A method and system for processing civil aviation navigation data information
By performing entry verification and occlusion detection on the data packets of aircraft images transmitted from fixed cameras, and comparing them with those of the local and adjacent cameras, the problem of unreliable aircraft identification results on the tarmac at night in winter was solved, and accurate binding of the image recording chain was achieved.
Patent Information
- Application Number
- CN202611580517.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-13
- Publication Date
- 2026-11-10
AI Technical Summary
On the tarmac late at night in winter, the unreliable identification results of the fuselage number lead to the misalignment between the image recording chain and the actual flight. This is mainly caused by factors such as continuous obstruction, adjacent aircraft of the same model having the same shape, liquid film trailing, and disordered transmission.
By acquiring the data packets of the aircraft's body image transmitted back from the fixed camera at the aircraft position, the entry verification of the shooting time, aircraft position number and acquisition unit identifier is performed. Obstructed areas are detected, and unobstructed segments are stitched together. By comparing the candidate numbers of the aircraft and adjacent aircraft positions, the binding status of the aircraft number and flight number is determined to ensure the accuracy of the binding.
This improved the accuracy of binding aircraft serial numbers to flight numbers, reduced the probability of incorrect associations in the image recording chain, and ensured the reliability of the image recording chain.
Smart Images

Figure CN123072235A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of civil aviation image processing technology, specifically to a method and system for processing civil aviation navigation data information. Background Technology
[0002] During civil aviation flight operations, airports need to capture images of the aircraft's fuselage using fixed cameras after the aircraft parks at its position, identify the aircraft's serial number, and then bind the identified serial number to the flight number in the flight data. This forms an image recording chain that links the parking position view, de-icing view, and walk-around inspection view together. The binding process is the entry point for the entire image recording chain. If an incorrect binding occurs, all subsequent de-icing and inspection views will be incorrectly associated. Therefore, the correctness of the binding directly determines the usability of the image recordings.
[0003] On a winter night at an airport tarmac, departing flights are queuing up. Two identical aircraft from the same airline are parked at adjacent gates. Shuttle buses continuously traverse the passageways beside the gates, obscuring the numbered areas in each frame of the aircraft's fuselage. The system consistently fails to capture images with complete numbered frames. Conventional identification methods remove the obscured frames and fill in the gaps with preceding and following frames. However, this approach fails to address the lack of a precondition for completion when obscuration is continuous, resulting in inconsistent identification results across multiple flights. Furthermore, conventional methods directly bind the identification results without verification, neglecting to check the reliability of the image source or the correctness of the transmission order. Simultaneous transmission from multiple gates at night leads to network congestion, disrupting the image arrival order and causing misattribution when images are entered into the database based on arrival order.
[0004] After the aircraft is sprayed with de-icing fluid on the de-icing pad, the liquid film flows downwards along the fuselage. The edges of the serial number strokes are dragged by the liquid, creating a trailing effect. Conventional enhanced identification processes treat this liquid film as normal reflection, resulting in inconsistent identification results. Furthermore, images captured by handheld cameras during aircraft inspections have higher resolution than images taken from a distance at the aircraft position. However, the standard procedure stores these handheld images only as attachments to the inspection results, excluding them from serial number verification. These deficiencies perpetuate the risk of misalignment between the image recording chain and the actual flight information. Summary of the Invention
[0005] In view of the above-mentioned problems, the present invention provides a method and system for processing civil aviation navigation data information.
[0006] Therefore, the technical problem solved by this invention is to address the issue of unreliable fuselage number identification results leading to misalignment between the image recording chain and the actual flight under conditions of continuous tarmac obstruction late at night in winter, identical shapes of adjacent aircraft of the same model, and disordered transmission of liquid film trailing.
[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a civil aviation flight data information processing method, comprising: acquiring a fuselage image data packet transmitted by a fixed camera at an aircraft stand, wherein the fuselage image data packet contains a shooting time, an aircraft stand number, and an acquisition unit identifier; performing an entry verification on the fuselage image data packet based on the shooting time, the aircraft stand number, the acquisition unit identifier, and flight data to obtain an entry verification result; performing occlusion area detection on the fuselage images that pass the entry verification, and writing the unoccluded fragments of numbered regions in each frame and the shooting time into a fragment sequence arranged in chronological order of shooting time; splicing the fragment sequence to obtain a candidate number; comparing the candidate number with a first comparison benchmark and adjacent aircraft stand candidate numbers, and determining the binding status identifier between the fuselage number and the flight number in the flight data based on the comparison result.
