Airport passenger abnormal luggage processing method and system based on hybrid retrieval strategy

Abstract

The invention relates to an airport passenger abnormal luggage processing method and system based on a hybrid retrieval strategy, and the method comprises the following steps: S1, constructing a hybrid retrieval knowledge base for obtaining historical data which comprises a claim process and a compensation standard when the historical luggage of an airport is lost, damaged and delayed, and synchronously establishing a luggage compensation standard model; s2, performing mixed strategy retrieval analysis based on the acquired data, and formulating compensation standards and processes; s3, luggage case data are obtained through multiple technical means and used for carrying out danger level evaluation on the luggage case, due to the fact that the obtained data features are sufficient, feature extraction can be carried out on the obtained data through a log table, a perfect luggage case approximate value evaluation model can be established through obtaining and storing original data, and the risk level evaluation accuracy of the luggage case is improved. The method is used for establishing a database of luggage cases passing through an airport security check system in a current period and providing basic data support for subsequent future reference and compensation.

Description

Technical Field

[0001] The present invention relates to the technical field of public safety, and particularly to an abnormal luggage processing method and system for airport passengers based on a hybrid retrieval strategy. Background Art

[0002] During the operation of civil aviation airports, abnormal luggage incidents of passengers occur frequently, including luggage loss, damage, delay, etc. When such incidents occur, airport staff need to quickly establish cases for handling abnormal luggage for passengers and handle claims according to different rules of each airline. However, there are many problems currently faced: on the one hand, the processes, standards, and claim policies for handling abnormal luggage by different airlines vary greatly, and it is difficult for staff to comprehensively and accurately remember and apply this complex information; on the other hand, traditional processing methods rely on manual access to a large number of paper or electronic document materials, the case establishment process is cumbersome, the claim calculation and processing are time-consuming, the efficiency is low, it is easy to cause passengers to wait for a long time, which greatly affects the travel experience of passengers, and at the same time increases the airport operation and management costs. Therefore, there is an urgent need for an innovative technical means to integrate the information of each airline and help airport staff achieve rapid and accurate case establishment and claim handling for abnormal luggage. Summary of the Invention

[0003] The purpose of the present invention is to provide an abnormal luggage processing method and system for airport passengers based on a hybrid retrieval strategy to solve the problems existing in the above-mentioned prior art.

[0004] The above technical purpose of the present invention is achieved through the following technical solutions: An abnormal luggage processing method for airport passengers based on a hybrid retrieval strategy includes the following steps: Step S1, constructing a hybrid retrieval knowledge base for obtaining historical data, where the historical data includes claim processes and compensation standards when there are historical luggage losses, damages, and delays at the airport, and simultaneously establishing a luggage compensation standard model, which is used to determine the claim process and compensation standard according to the compensation specification documents of different airlines; Step S2, establishing a relational knowledge base, obtaining the basic data of the luggage through internal scanning of the luggage and saving it in the form of pictures, marking and storing data for each luggage, and simultaneously extracting feature data, where the data marking is to mark the luggage size information, luggage color information, and weighted information; performing hybrid strategy retrieval analysis on the pictures to determine whether there are dangerous objects, if so, issuing an alarm, if not, not issuing an alarm; Step S3, establishing a vector knowledge base, setting ultrasonic detectors and temperature detectors according to regions, performing hybrid inspection and obtaining hybrid features of the luggage through the data of the ultrasonic detectors and temperature detectors, for improving the basic data information of the compensation system, and presetting the compensation standard for each luggage according to the compensation specification documents of different airlines; Step S4, formulate a hybrid retrieval strategy, which includes: establishing a feature database for acoustic wave amplitude and frequency; establishing a database for the relationship between the size and weight of the luggage, judging the danger level of the luggage by retrieving the multi-dimensional feature data of the luggage, and adjusting the retrieval strategy through training.

[0005] By adopting the above technical solution, because the obtained data features are sufficient enough, the features of the obtained data can be extracted through the log table and used to establish a database of the luggage passing through the airport security system during the current period, providing basic data support for subsequent reference and compensation; through this process, the collection of the specification policy documents related to the abnormal luggage of different airlines and the preprocessing of the specification document data are carried out. Among them, the collection of policy documents can regularly obtain unstructured documents such as the "Baggage Transportation Terms", "Abnormal Handling Manual", and "Baggage Compensation Standard" of each airline through channels such as the official websites of each airline, internal documents, and industry specification documents. Then, it is necessary to parse and process the collected specification document data. By deploying the localized open-source tool OmniParse, the unstructured document data such as the specification documents of each airline collected are parsed into structured and operable data, and a storage structure of airline-abnormal type-related policy triples is designed with the airline as the primary key, and the parsed document data is stored in a relational database (such as MySQL).

