A lost property finding system and method

Abstract

This invention relates to the fields of data processing and the Internet of Things (IoT), specifically disclosing a lost and found system and method. The user end collects and uploads images of items; the server end extracts feature vectors of the items using a convolutional neural network with an attention mechanism, employing a hierarchical strategy of "type coarse screening + feature fine screening" for matching, and combining reward management and order management modules to freeze and distribute rewards and record service processes; the offline server end is responsible for item verification. This invention can focus on key features of items, reduce comparison complexity, build a complete online and offline service loop, prevent fraudulent uploads, protect user privacy, and has a modular design with strong scalability.

Description

Technical Field

[0001] This invention relates to the fields of data processing and Internet of Things (IoT) technology, specifically to a lost item retrieval system and method. Background Technology

[0002] Existing lost and found management systems mainly fall into several categories: QR code-based anti-loss systems, Bluetooth-based anti-loss devices, and various lost and found platforms. These existing technologies share the following common drawbacks in application: Inaccurate feature extraction: When dealing with small items such as keys, existing image comparison technologies generally have an accuracy rate of less than 70% due to differences in shooting angle and lighting, which cannot meet the needs of practical applications.

[0003] Lack of offline verification loop: Purely online solutions cannot verify "whether it is really lost" or "whether it has really been returned", making it difficult to connect with insurance claims or offline service processes.

[0004] The contradiction between incentives and abuse: The "upload as a reward" mechanism set up to incentivize finders to return the finders has led to a large number of invalid uploads, which consume server resources and incur manual review costs.

[0005] Privacy versus convenience: Directly exposing contact information poses privacy risks, but complete anonymity makes it impossible to contact the owner of the lost item.

[0006] Therefore, a lost item retrieval system and method has become an urgent problem to be solved. Summary of the Invention

[0007] The purpose of this invention is to provide a lost and found system and method that aims to solve the problems of low accuracy, poor efficiency, disconnect between online and offline methods, imperfect incentive mechanisms, and insufficient privacy protection in existing lost and found methods, so as to achieve rapid and accurate retrieval of lost items, while reducing service costs and protecting user privacy.

[0008] To achieve the above objectives, the technical solution provided by this invention is: a lost item retrieval system, comprising: The user-side, including the owner's side and the pick-up side, is used for collecting and uploading images of items; The server-side includes: an image processing module for receiving images of the items and extracting feature vectors of the items using a convolutional neural network with an attention mechanism; a matching module for performing a coarse screening based on item type, matching by calculating the similarity between feature vectors, and outputting the matching results; a reward management module for freezing reward funds after a successful match and distributing reward funds after offline verification and confirmation; and an order management module for generating service orders and recording service processes. The offline service terminal consists of multiple franchisee terminals connected to the server network, used to scan verification codes, collect verification images, and perform offline verification of items.

[0009] Furthermore, the image processing module includes: The preprocessing unit is used to denoise and normalize the object image; The feature extraction network introduces an attention layer after the convolutional layer to focus on the local feature regions of the object and output a feature vector of fixed dimension. The feature storage unit is used to associate and store feature vectors with user IDs and to establish vector indexes.

[0010] Furthermore, the matching module employs a hierarchical retrieval strategy, including: The coarse screening unit is used to narrow down the comparison range based on the type of item. The fine screening unit is used to perform an approximate nearest neighbor search based on vector index within the narrowed range and calculate the cosine similarity between feature vectors. The dynamic threshold adjustment unit is used to automatically adjust the similarity threshold according to the item type. When the similarity is higher than the threshold, the match is considered successful.

[0011] Furthermore, the reward management module includes a state machine, which maintains the reward status, including untriggered status, matched and frozen status, verification pending confirmation status, issued status, and expired status. The reward management module is also used to place the reward money in the matched frozen state after a successful match, and to switch the reward money to the issued state after receiving an offline verification confirmation signal.