[0008] In a preferred embodiment of the civil aviation navigation data information processing method of the present invention, the following steps are performed: Based on the shooting time, the aircraft stand number, the acquisition unit identifier, and the navigation data, an entry verification is performed on the fuselage image data packet to obtain an entry verification result. This includes: comparing the shooting time with a preset time range corresponding to the expected docking time in the navigation data; comparing the aircraft stand number with the aircraft stand number in the navigation data; and comparing the acquisition unit identifier with a preset aircraft stand acquisition unit registration form. When all comparisons match, a confirmation message is returned to the fixed camera at the aircraft stand to obtain an entry verification result that allows recognition. When any comparison does not match, the fuselage image data packet is placed in a verification queue, and a first prompt message containing the aircraft stand number and shooting time is sent to the apron duty terminal to obtain an entry verification result that suspends recognition.
[0009] As a preferred embodiment of the civil aviation navigation data information processing method described in this invention, the following steps are included: performing occlusion area detection on the fuselage image that has passed the entry verification, and writing the unoccluded fragments of numbered regions in each frame of the image along with the shooting time into a fragment sequence arranged chronologically by shooting time. This includes: detecting the position marker of the occluded part of the numbered region in each frame of the fuselage image, writing the unoccluded fragments along with the shooting time into the fragment sequence; counting the cumulative visibility count of each part in the fragment sequence; and obtaining the result when the cumulative visibility count of each part reaches a preset minimum count. A fragmented sequence that can be used for splicing; when there are parts that have not reached the preset minimum number of times, an instruction to increase the frame rate is sent to the fixed camera, and a supplementary sampling period is arranged outside the time period when the shuttle bus passes through the channel. After the supplementary sampling, the fragments obtained by the supplementary sampling are written into the fragmented sequence and the cumulative number of visible times is recounted until the cumulative number of visible times of each part reaches the preset minimum number of times, thus obtaining the fragmented sequence that can be used for splicing; when there are still parts that have not reached the preset minimum number of times after the supplementary sampling, an insufficient sampling flag is generated, and the written fragments are retained in the fragmented sequence.
[0010] As a preferred embodiment of the civil aviation navigation data information processing method of the present invention, the following steps are included: comparing the candidate number with a first comparison benchmark and the candidate numbers of adjacent aircraft stands, and determining the binding status identifier of the fuselage number and the flight number in the navigation data based on the comparison results. This includes: extracting the complete fuselage image of the aircraft taken when the corresponding flight taxis into the aircraft stand as the first comparison benchmark; comparing the candidate number with the visible number parts in the first comparison benchmark segment by segment; sending a query request to the adjacent aircraft stand acquisition unit and receiving the candidate numbers of the adjacent aircraft stands; when the candidate number matches the first comparison benchmark and is different from the candidate numbers of the adjacent aircraft stands, determining that the candidate number is valid and writing a binding status identifier indicating successful binding; when the candidate number is the same as or has a similarity higher than a preset similarity condition with the candidate numbers of the adjacent aircraft stands, backtracking to query the earliest captured fragment in the fragment sequence for a secondary judgment; when the secondary judgment fails, determining that the splicing has failed and transferring to the supplementary acquisition of the fragment sequence, and writing a binding status identifier indicating that binding is suspended.
[0011] As a preferred embodiment of the civil aviation navigation data information processing method described in this invention, the following steps are included: obtaining candidate numbers by splicing the fragment sequence, comprising: detecting the liquid film coverage state frame by frame in the numbered area of the de-icing image to obtain liquid film status identifiers of liquid film flow state, liquid film still state, or liquid film receding state; selecting frames for identification from each frame in order of priority for liquid film receding state and second priority for liquid film still state, wherein the frames of liquid film flow state are only used to provide position information in the fragment sequence; when the characters identified in the liquid film receding state frame and the liquid film still state frame at the same numbered location are different, backtracking to query the dry image before the start of de-icing, and using the identification result of the dry image for splicing; when no dry image is available before de-icing, marking the characters at the corresponding location as undetermined characters; and splicing the characters identified at each location to obtain the candidate number.