[0006] In a further embodiment, step S3 further includes: Step S31, obtain the frequency band of the internal objects of the luggage through an ultrasonic detector; Step S32, obtain the internal thermal radiation distribution map of the luggage through a temperature detector; Step S33, a weighing module is arranged on the conveyor belt of the X-ray detection device, and ranging gratings are arranged at both the inlet and outlet ends of the X-ray. Whether the luggage leaves and enters the conveyor belt is judged through the ranging gratings, and the weighing of the luggage is realized through the mutual cooperation of the weighing module and the ranging gratings.

[0007] Traditional X-ray scanning of luggage can obtain a floor plan of the inside of the luggage. From the floor plan, it can be roughly seen whether there are dangerous goods inside the luggage according to the different brightness. However, with the development of technology, there are more and more ways to avoid X-ray scanning. Therefore, this application proposes a hybrid strategy based on X-ray scanning for evaluating the danger level of luggage. By substituting the data of the ultrasonic detector and the temperature detector into the hybrid retrieval strategy, the safety level value of the luggage can be quickly obtained to assist the personnel in judging whether the luggage needs to be opened for inspection.

[0008] By adopting the above technical solution, the X-ray module obtains the scan imaging plan view directly above the luggage, the scan imaging plan view on the left side, and the scan imaging plan view on the right side through multi-dimensional scanning. X-rays have strong penetration ability and can penetrate low-density substances such as clothes and plastics, but will be absorbed by high-density substances such as metals and liquids. Different substances have different absorption degrees of X-rays, forming light and dark differences. The absorption of X-rays by substances is related to the atomic number and density. High-density substances (such as metals) absorb more X-rays and appear dark; low-density substances (such as clothes and plastics) absorb less and appear light. Therefore, the top view, left view, and right view of the luggage can be obtained, and the grayscales of the three views in different directions are also different. Through 3D modeling and comparison, the three-dimensional shape of the items is restored. The restoration steps are as follows: extend the view in the acquisition direction according to the grayscale of the plan view. For example, the scan image above the luggage is extended downward according to the grayscale. Set the grayscale of a certain pixel point in the plan view as P. Obtain the length X, width Y, and height H of the luggage through the ranging grating. The grayscale is linearly mapped by Gray = (R + G + B) / 3, where R, G, and B are the values of the red, green, and blue channels of a certain pixel in the image (usually the value range is 0-255). This formula generates the corresponding grayscale value by averaging the values of the three color channels. Then the stretching in the height direction is L = (P / 255)H. By intersecting the grayscales of the plan views in the three directions and taking the intersecting space, the approximate spatial shapes of the objects inside the box can be obtained. Mark the areas with frequency characteristics that need attention through the ultrasonic detector, and then map them to the space formed by the X-rays to isolate the space containing these frequency characteristics that need attention. Identify the temperature of these areas through the temperature detector. If the temperature of these areas is also abnormal, set them as abnormal luggage and issue an alarm. Through step-by-step investigation and associated hybrid retrieval, high-precision scanning and analysis of the inside of the luggage are realized, and the difficulty of manual review is reduced.

[0009] In a further embodiment, it further includes step S4 of formulating a hybrid retrieval strategy. The hybrid retrieval strategy includes: establishing a feature database of sound wave amplitude and frequency; establishing a database of the relationship between the size and weight of the luggage, judging the danger level of the luggage by retrieving the multi-dimensional feature data of the luggage, and adjusting the retrieval strategy through training.

[0010] By adopting the above technical solution, the retrieval strategy is executed after the establishment of the grayscale three-dimensional model inside the luggage. X-ray scanning and modeling are used to provide basic data, and then the ultrasonic detector is used to retrieve and analyze this data to judge the danger. The temperature detector is used to retrieve the temperature of each area and analyze the danger level of each area. The X-ray detection, ultrasonic detection, temperature detector detection, and gravity detection do not affect each other. Finally, these detection methods divide each area inside the luggage into different danger levels and give the luggage an overall danger level. Through the weighted mixing calculation method, the danger level values of each area are superimposed and calculated. Finally, the luggage reaching the warning value is marked and an alarm is issued. Since different detection methods have different effects on different dangerous goods, it is necessary to control the weighting value according to the remaining items inside the luggage. In this way, misjudgment can be reduced while avoiding system loopholes.