[0012] Furthermore, the order management module is also used for: A service record is automatically generated upon completion of the service. The service record includes the service time, service location, item image, and identity verification information. In response to the confirmation signal sent by the owner, the system triggers the automatic claim form generation interface of the insurance system, submitting the service record as the basis for the claim.

[0013] The present invention also provides a method for recovering lost items, applied to the aforementioned lost item recovery system, comprising the following steps: S1. Evidence preservation steps: The owner of the lost item collects and uploads multi-angle images of the item through the owner's terminal, and the server extracts and stores the feature vector of the item. S2. Matching steps: The picker uploads images of the picked items through the picking device. The server performs a rough screening based on the item type and then performs matching based on feature vector similarity calculation. S3. Verification steps: After a successful match, the server generates a verification code and freezes the reward money, guiding the picker to deliver the item to the offline server. The offline server scans the verification code and collects the verification image. The server then performs a second comparison between the verification image and the feature vector of the evidence. S4. Settlement Steps: After the second comparison is successful, the server will issue the reward to the picker's account and settle the service fee with the offline server, generating a service record.

[0014] The advantages of this invention compared to the prior art are: This invention employs an improved convolutional neural network with an attention mechanism, which can automatically focus on key local features of an object and avoid interference from irrelevant backgrounds. At the same time, pre-training based on multi-condition samples makes the system more robust to changes in object rotation, scaling, and lighting.

[0015] This invention adopts a hierarchical retrieval strategy of "coarse screening by item type + fine screening by feature vector index", which optimizes the traditional linear comparison into hierarchical retrieval and greatly reduces the comparison complexity.

[0016] This invention introduces offline franchisees as verification and delivery nodes, constructing a complete service loop of "online evidence storage - intelligent matching - offline verification - automatic settlement," thus solving the problem of online and offline disconnect in existing technologies.

[0017] This invention adopts a mechanism of "successful matching + offline verification and confirmation" before issuing rewards, which effectively constrains the uploading behavior of pickers, reduces the proportion of invalid and fake uploads, greatly reduces the storage pressure and computing power consumption of the server, and effectively prevents the behavior of fraudulently obtaining rewards from a technical point of view, ensuring the rationality and sustainability of the incentive mechanism.

[0018] This invention uses item feature vectors to replace the original images for core comparison and storage, avoiding privacy and security issues caused by the leakage of original images and effectively protecting user privacy. At the same time, the system adopts a modular design, with each module working independently and cooperating with each other, possessing high scalability and allowing for flexible addition of functional modules according to user needs. Attached Figure Description

[0019] Figure 1 This is a system block diagram of a lost item retrieval system according to the present invention.

[0020] Figure 2 This is a flowchart of a lost item retrieval method according to the present invention. Detailed Implementation

[0021] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0022] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0023] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0024] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0025] The following detailed description of a lost item retrieval system and method according to the present invention, with reference to the accompanying drawings, provides further insight.

[0026] Combined with appendix Figure 1-2 The specific implementation process of the lost item retrieval system and method of the present invention is as follows: A lost and found system includes a user terminal, a server terminal, and an offline service terminal; The user interface includes both the owner's side and the finder's side, both of which support login via mobile or computer terminals. They are primarily used for uploading evidence of lost items, uploading found items, and confirming service details. Specific functions are as follows: The owner's app allows owners to perform actions such as initial registration of personal belongings, viewing matching results, receiving verification codes, confirming item collection, initiating claims, and purchasing insurance. Pickup app: Allows pickers to upload images of picked items, receive matching notifications, view verification codes, deliver items to offline service centers, and receive rewards.

[0027] The server is the core control unit of the system, establishing communication connections with the user end and offline servers via the network. It includes image processing, matching, reward management, and order management modules. Each module is functionally independent yet works collaboratively. Image processing module: Used to receive images of items uploaded by users, preprocess the images, extract features, and store features to provide core data support for subsequent matching; Matching module: Used to perform hierarchical retrieval and comparison of images of picked items uploaded by the pickup terminal with the characteristics of items stored by the owner, to achieve accurate matching of items and return the matching results; Rewards Management Module: Used to manage the freezing and distribution of rewards and the settlement of franchisee service fees, standardize the incentive mechanism, and prevent the abuse of rewards; Order Management Module: Used to automatically generate service records and claim forms, enabling full-process order management and automatic claim processing.