[0012] As a preferred embodiment of the civil aviation navigation data information processing method described in this invention, the method for acquiring the fuselage image data packet transmitted by the fixed camera at the aircraft position includes: after receiving the fuselage image data packet, extracting the shooting time and acquisition unit identifier from the fuselage image data packet, and reconstructing the shooting sequence of the images from each aircraft position according to the shooting time; determining whether the shooting time interval of adjacent data packets in the same acquisition unit exceeds a preset multiple range corresponding to the fixed frame rate interval; if it exceeds the limit, checking the acquisition unit identifiers of all data packets that have arrived within the corresponding time period, and repositioning the images belonging to this aircraft position; when images with the same identifier belonging to different aircraft positions appear at the same shooting time, sending a first self-check request to the corresponding acquisition unit respectively, taking the image with the earlier shooting time in the response as the standard, and delivering the reconstructed image as the fuselage image data packet to be verified to the entry verification.
[0013] As a preferred embodiment of the civil aviation navigation data information processing method of the present invention, the method includes: when no dry footage is available before de-icing, marking the characters in the corresponding parts as pending characters; splicing the characters identified from each part to obtain the candidate number, including: when the pending characters exist, sending a first reshoot request specifying the location of the fuselage number and the suggested shooting angle to the handheld shooting device of the on-duty aircraft inspection personnel; receiving the reshoot footage and shooting location identifier returned by the handheld shooting device; comparing the number identification result in the reshoot footage with the splicing result of the fragment sequence character by character; when all matches, removing the pending characters to obtain a determined candidate number; when partially matching, updating the fragment sequence based on the reshoot footage and re-sponsoring to obtain an updated candidate number; when completely mismatched, reviewing the verification record of the liquid film status identifier and the shooting sequence and rejecting the candidate number.
[0014] As a preferred embodiment of the civil aviation navigation data information processing method of the present invention, the method further includes: comparing the candidate number with a first comparison benchmark and the candidate numbers of adjacent aircraft positions, and determining the binding status identifier of the fuselage number and the flight number in the navigation data based on the comparison result; and responding to the insufficient data collection identifier, the conflict of the same identifier image, or the difference in the re-capture comparison exceeding a preset difference limit, respectively determining the abnormal types of insufficient data collection, questionable attribution, and contradictory identification, and processing them in the order of questionable attribution, contradictory identification, and insufficient data collection; when the attribution is questionable, freezing the corresponding... Once all camera positions have been entered into the database, a manual assignment request is sent to the apron duty terminal. Based on the response, the camera is re-bound and a binding status flag indicating successful binding or invalid binding is written. In cases of conflicting identification, a second reshoot request containing the difference in character positions is sent to the handheld shooting device, and a binding status flag is written based on the reshoot comparison results. In cases of insufficient data collection, a channel control inquiry is sent to the apron dispatch terminal. Upon receiving a positive response, data is collected centrally in the clear window and a binding status flag indicating successful binding is written. Upon receiving a negative response, the process switches to manual binding and a binding status flag indicating pending binding is written.
[0015] As a preferred embodiment of the civil aviation navigation data information processing method described in this invention, the method further includes: establishing an image recording chain for the corresponding flight based on the successful binding status identifier; sequentially attaching the aircraft stand image, de-icing image, and aircraft walk-around inspection image, with the binding status identifier written at the head of the chain; checking the binding status identifier at the head of the chain before each de-icing image or aircraft walk-around inspection image is entered into the database; automatically attaching the image when the binding status identifier at the head of the chain indicates successful binding; entering the attachment queue when the binding is suspended; refusing attachment and sending a first stop acquisition command to the acquisition side when the binding is invalid; and on flights with incorrect binding, checking the liquid film status identifier, acquisition unit identifier, and shooting time of each frame along the image recording chain to locate the error level and generate a correction record that retains the original incorrect binding content and the corrected content.
[0016] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a civil aviation navigation data information processing system, comprising a data packet acquisition module, an entry verification module, a fragment sequence generation module, and a number binding module; the data packet acquisition module is used to acquire fuselage image data packets transmitted back by fixed camera at the aircraft position, wherein the fuselage image data packet contains the shooting time, aircraft position number, and acquisition unit identifier; the entry verification module is used to perform entry verification on the fuselage image data packet according to the shooting time, aircraft position number, acquisition unit identifier, and navigation data, and obtain the entry verification result; the fragment sequence generation module is used to perform occlusion area detection on the fuselage image that has passed the entry verification, and write the fragments of the numbered areas in each frame that are not occluded and the shooting time into a fragment sequence arranged in chronological order of shooting time; the number binding module is used to splice the fragment sequence to obtain candidate numbers, compare the candidate numbers with a first comparison benchmark and adjacent aircraft position candidate numbers, and determine the binding status identifier of the fuselage number and the flight number in the navigation data according to the comparison result.