[0011] In a further embodiment, a data feature record database is established to record all data features and update them in real time.

[0012] By adopting the above technical solution, for example, if the echo data frequency and amplitude of an item detected by the ultrasonic detector are not recorded in the database, but it is indeed a dangerous good after opening the box for inspection, then its features need to be updated and recorded. Another example is that if the grayscale value scanned by the X-ray is not within the range of abnormal warning, but it is actually a dangerous good, then it also needs to be found and recorded in the database. All features recorded in the database are the key features to be queried.

[0013] In a further embodiment, the abnormal data is associated and recorded through a hybrid retrieval strategy.

[0014] By adopting the above technical solution, for example, when it is found that a certain area with a grayscale of P inside the luggage is a dangerous good after detection, then the temperature, amplitude, and frequency of this area need to be associated. When it is found again next time, direct associated retrieval is performed. If all features are the same, an alarm is directly issued.

[0015] In a further embodiment, step S1 further includes performing YOLOV recognition on the scanned pictures to fit the boundaries of the items.

[0016] In a further embodiment, step S5 is further included, which is to establish a biological invasion detection database for performing biometric detection on the echoes of some specific bands inside the luggage.

[0017] In a further embodiment, a temperature detection database is set in the biological invasion detection database, and the temperature detection database is used to record the biological dormancy temperature of specific bands.

[0018] By adopting the above technical solution, due to the risk of biological invasion, the echo characteristics and temperature correlation are combined to detect the living organisms in the dormant state that may exist on the luggage, reducing the risk of biological invasion.

[0019] The present invention also provides a system for an abnormal luggage handling method of airport passengers based on a hybrid retrieval strategy, including: At least one X-ray scanning device, which is used to scan the luggage; Range-finding gratings are arranged at the inlet and outlet of the X-ray scanning device, which are used to measure the length, width and height values of the luggage, and are also used to measure the position of the luggage; An ultrasonic detector is arranged on one side of the X-ray scanning device; And a temperature detector is arranged on one side of the X-ray scanning device.

[0020] In summary, the present invention has the following beneficial effects: 1. Through the hybrid strategy of X-ray scanning, it is used to evaluate the risk level of the luggage. By substituting the data of the ultrasonic detector and the temperature detector into the hybrid retrieval strategy, the safety level value of the luggage can be obtained quickly, assisting the personnel to judge whether the luggage needs to be opened for inspection. At the same time, because the acquired data features are sufficient, the data obtained can be subjected to feature extraction through a log table, which is used to establish a luggage database of the luggage passing through the airport security system during the current period, providing basic data support for subsequent reference and compensation; through this process, the collection of relevant specification policy documents for abnormal luggage of different airlines and the preprocessing of specification document data are carried out. Among them, the collection of policy documents can regularly obtain unstructured documents such as the "Baggage Transportation Terms", "Abnormal Handling Manual", and "Baggage Compensation Standard" of each airline through channels such as the official websites of each airline, internal documents, and industry specification documents. Then, the collected specification document data needs to be parsed and processed. By deploying the localized open-source tool OmniParse, the unstructured document data such as the specification documents of each airline collected are parsed into structured and operable data, and a triple storage structure of airline - abnormal type - relevant policy is designed with the airline as the primary key, and the parsed document data is stored in a relational database (such as MySQL). Description of the Drawings

[0021] Figure 1 is a partial flow block diagram of the system of the present invention; Figure 2 is a flow block diagram of the data acquisition method of the security inspection system of the present invention; Figure 3 is a flow block diagram of the analysis of abnormal data detected by the ultrasonic detector of the present invention; Figure 4is a flowchart of the temperature detector of the present invention detecting abnormal characteristics; Figure 5 It is a flowchart for embodying the third embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0023] The same parts are denoted by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to the attached Figure 1 In the description of this specification, the words "bottom" and "top", "inside" and "outside" refer to directions toward or away from a particular component geometry, respectively. In addition, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this specification, "plurality" means two or more, unless the direction of the center is otherwise clearly and specifically defined.

[0024] Example 1: Figures 1 - 5 As shown, the present invention discloses a hybrid retrieval strategy for airport luggage. When a staff member initiates a query request, the following steps are required to formulate the strategy and process.