[0028] The offline service terminals are multiple franchisee terminals connected to the service terminal network. Franchisees include designated locksmith shops, convenience stores, community service centers and other offline entities. They are mainly used for item verification, identity verification and service delivery, and are the key nodes connecting online matching and offline delivery.

[0029] This invention also provides a method for recovering lost items based on the above-mentioned lost item recovery system. The method realizes the entire process of lost item recovery through the following steps, which are clear, easy to operate, and are detailed below: S1: Item Storage Evidence The owner of lost property performs initial evidence preservation of their personal belongings through the owner's app. This process requires the user to take images of the item from at least three different angles, ensuring the images fully capture the item's key features. After receiving the images uploaded by the user, the server-side image processing module performs the following operations in sequence: S11. Preprocessing: The received image is denoised and normalized to remove noise interference, unify image size and pixel specifications, and provide standardized image data for subsequent feature extraction. S12. Feature Extraction: An improved convolutional neural network with an attention mechanism is used to extract features from the preprocessed image. This neural network automatically focuses on the key local feature regions of the object through the attention mechanism, ignoring irrelevant background interference. At the same time, the network is trained based on more than 100,000 pre-trained multi-condition samples, which has strong robustness to image rotation, scaling and illumination changes, and can accurately extract the stable features of the object. S13. Feature Storage: The extracted item feature vectors are associated with the owner's user ID and stored in the system's feature vector database, with an index created for easy and quick retrieval later. The original images uploaded by users are only for display purposes and do not participate in the core comparison process, effectively protecting user privacy.

[0030] S2: Pickup Matching The scavenger uploads images of the collected items via the scavenging app. When uploading, the scavenger can select the item type. The server-side matching module receives the images and executes a hierarchical search strategy to achieve fast and accurate matching. The specific operation is as follows: S21. Coarse screening: First, filter according to the type of item, group the picked items and the same type of evidence items into a subset, greatly narrowing the comparison range and reducing the comparison complexity; S22. Fine Screening: In the subset of items after coarse screening, a precise comparison is made by calculating the cosine similarity between the feature vectors of the picked items and the feature vectors of the stored items; the comparison process adopts a dynamic threshold adjustment algorithm, which automatically adjusts the similarity threshold according to the item type; S23. Result Return: If there is a similarity item with a similarity higher than the set threshold, the match is considered successful. The server immediately sends a matching notification to both the owner and the finder, informing them of the matching result and subsequent operation procedures. If the match fails, the feature vector of the found item is stored in the matching pool, waiting for new items of the same type to be found and compared again.

[0031] S3: Reward and Verification Mechanism Upon successful matching, the server-side reward management module is triggered, and the owner's pre-set reward automatically enters a "frozen" state. The reward cannot be withdrawn at this time and will only be unfrozen and distributed after offline verification and item delivery are completed. Simultaneously, the system generates a unique verification code containing the item's unique identifier and matching information, and sends it to both the pickup point and the owner's end, serving as the core credential for offline verification. The specific process is as follows: S31. Item Delivery Guidance: Based on the picker's geographical location, the system recommends the nearest offline service terminal and guides the picker to deliver the picked-up items to the franchisee. S32. Offline Verification Procedure: Franchisees initiate the verification process by scanning the verification code provided by the finder through an offline service terminal, and then perform the following operations in sequence: Identity verification: The identity information of the finder is confirmed through the facial recognition function of the terminal to ensure that the finder's identity is authentic and valid; Item verification: The franchisee takes another picture of the picked-up item. The image processing module on the server extracts features from the image and compares it with the feature vector of the item in the owner's certificate to confirm that the picked-up item is consistent with the item in the owner's certificate, thus avoiding fraudulent or wrongful claims. S33. Service Delivery: After the items are verified to be correct, the system sends a collection notification to the owner. The owner can go to the franchisee to collect the items, or the franchisee can provide extended services such as key duplication depending on the type of items. S34. Reward and Service Fee Settlement: After the owner confirms receipt of the item, they submit a receipt confirmation instruction to the system. The reward management module immediately distributes the frozen reward funds to the finder's account. At the same time, the system automatically calculates the franchisee's service fee and completes the settlement, achieving a win-win situation for all parties.