[0017] The beneficial effects of this invention are as follows: In existing technologies, the binding process lacks verification, the image source and transmission order are not checked, obscured frames are completely removed, liquid film ghosting causes recognition jitter, and the image recording chain is misaligned with the actual flight. Before recognition, the invention uses the three elements of the flight data—shooting time, aircraft position number, and acquisition unit identifier—from the fuselage image data packet for entry verification, preventing unreliable images from being detected. The fragment sequence organizes unobscured fragments of each frame according to their shooting time, ensuring that the object being identified still has a data basis even when obscuration is continuous. Candidate numbers are written to the binding status identifier after double comparison between the complete fuselage image of the device and candidate numbers of adjacent aircraft positions, eliminating the possibility of piecing together adjacent aircraft numbers. The binding status identifier controls subsequent image attachment, preventing entry errors within the image recording chain. This application improves the accuracy of binding fuselage numbers and flight numbers and reduces the probability of incorrect associations in the image recording chain. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart of a civil aviation navigation data information processing method provided in one embodiment of the present invention.
[0020] Figure 2 This is a schematic diagram illustrating the execution logic of the fuselage screen data packet entry verification in a civil aviation navigation data information processing method according to an embodiment of the present invention.
[0021] Figure 3 This is a schematic diagram of the dual verification logic of candidate number, comparison benchmark and adjacent gate number in a civil aviation navigation data information processing method provided in an embodiment of the present invention.
[0022] Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided by the present invention. Detailed Implementation
[0023] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0024] Example 1, referring to Figures 1-3 This is one embodiment of the present invention, which provides a method for processing civil aviation navigation data information, including: The application scenarios of this application include data receiving servers, fixed cameras at aircraft positions, data acquisition units, flight data interfaces, apron duty terminals, apron dispatch terminals, handheld shooting devices, and local databases.
[0025] Fixed cameras are installed on the top poles of each aircraft stand on the airport apron. The fixed cameras are connected to the acquisition unit next to the aircraft stand via cables. The acquisition unit transmits the aircraft image data packets generated by the fixed cameras back to the data receiving server via the airport intranet.
[0026] The data receiving server communicates with the aeronautical data interface to read the flight number, gate number, and estimated docking time. The data receiving server also communicates wirelessly with the apron duty terminal and apron dispatch terminal to send notification messages, request manual assignments, and inquire about access control. Handheld cameras, carried by personnel conducting walk-around inspections, communicate with the data receiving server via the airport's wireless network to transmit re-enhanced footage and camera location markers.
[0027] The local database is deployed within the data receiving server and stores fragment sequences, candidate numbers, binding status identifiers, and image recording chains.
[0028] After a departing flight docks at its stand, the fixed camera at the stand continuously captures images of the aircraft. The acquisition unit writes the capture time, stand number, and acquisition unit identifier into each frame before transmitting it back. The data receiving server performs entry verification, occlusion detection, fragment splicing, number comparison, and writes binding status identifiers on the images. Finally, an image recording chain is established for attaching de-icing images and aircraft inspection images.
[0029] Step S1: Obtain the camera image data packet transmitted back by the fixed camera position. Write the shooting time, camera position number and acquisition unit identifier into the camera image data packet.
[0030] It should be noted that during winter nights, departing flights queue continuously on the tarmac, with multiple fixed cameras at various positions simultaneously transmitting images back to the data receiving server. If the transmitted images lack source information, the subsequent identification process cannot distinguish which camera position and which acquisition unit the image belongs to. Therefore, this step writes the shooting time, camera position number, and acquisition unit identifier into the data packet to provide a basis for comparison during entry verification. The design aims to ensure that each image data carries these three verifiable elements from the moment it leaves the camera. The operations following this step include reconstructing the shooting sequence of images from each camera position according to the shooting time, determining whether the shooting time interval between adjacent data packets is abnormal, reassigning images from abnormal periods, and handling images with the same identifier belonging to different camera positions at the same shooting time. The specific execution methods of the above operations are detailed in S1.1 to S1.4.
[0031] Step S1.1: After receiving the camera image data packet, extract the shooting time and acquisition unit identifier from the camera image data packet, and reconstruct the shooting sequence of each camera position according to the shooting time.