[0025] The present invention mainly includes two parts. The first part is to obtain the data of the luggage from the airport security inspection system, and perform feature extraction, recording and storage on the obtained data; The second part is to provide basic data support for the case of lost or damaged luggage at the airport through the data obtained by the security inspection system, which is used to ensure the integrity and objectivity of the compensation system, so that the compensation system has a system to follow; Because with the development of science and technology, more and more high-tech is bypassing the airport security system, so security checks need to be carried out in a variety of ways. Similarly, the data obtained is also diverse, so it is necessary to determine the complete status of a suitcase through a mixed retrieval method. For example, X-ray detection data can determine the approximate material type of objects inside the suitcase, including those materials and whether there is metal. Ultrasonic detection can identify characteristic waveforms to determine whether there are special items inside the suitcase, whether there are dangerous items and the approximate type of items. Because the echo of ultrasonic detection is affected by many factors, it is not very ideal for detecting the complex spatial structure inside the suitcase. Therefore, ultrasonic detection is mainly used in conjunction with X-ray detection to supplement multiple X-ray detection data. Temperature detection is more about whether there are living things inside the suitcase to avoid biological invasion.

[0026] likeFigures 1 - 5 The steps are as follows: Step S1: Build a hybrid retrieval knowledge base and obtain the scanned pictures of the luggage through the X-ray detection device. The hybrid retrieval knowledge base includes building a relational knowledge base based on keyword retrieval and building a vector knowledge base based on vector retrieval. To provide more accurate processing suggestions for the airport staff handling abnormal luggage, a relational knowledge base for the processing procedures and compensation standards of different airlines for different types of abnormal luggage needs to be established. This process includes the collection of the specification policy documents related to abnormal luggage of different airlines and the preprocessing of the specification document data. Among them, the collection of policy documents can regularly obtain unstructured documents such as the "Baggage Transportation Terms", "Abnormal Handling Manual", and "Baggage Compensation Standard" of each airline through channels such as the official websites of each airline, internal documents, and industry specification documents. Then, the collected specification document data needs to be parsed and processed. By deploying the localized open-source tool OmniParse, the unstructured document data such as the specification documents of each airline is parsed into structured and operable data, and a storage structure of airline-abnormal type-related policy triples is designed with the airline as the primary key, and the parsed document data is stored in a relational database (such as MySQL); Step S2: Establish a relational knowledge base, perform a hybrid strategy retrieval analysis on the pictures to determine whether there are dangerous goods. If so, an alarm is issued; if not, no alarm is issued. To provide more accurate processing suggestions for the airport staff handling abnormal luggage, a vector knowledge base needs to be established for the historical processing case data of different airlines for different types of abnormal luggage, to help the staff give reference suggestions based on the historical processing case with the highest similarity match. This process includes the collection of historical abnormal luggage case data and the construction of the vector knowledge base. First, by building a local passenger abnormal luggage processing system at the airport terminal, the passenger abnormal luggage cases at this airport are collected; at the same time, the system of the present invention also provides an external API interface, which can realize the docking of the passenger abnormal luggage case data of other airports, and the historical case data is stored in the local relational database of the airport. Then, the historical case data of the passenger abnormal luggage in the local relational database is regularly extracted for preprocessing and vectorization. Identify the field types in the database table, delete the missing values, unify the date type field format, standardize the numerical type fields, and use the pre-trained model BERT (Bidirectional Encoder Representations from Transformers) to generate sentence vectors for the text type field data; then splice the processed numerical features of each field into a single vector; finally, use the multi-modal BERT to generate a joint vector and store it in the ElasticSearch index library; Step S3: Establish a vector knowledge base. Set up ultrasonic detectors and temperature detectors according to the region, and conduct a mixed inspection of the luggage based on the data from the ultrasonic detectors and temperature detectors. To provide more accurate processing suggestions for the airport staff handling abnormal luggage, it is necessary to establish a vector knowledge base for the historical processing case data of different airlines and different types of abnormal luggage, to help the staff give reference suggestions based on the historical processing case with the highest similarity match. This process includes the collection of historical abnormal luggage case data and the construction of the vector knowledge base; when a query request is received, according to the case description of the new passenger's abnormal luggage input by the airport staff, using the flexibility of keyword retrieval, first, accurately find the processing process and compensation standards of the described airline and abnormal type of luggage in the relational database through keyword matching. At the same time, using the efficiency of vector retrieval, the characteristics of the passenger's abnormal luggage described by the staff are characterized as a query vector (query_vector) through the Embedding model, and the similarity score score between the query vector and each vector in each shard in the ElasticSearch index library is calculated based on the cosine similarity algorithm in the Knn algorithm. The qiore calculation formula is (1 + cosine(query_vector, index_vector)) / 2. To further improve the accuracy of retrieval, recall the data of the Top10 similarity scores calculated for each shard, and sort all the recalled data in reverse order according to the scores to obtain a similarity result set. Finally, select the Top5 similar data with the highest scores in the result set. Therefore, this method requires a powerful and suitable large language model, and develop a retrieval-enhanced generation function module based on RAG technology and integrate it with the large model. Use RAG technology to retrieve from the established abnormal luggage business domain knowledge base, generate prompt words (prompt) based on the retrieval results as the input of the large model, and combine the understanding and reasoning capabilities of the large model. Finally, generate decision-making suggestions for the airport staff handling abnormal luggage through the visualization function page; Step S4, formulate a hybrid retrieval strategy, which includes: establishing a feature database for sound wave amplitude and frequency; establishing a database for the relationship between the size and weight of the luggage, judging the danger level of the luggage by retrieving multi-dimensional feature data of the luggage, and adjusting the retrieval strategy through training; the retrieval strategy is executed after the establishment of the grayscale three-dimensional model inside the luggage, X-ray scanning and modeling are used to provide basic data, and then the ultrasonic detector is used to retrieve and analyze this data to judge the danger, the temperature detector is used to retrieve the temperature of each area and analyze the danger level of each area, X-ray detection, ultrasonic detection, temperature detector detection and gravity detection do not affect each other. Finally, these detection methods divide each area inside the luggage into different danger levels and give the luggage an overall danger level. Through the weighted mixing calculation method, the danger level values of each area are superimposed and calculated. Finally, the luggage that reaches the warning value is marked and an alarm is issued. Because different detection methods have different effects on different dangerous goods, it is necessary to control the weighted value according to the remaining items inside the luggage. In this way, misjudgment can be reduced while avoiding system loopholes. To provide more accurate processing suggestions for the airport abnormal luggage handling staff, it is necessary to establish a vector knowledge base for the historical processing case data of different airlines and different abnormal luggage types to help the staff give reference suggestions according to the historical processing case with the highest similarity matching. This process includes the collection of historical abnormal luggage case data and the construction of the vector knowledge base.