[0032] S4: Order and Claims Automation Once the item is retrieved or the extended service is delivered, the order management module on the server automatically generates a complete service record. This record includes evidence of loss, matching process, verification results, service content, reward amount, and service fee, ensuring full traceability of the service process. If the owner purchased relevant lost and found insurance during the evidence collection process, the system will automatically generate a claim form and submit it to the insurance system using the service record, verification data, and item information as the basis for the claim, automating the claims process without manual intervention.

[0033] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A lost item retrieval system, characterized in that, include: The user-side, including the owner's side and the pick-up side, is used for collecting and uploading images of items; The server-side includes: an image processing module for receiving images of the items and extracting feature vectors of the items using a convolutional neural network with an attention mechanism; a matching module for performing a coarse screening based on item type, matching by calculating the similarity between feature vectors, and outputting the matching results; a reward management module for freezing reward funds after a successful match and distributing reward funds after offline verification and confirmation; and an order management module for generating service orders and recording service processes. The offline service terminal consists of multiple franchisee terminals connected to the server network, used to scan verification codes, collect verification images, and perform offline verification of items.

2. The lost item retrieval system according to claim 1, characterized in that: The image processing module includes: The preprocessing unit is used to denoise and normalize the object image; The feature extraction network introduces an attention layer after the convolutional layer to focus on the local feature regions of the object and output a feature vector of fixed dimension. The feature storage unit is used to associate and store feature vectors with user IDs and to establish vector indexes.

3. The lost item retrieval system according to claim 2, characterized in that: The matching module employs a hierarchical retrieval strategy, including: The coarse screening unit is used to narrow down the comparison range based on the type of item. The fine screening unit is used to perform an approximate nearest neighbor search based on vector index within the narrowed range and calculate the cosine similarity between feature vectors. The dynamic threshold adjustment unit is used to automatically adjust the similarity threshold according to the item type. When the similarity is higher than the threshold, the match is considered successful.

4. A lost item retrieval system according to claim 3, characterized in that: The reward management module includes a state machine, which maintains the reward status, which includes untriggered status, matched and frozen status, verification pending confirmation status, issued status, and expired status. The reward management module is also used to place the reward money in the matched frozen state after a successful match, and to switch the reward money to the issued state after receiving an offline verification confirmation signal.

5. A lost item retrieval system according to claim 4, characterized in that: The order management module is also used for: A service record is automatically generated upon completion of the service. The service record includes the service time, service location, item image, and identity verification information. In response to the confirmation signal sent by the owner, the system triggers the automatic claim form generation interface of the insurance system, submitting the service record as the basis for the claim.

6. A method for recovering lost items, applied to the lost item recovery system according to any one of claims 1 to 5, characterized in that, Includes the following steps: S1. Evidence preservation steps: The owner of the lost item collects and uploads multi-angle images of the item through the owner's terminal, and the server extracts and stores the feature vector of the item. S2. Matching steps: The picker uploads images of the picked items through the picking device. The server performs a rough screening based on the item type and then performs matching based on feature vector similarity calculation. S3. Verification steps: After a successful match, the server generates a verification code and freezes the reward money, guiding the picker to deliver the item to the offline server. The offline server scans the verification code and collects the verification image. The server then performs a second comparison between the verification image and the feature vector of the evidence. S4. Settlement Steps: After the second comparison is successful, the server will issue the reward to the picker's account and settle the service fee with the offline server, generating a service record.

Images