[0032] The camera image data packet refers to the data unit generated by the fixed camera position, encapsulated by the acquisition unit, and transmitted back. The camera image data packet contains image frame data and three fields written in the packet header: shooting time, camera position number, and acquisition unit identifier. The shooting order refers to the sequence of frames from the same camera position, arranged from earliest to latest shooting time.
[0033] Specifically, after receiving the camera image data packet, the data receiving server reads the shooting time field and the acquisition unit identifier field from the packet header. It extracts the year, month, day, hour, minute, second, and millisecond values from the shooting time field and the number string from the acquisition unit identifier field. Using the extracted shooting time millisecond values and the acquisition unit identifier as sorting criteria, a buffer queue is created in the local database for the local camera position, ordered in ascending order of millisecond values. The current camera image data packet is inserted into the buffer queue at the position matching the millisecond value. After insertion, the buffer queue is traversed to obtain the reconstructed shooting sequence. The reconstructed frames are then used for interval determination in S1.2.
[0034] Step S1.2: Determine whether the shooting time interval of adjacent data packets in the same acquisition unit exceeds the preset multiple range of the interval corresponding to the fixed frame rate.
[0035] It should be pointed out that when multiple cameras transmit back at the same time late at night, the lines become congested and the order in which the images arrive is disrupted. If the images are processed directly according to the order of arrival, it will cause the images to be misattributed. Therefore, after rearranging them according to the order of shooting, it is necessary to further check the continuity of the shooting time itself to expose the transmission anomalies.
[0036] Specifically, a fixed frame rate refers to the number of frames per second (fps) that the fixed-position camera is factory-calibrated for. The interval corresponding to the fixed frame rate is the difference between the shooting times of two adjacent frames, calculated from the reciprocal of the fps shooting time. For the buffer queue after reconstruction, the shooting times of two adjacent data packets are read sequentially. The shooting time interval is obtained by subtracting the millisecond value of the previous shooting time from the millisecond value of the subsequent shooting time. The shooting time interval is divided by the interval corresponding to the fixed frame rate to obtain the interval multiple. The interval multiple is compared with the lower and upper limits of a preset multiple range. When the interval multiple falls within the preset multiple range, the shooting times of adjacent data packets are determined to be continuous, and the current adjacent data packets are marked as a normal pair. When the interval multiple exceeds the preset multiple range, it is determined that there may be out-of-order or missing frames between adjacent data packets, and the process proceeds to S1.3 for repositioning. In this application, the preset multiple range is obtained by statistically analyzing the distribution range of the interval multiple during the normal backhaul period at the test station; the preferred setting range is 0.5 times to 2 times; in this embodiment, considering the transmission jitter at night, 0.5 times to 2 times is selected as the preset multiple range and written into the interval judgment logic.
[0037] Step S1.3: When the time limit is exceeded, check the acquisition unit identifiers of all data packets that have arrived within the corresponding time period and return the screen belonging to the local unit to its original position.
[0038] Furthermore, when S1.2 determines that the interval multiple of adjacent data packets exceeds a preset multiple range, the start and end times of the shooting time corresponding to the adjacent data packets are recorded, and the start and end times of the shooting time are used to form the time period to be reviewed. For example, the shooting times of adjacent data packets are 22:31:05.200 and 22:31:07.900, respectively. The fixed frame rate corresponds to an interval of 40 milliseconds, and the interval multiple is 67.5, which exceeds the preset multiple range. The time period to be reviewed is the period between the two shooting times.
[0039] Specifically, the process iterates through all data packets that have arrived in the local database within the time period to be reviewed. For each arriving data packet, the acquisition unit identifier is read, and a string comparison is performed between the read acquisition unit identifier and the acquisition unit identifier registered on the local camera. Arriving data packets that match are considered to belong to the local camera, while those that do not match are kept in their original queue without being moved. The images determined to belong to the local camera are then re-inserted into the corresponding positions in the local camera's buffer queue in ascending order of capture time. After insertion, the interval check in S1.2 is executed again, and the re-inserted images continue to participate in subsequent processes.
[0040] Step S1.4: When images with the same identifier appear at the same shooting time but belonging to different camera positions, send the first self-test request to the corresponding acquisition unit respectively. The image with the earlier shooting time in the response shall be used as the standard, and the image after reconstruction order shall be delivered to the entry point for verification as the camera image data packet to be verified.