[0027] First, by building a local passenger abnormal luggage handling system at the airport terminal, collect the passenger abnormal luggage cases at this airport; at the same time, the system of the present invention also provides an external API interface, which can realize the docking of passenger abnormal luggage case data of other airports, and the historical case data is stored in the relational database of the airport local.

[0028] Then, regularly extract the historical case data of passenger abnormal luggage in the local relational database for preprocessing and vectorization. Identify the field types in the database table, delete the missing values, unify the date type field format, standardize the numerical type fields, and use the pre-trained model BERT (Bidirectional Encoder Representations from Transformers) to generate sentence vectors for the field data of the text type; then splice the processed numerical features of each field into a single vector; finally, use the multi-modal BERT to generate a joint vector and store it in the ElasticSearch index library; Step S5, establish a biological invasion detection database for biometric detection of echoes in some specific frequency bands inside the luggage. Traditional X-ray scanning of luggage can obtain a floor plan of the inside of the luggage. From the floor plan, it is possible to roughly see whether there are dangerous goods inside the luggage based on different brightness levels. However, with the development of technology, there are more and more ways to avoid X-ray scanning. Therefore, this application proposes a hybrid strategy based on X-ray scanning for risk level assessment of luggage. By substituting the data from ultrasonic detectors and temperature detectors into the hybrid retrieval strategy, the safety level value of the luggage can be quickly obtained to assist personnel in determining whether the luggage needs to be opened for inspection. It further includes step S6, establish a luggage claim mechanism and historical data record management rules. The luggage claim mechanism is used to estimate the items and value inside the luggage based on historical data, and the historical data record management rules are used to regularly clean and update the repository.

[0029] In a further embodiment, step S3 further includes: Step S31, obtain the frequency band of the internal object of the luggage through an ultrasonic detector. Step S32, obtain the internal thermal radiation distribution map of the luggage through a temperature detector. Step S33, a weighing module is provided on the conveyor belt of the X-ray detection device, and ranging gratings are provided at both the inlet and outlet ends of the X-ray. The ranging gratings are used to determine whether the luggage leaves and enters the conveyor belt, and the weighing of the luggage is achieved through the mutual cooperation of the weighing module and the ranging gratings.

