, including a command to cause to be executed.

16. A device that determines location based on on-device AI based on image and composite positioning information, At least one processor; and It includes at least one memory that stores computer-executable instructions, The computer-executable instruction stored in the above at least one memory is, by the above at least one processor, (a) A step of collecting one or more data for positioning; (b) a step of evaluating the reliability of the data collected in step (a) above; (c) a step of calculating the location by fusing the data evaluated in step (b) based on AI; and, (d) A step of determining the user's location based on the result calculated in step (c) above. On-device AI-based location finding device based on image and composite positioning information that enables execution.

Images