[0041] It is understandable that if two adjacent camera positions share the same type of acquisition unit and the numbering is configured incorrectly, two images with the same acquisition unit identifier but different camera position numbers will appear at the same shooting time. If they are sent for identification without any handling, the image ownership will be permanently misplaced. Therefore, it is necessary to make a decision on the conflict of the same identifier before delivery to the entry point for verification.
[0042] Specifically, the buffer queue after reconstruction is traversed, and the shooting time field and acquisition unit identifier field of any two frames are compared pairwise. When the comparison result shows that the shooting time field value is the same and the acquisition unit identifier field string is the same, but the camera position number field string is different, it is determined that a conflict of images with the same identifier has occurred. For the conflict of images with the same identifier, the network address of the acquisition unit corresponding to each camera position number involved in the conflict of images with the same identifier is read from the local database, and a first self-test request is sent to each acquisition unit. The first self-test request is written with the conflicting shooting time and the conflicting acquisition unit identifier. The first self-test request is used to command the acquisition unit to send back the images of the two frames before and after the shooting time in the local cache, as well as the camera self-test status field.
[0043] Furthermore, within a preset waiting time after issuing the first self-test request, the system receives response images returned by each acquisition unit, reads the capture time in the header of each response image, compares the capture times of each response image, retains the response image with the smallest capture time value, and discards the remaining response images with larger capture time values. If only one response image is received within the preset waiting time, that response image is directly used as the decision result; if no response image is received within the preset waiting time, both conflicting images are placed in the verification queue and a questionable attribution mark is generated, awaiting manual handling. The images retained by the decision are rewritten into the buffer queue to complete the reconstruction order, and the reconstructed images, along with the three elements in the header, are encapsulated into a body image data packet to be verified and delivered to the entry verification of S2. In this application, the preset waiting time is obtained by statistically analyzing the historical response delay of the request message by the acquisition unit; the preferred setting range is 2 seconds to 10 seconds; here, 5 seconds is selected as the preset waiting time and written into the self-test waiting logic.
[0044] Step S2: Based on the shooting time, camera position number, acquisition unit identifier, and flight data, perform entry verification on the aircraft image data packet to obtain the entry verification result.
[0045] It should be noted that most image recording chain binding errors stem from including images of unknown origin in the recognition process. This step is designed to establish a gate before recognition begins, cross-checking the image's inherent capture time, gate number, and acquisition unit identifier against the flight data. Only images that pass the cross-check are allowed to proceed to occlusion detection and numbering / segmentation; images that fail the cross-check are suspended. The operations following this step include comparing the three elements with their corresponding fields in the flight data, sending a confirmation message to the fixed camera at the gate, and sending a notification message to the apron duty terminal. The specific execution methods of these operations are detailed in S2.1 to S2.3.
[0046] Step S2.1: Compare the shooting time with the preset time range corresponding to the expected docking time in the navigation data, compare the gate number with the gate number in the navigation data, and compare the acquisition unit identifier with the preset gate acquisition unit registration form.
[0047] It should be noted that the three comparison actions of the entry verification are aimed at three types of risks of the image source. The shooting time comparison is used to intercept irrelevant images outside the docking time, the camera position number comparison is used to intercept images with incorrect camera position input, and the acquisition unit identifier comparison is used to intercept images with cross-connection of acquisition units. The three comparison actions constitute a complete source verification chain.
[0048] Specifically, the system reads the estimated docking time of the flight corresponding to the current aircraft stand from the flight data interface. Extending this estimated docking time forward and backward by a preset time interval, a preset time range is obtained. The system reads the capture time from the aircraft view data packet header and compares the hour, minute, and second values of the capture time with the upper and lower limits of the preset time range. A match is determined if the capture time falls within the preset time range, and otherwise, a mismatch. The system compares the aircraft stand number field in the aircraft view data packet with the aircraft stand number field in the flight data. A match is determined if the strings are identical. The system reads a preset aircraft stand acquisition unit registration table from the local database. This table records the correspondence between each aircraft stand number and a unique acquisition unit identifier. The system searches for the row corresponding to the current aircraft stand number in the preset aircraft stand acquisition unit registration table and compares the acquisition unit identifier in that row with the acquisition unit identifier in the image header. A match is determined if they are identical. In this application, the preset time range is obtained by statistical analysis based on the average time spent on the bridge operation; the preferred setting range is 30 minutes before and after the bridge to 60 minutes before and after the bridge; in this embodiment, 45 minutes before and after the bridge is selected as the preset time range and written into the comparison logic.