[0030] By adopting the above technical solution, the X-ray module obtains the scanning imaging plan views directly above the luggage, on the left side, and on the right side of the luggage through multi-dimensional scanning. X-rays have strong penetration ability and can penetrate low-density substances such as clothes and plastics, but will be absorbed by high-density substances such as metals and liquids. Different substances have different absorption degrees of X-rays, forming light and dark differences. The absorption of X-rays by substances is related to the atomic number and density. High-density substances (such as metals) absorb more X-rays and appear dark; low-density substances (such as clothes and plastics) absorb less and appear light. Therefore, the top view, left view, and right view of the luggage can be obtained, and the gray levels of the three views in different directions are also different. Through 3D modeling and comparison, the three-dimensional shape of the items is restored. The restoration steps are as follows: through the gray level of the plan view, the view is extended in the acquisition direction. For example, the scanning image above the luggage is extended downward according to the gray level. Set the gray level of a certain pixel point in the plan view as P. The length X, width Y, and height H of the luggage are obtained through the ranging grating. The gray level is linearly mapped by Gray = (R + G + B) / 3, where R, G, and B are the values of the red, green, and blue channels of a certain pixel in the image (usually the value range is 0-255). This formula generates the corresponding gray level value by averaging the values of the three color channels. Then the stretching in the height direction is L = (P / 255)H. By intersecting the gray levels of the plan views in the three directions and taking the intersecting space, the approximate space shapes of the objects inside the box can be obtained. The area with the frequency characteristics that need attention is marked by the ultrasonic detector, and then it is mutually mapped with the space formed by the X-rays to isolate the space containing these frequency characteristics that need attention. The temperature of these areas is identified by the temperature detector. If the temperature of these areas is also abnormal, it is set as abnormal luggage and an alarm is issued. Through step-by-step investigation and associated hybrid retrieval, high-precision scanning and analysis of the inside of the luggage are realized, and the difficulty of manual inspection is reduced.

[0031] Embodiment 2: As Figures 1 - 3 shown, an abnormal luggage processing method and system for airport passengers based on a hybrid retrieval strategy includes: an X-ray detection module: composed of an X-ray emitter, a receiver, and a multi-angle imaging unit, installed above and on both sides of the luggage conveyor belt, and used to obtain the top view, left view, and right view scanning plan views of the luggage; A hybrid sensor module: includes an ultrasonic detector, and the ultrasonic detector needs to cover a frequency range of 20 kHz - 100 MHz; an infrared temperature detector, and the detection accuracy is preferably as high as possible; a weighing module and a ranging grating; A data processing module: built-in with a 3D modeling engine and a hybrid retrieval strategy, and the hybrid retrieval strategy is optimized based on the ResNet-50 architecture and a hybrid retrieval strategy database; Alarm and Interaction Module: It includes a visualization interface for danger levels and a hierarchical alarm device.

[0032] As Figures 1 - 4 shown, when the luggage enters the conveyor belt, the ranging grating triggers the X-ray module to synchronously obtain gray-scale images of three channels: top view, left view, and right view. The ultrasonic detector emits pulse signals (pulse width 10 μs, repetition frequency 1 kHz) and receives the time-domain and frequency-domain characteristics of the echo reflected inside the luggage. The infrared temperature detector obtains the thermal radiation distribution map of the luggage surface (thermal sensitivity 0.05 °C) at a sampling rate of 5 Hz. The weighing module combines the grating trigger signal and records the mass data when the luggage completely enters the weighing area. Based on the gray-scale values (Gray = 0.299R + 0.587G + 0.114B) of the top view, left view, and right view plane graphs, a three-dimensional model is generated using the voxel reconstruction algorithm: Define the voxel size as 5 mm × 5 mm × 5 mm; assign weight values to each voxel: where αi is the viewing angle weight coefficient (0.5 for top view, 0.25 for each side view), Gref is the reference gray-scale threshold for dangerous goods, map the ultrasonic spectrum characteristics (extract MFCC coefficients) and the temperature distribution map (spatial gradient analysis) to the three-dimensional model to generate a multi-modal fusion feature matrix; Execute a hybrid retrieval strategy. At the first level, perform binary classification based on the model (input size 512 × 512 × 4 channels, including gray-scale + temperature map of three views); output the preliminary screening dangerous probability P. At the second level, perform dynamic time warping (DTW) matching on the ultrasonic spectrum and calculate the similarity Sus with the dangerous goods feature library. At the third level, execute the mass-volume correlation verification ; where ρv is the estimated density of voxel v (obtained by looking up the table according to the gray scale). If Δ > threshold (such as 200 g), then trigger an abnormal mark; comprehensive scoring: when Risk > 0.85, trigger a red alarm; dynamic strategy optimization: actively learn misreported / missed reported samples every week and update the feature extractor through the gradient reversal layer (GRL); The ultrasonic feature library adopts an incremental k-d tree structure to support online insertion of new dangerous goods spectrum characteristics; combine the temperature sensor data to establish a bioheat model, ; where represents the metabolic intensity, is the heat dissipation coefficient, and compare with the database; an attention mechanism can be used to weight specific areas. If a suspicious area is found, perform local CT scanning on the suspicious area and obtain a high-resolution cross-sectional image through the FDK reconstruction algorithm; Introduce the learning module into the system, define the state space as , the adjustment of the hybrid weight vector is the action space, and design the reward function , where TP / FP / FN are the detection statistics of the day, and the policy network parameters are updated online through the Q-learning algorithm.