[0049] Step S2.2: When all comparisons match, a confirmation message is returned to the fixed camera at the camera position to obtain the entry verification result that allows recognition.
[0050] Furthermore, the judgment results of the three items in S2.1—shooting time comparison, camera position number comparison, and acquisition unit identifier comparison—are statistically analyzed. When all three judgment results match, the return address of the fixed camera at the current position is read from the local database, a confirmation message containing the camera position number and data packet number is generated, and the confirmation message is sent to the fixed camera at the position. After receiving the confirmation message, the fixed camera at the position continues to transmit subsequent images at the original frame rate. The status field of the current aircraft image data packet is written into the local database as an allowable recognition flag. The entry verification result with the allowable recognition flag written into it, along with the aircraft image data packet, is delivered to the occlusion area detection in S3. For example, if the shooting time of a certain image is 22:45:10, the estimated docking time in the navigation data is 22:43, and the preset time range is 21:58 to 23:28, the shooting time falls within the range, and the comparison of the camera position number and the acquisition unit identifier in the registration form are both consistent. All three items match, and the image obtains the allowable recognition entry verification result and enters the occlusion area detection.
[0051] Step S2.3: If any comparison fails, the camera image data packet is placed in the verification queue, and a first prompt message containing the camera position number and shooting time is sent to the apron duty terminal to obtain the entry verification result of the suspension recognition.
[0052] On the other hand, when any one of the three comparisons in S2.1 results in a mismatch, a suspension recognition decision is generated.
[0053] Specifically, a list of teams to be checked is established in the local database. The current aircraft image data packet, along with the three fields of shooting time, aircraft position number, and acquisition unit identifier in the packet header, is written into a row of the list of teams to be checked, and the status field of this row is set to "suspended identification". The aircraft position number field and shooting time field are read from the list of teams to be checked, concatenated to generate the body of the first prompt message, and the first prompt message is sent to the apron duty terminal via wireless link. The screen of the apron duty terminal displays the aircraft position number, shooting time, and name of the mismatch, allowing duty personnel to verify whether it is a flight data entry error or an error in the image source. Before manual verification is completed and the handling result is returned, the suspended aircraft image data packet does not enter the S3 obstruction area detection. The entry verification result of the suspended identification is synchronously written into the candidate status of the bound status identifier for subsequent review.
[0054] Figure 2 This demonstrates the execution logic of the aircraft image data packet entry verification. After acquiring the aircraft image data packet, the shooting time, aircraft position number, and acquisition unit identifier written in the data packet are compared with the flight data. When all three comparisons match, a confirmation message is returned to the fixed camera at the aircraft position, thus obtaining the entry verification result allowing recognition. When any comparison fails, the aircraft image data packet is placed in the verification queue, and a prompt message is sent to the apron duty terminal, thus obtaining the entry verification result for suspended recognition. This process intercepts unreliable images before recognition begins, ensuring the accuracy of subsequent data processing.
[0055] Step S3: Perform occlusion area detection on the camera body image that has passed the entry verification, and write the unoccluded fragments in the numbered areas of each frame and the shooting time into a fragment sequence arranged in chronological order of shooting time.
[0056] Understandably, during winter nights, shuttle buses continuously pass through the passageway next to the camera positions, obscuring a portion of the numbered area of each frame's image. If the obscured portion of the entire frame is removed, the identification will lack a data foundation. Therefore, the purpose of this step is to extract and save the unobscured portion of each frame separately, forming a data structure organized by the shooting time. The operations following this step include marking the obscured areas, counting the cumulative number of times the area is visible, determining whether the count has reached the target, triggering frame rate increases and re-capture, and generating an insufficient capture flag if re-capture is still insufficient. The specific execution methods of the above operations are detailed in S3.1 to S3.5.
[0057] Step S3.1: For each frame of the camera body image, detect the location marker of the obscured part in the numbered area, and write the unobscured fragments and the shooting time into the fragment sequence.
[0058] It should be noted that the serial number is composed of a fixed number of character parts. For example, the serial number of a common model consists of six character parts. The object of occlusion detection is not the entire image, but each character part within the numbered area. Only by marking each part can subsequent splicing be supported.