[0033] Embodiment 3: As Figure 5 shown, to solve the above problems, the object of the present invention is to provide a method and system for processing abnormal luggage of passengers based on a hybrid retrieval strategy, including the following steps: (1) Development of the RAG function module First, select the large language model DeepSeek and develop the RAG function module applied to the knowledge base in the field of abnormal luggage business. This function module can accurately retrieve the abnormal luggage handling rules, handling processes, and compensation terms related to luggage damage of a certain airline from the relational knowledge base according to the information related to the abnormal luggage of passengers of a certain airline input by airport staff (such as airline code, luggage abnormal type - luggage damage). At the same time, it can convert the staff's input into a query vector and perform similarity retrieval through vector retrieval in the vector library to retrieve supplementary evidence semantically related to the input, such as knowledge fragments such as the degree of similar damage, past handling cases under the flight type, and compensation amounts in the historical case data of the airline.

[0034] (2) Development of the visual interaction interface Then, develop a visual interaction page integrating the DeepSeek model, input the knowledge fragments retrieved by the RAG module as prompt words into the large model DeepSeek, utilize the understanding and reasoning ability of the large model to generate detailed processing process suggestions, structured processing steps, liability determination logic, and compensation basis for airport staff, and present the generated results to airport staff in a visual interface, including processing steps, measure descriptions, and compensation plans.

[0035] Embodiment 4: Scenario description: A passenger took flight CA1510 of Air China from Shanghai to Beijing. After the passenger arrived at the destination airport, Beijing Capital International Airport, it was found that the checked luggage was lost, and the passenger went to the luggage inquiry counter at the arrival airport for processing. After the staff at the airport luggage inquiry counter fully understood the abnormal case reported by the passenger, they input the case information through the system visual interface: "Passenger of Air China domestic flight CA1510, flying from Shanghai to Beijing, luggage lost, containing emergency supplies and valuable electronic products, applying for compensation".

[0036] First, the system automatically extracts keywords "Airline: Air China", "Abnormal type: Luggage lost", "Domestic flight", "Compensation", and accurately matches the handling process specifications and compensation policies for lost luggage on domestic flights of Air China in the MySQL relational knowledge base through keyword matching: (1)Immediate Declaration: In case of lost luggage, a declaration must be made to the Air China Luggage Service Counter within 24 hours, and valid ticket, luggage tag, passport / identity document, list of items in the luggage, purchase vouchers or photos should be provided.

[0037] (2)Luggage Search: The staff will enter the case information into the Abnormal Luggage Handling System; for domestic flights, the longest search period is 7 days. If the luggage is not found, the compensation process will be initiated.

[0038] (3)Compensation Application: The compensation standard for lost luggage is 100 yuan per kilogram, with a maximum of 3000 yuan; the advance payment limit for emergency supplies is 500 yuan for purchasing daily necessities, and the passenger needs to sign a temporary agreement.

[0039] At the same time, the system will vectorize the staff's case description text through the BERT model to generate a query vector, and retrieve the top 5 similar historical Air China luggage loss cases in the ElasticSearch vector library as follows: Finally, the system will automatically fuse the results retrieved from the relational knowledge base and the vector knowledge base, and input them into the DeepSeek model. After the model's reasoning, the following suggestions will be generated for the staff in the form of a visualization page: (1)Liability Determination: According to Air China's terms, the liability for lost luggage belongs to the airline; (2)Emergency Handling: The system registers the case information of the passenger's lost luggage and guides the passenger to sign the "Temporary Advance Payment Agreement", and advances 500 yuan for emergency supplies; Compensation Calculation: Calculated based on the luggage weight of 28 kilograms, the preliminary compensation amount is 2800 yuan; if the luggage is not retrieved within 7 days, the full compensation process will be initiated.

[0040] In the embodiments disclosed in the present invention, terms such as "installation", "connection", "linkage", "fixation" should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; "linkage" can be a direct connection or an indirect connection through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments disclosed in the present invention can be understood according to specific circumstances.

[0041] This specific embodiment is only an interpretation of the present invention and is not a limitation of the present invention. Those skilled in the art can make modifications without creative contributions to this embodiment according to their needs after reading this specification, but as long as they are within the scope of the claims of the present invention, they are protected by the patent law.

Claims

1. An abnormal luggage handling method for airport passengers based on a hybrid retrieval strategy, characterized in that It includes the following steps: Step S1, construct a hybrid retrieval knowledge base for obtaining historical data, where the historical data includes the claim processes and compensation standards for historical baggage loss, damage, and delay at airports, and simultaneously establish a baggage compensation standard model, which is used to determine the claim process and compensation standard according to the compensation specification documents of different airlines; Step S2, establish a relational knowledge base, obtain the basic data of the luggage through internal scanning of the luggage and save it in the form of pictures, perform data marking and storage for each piece of luggage, and at the same time extract feature data, perform hybrid strategy retrieval analysis on the pictures to judge whether there are dangerous goods. If so, an alarm is issued; if not, no alarm is issued; Step S3, establish a vector knowledge base, set ultrasonic detectors and temperature detectors according to regions, perform hybrid investigation and obtain hybrid features of the luggage through the data of the ultrasonic detectors and temperature detectors to improve the basic data information of the compensation system, and preset the compensation standard for each piece of luggage according to the compensation specification documents of different airlines; Step S4, formulate a hybrid retrieval strategy, and the hybrid retrieval strategy includes: establish a feature database of sound wave amplitude and frequency; Establish a database of the relationship between luggage size and weight, judge the danger level of the luggage by retrieving the multi-dimensional feature data of the luggage, and adjust the retrieval strategy through training.

2. The method for handling abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 1, wherein The step S3 further includes obtaining luggage data, and step S3 further includes: Step S31, obtain the frequency band of the internal objects of the luggage through an ultrasonic detector; Step S32, obtain the internal thermal radiation distribution map of the luggage through a temperature detector; Step S33, a weighing module is provided on the conveyor belt of the X-ray detection device, and ranging gratings are provided at both the inlet and outlet ends of the X-ray. The ranging gratings are used to judge whether the luggage leaves and enters the conveyor belt, and the weighing of the luggage is realized through the mutual cooperation of the weighing module and the ranging gratings.

3. The method for processing abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 2, wherein, The step S2 includes: perform associated tracking according to the size, color, and quality of the luggage. When loss information is reported, the staff will enter the case information into the abnormal luggage processing system; for domestic flights, the longest search time is 7 days. If it is not found, the compensation process will be started. The abnormal state of airport luggage mainly reflects the luggage with dangerous goods during safety inspection and the damaged and lost luggage during claims.

4. The method for handling abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 2, characterized in that, It further includes step S6, establish a luggage claim mechanism and a historical data record management rule. The luggage claim mechanism is used to estimate the items and value inside the luggage according to historical data, and the historical data record management rule is used to regularly clean and update the storage repository.

5. The method for processing abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 1, wherein: Establish a data feature record database, record all data features and update them in real time. At the same time, it is necessary to associate the historical data with the latest recorded data, establish a claim strategy model, and optimize the claim strategy according to historical claim information.

6. The method for processing abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 1, characterized in that: Associate and record abnormal data through a hybrid retrieval strategy.

7. The method for handling abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 1, wherein: The step S1 further includes performing YOLOV recognition on the scanned pictures to fit the boundaries of the items.

8. The method for processing abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 1, characterized in that: It further includes step S5 of establishing a biological invasion detection database for performing biometric detection on the echoes of some specific bands inside the luggage case.

9. The method for processing abnormal luggage of airport passengers based on a hybrid retrieval strategy according to claim 8, characterized in that: A temperature detection database is set in the biological invasion detection database, and the temperature detection database is used to record the biological dormancy temperature of specific bands.

10. A system for the method of handling abnormal luggage of airport passengers based on a hybrid retrieval strategy according to any one of claims 1-9, characterized in that, It includes: At least one X-ray scanning device for scanning the luggage case; Range-finding gratings are arranged at the inlet and outlet of the X-ray scanning device for measuring the length, width and height values of the luggage case and also for measuring the position of the luggage case; An ultrasonic detector is arranged on one side of the X-ray scanning device; And a temperature detector is arranged on one side of the X-ray scanning device.

Images