[0059] Specifically, based on the fixed positions of the numbered areas in the body painting standard, numbered areas are divided on each frame of the camera body image. These numbered areas are then divided horizontally into intervals equal to the number of numbered characters, and each interval is assigned a sequential position identifier. For each interval, the difference between the maximum and minimum pixel brightness values and the color distance between the pixels within the interval and the reference value of the camera body background color are calculated. When the brightness difference exceeds a preset brightness difference threshold or the color distance exceeds a preset color distance threshold, the interval is determined to be occupied by an occluded part, and the occlusion status is written into the position identifier of that part. The area outside the occluded part is cropped into image segments. These image segments, along with the frame's shooting time and the position identifier, are written into a fragment sequence. The fragment sequence is sorted in ascending order by shooting time in the local database. Each record in the fragment sequence contains three fields: fragment image, shooting time, and position identifier. The fragment sequence is used for statistical analysis in S3.2.
[0060] Step S3.2: Count the cumulative number of times each part in the fragment sequence is visible in the sequence.
[0061] The cumulative number of visible occurrences refers to the total number of unobstructed fragments corresponding to the same position identifier in the fragment sequence.
[0062] Specifically, the entire sequence of fragments is traversed, and the location identifier carried in each record is read. A number of count fields, equal to the number of parts, are created in the local database. For each record read, the count field value corresponding to the location identifier is incremented by one. After traversal, the values of each count field are read to obtain the cumulative number of times each part is visible in the sequence. The cumulative number of visible times for each part is combined with its part number to form a visible count list, which is used for judgment in S3.3. For example, if there are six parts with different numbers, after traversing 120 fragment records, the count field values for the six parts are 97, 40, 86, 120, 63, and 97, respectively, forming the visible count list.
[0063] Step S3.3: When the cumulative number of visible times of each part reaches the preset minimum number, a fragment sequence that can be used for splicing is obtained.
[0064] It should be pointed out that single-frame fragments are subject to recognition errors due to the influence of nighttime lighting and de-icing liquid film. Only when a sufficient number of fragments accumulate in a certain part can multi-frame cross-validation have a data basis. Therefore, it is necessary to set a standard condition for the cumulative number of visible times.
[0065] Specifically, the value of each count field in the visible count list obtained in S3.2 is compared one by one with the preset minimum count. When the values of all count fields in the visible count list are greater than or equal to the preset minimum count, the standard is met. In the local database, the status field of the current fragment sequence is changed from "under acquisition" to "available for splicing". The fragment sequence with the status field "available for splicing" is the output of this step and is delivered to S4 for splicing. At the same time, the status rewriting event is recorded in the acquisition log. In this application, the preset minimum count is obtained by statistically analyzing the inflection point of the curve of the number splicing accuracy as the number of fragments increases. The preferred setting range is thirty to eighty times. In this embodiment, fifty times is selected as the preset minimum count and written into the standard-measuring logic.
[0066] Step S3.4: When there are parts that have not reached the preset minimum number of times, send an instruction to increase the frame rate to the fixed camera and arrange a supplementary sampling period outside the time period when the shuttle car passes through the channel. After the supplementary sampling, write the obtained fragments into the fragment sequence and recount the cumulative number of times visible until the cumulative number of times visible for each part reaches the preset minimum number of times, and obtain a fragment sequence that can be used for splicing.
[0067] Furthermore, when there are parts in the visible count list whose count field value is less than the preset minimum count, the position identifier of the corresponding part is read from the count field with a count less than the preset minimum count to form a gap part list.
[0068] Specifically, a command to increase the frame rate is sent to the fixed camera at the aircraft stand, containing the target frame rate value. Upon receiving the command, the fixed camera switches its shooting frame rate from the original fixed frame rate to the target frame rate. The schedule times for shuttle buses passing through the adjacent passageway of this aircraft stand are read from the airport operation plan database. The continuous time period closest to the current time but outside the scheduled time period is identified as the supplementary shooting period. This supplementary shooting period, along with a list of missing parts, is written into a supplementary shooting task message and sent to the fixed camera at the aircraft stand. During the supplementary shooting period, the fixed camera transmits supplementary images at the target frame rate. For each supplementary shooting image, the same occlusion detection as in S3.1 is performed. The supplementary fragments, along with the shooting time, are appended to the fragment sequence. After each batch of supplementary fragments is appended, the counting in S3.2 and the comparison in S3.3 are re-executed. When all count fields reach the preset minimum number of times, the fragment sequence status field is rewritten as usable for splicing, and the fragment sequence is delivered to S4. If the target is still not met, the next round of supplementary shooting tasks continues to be generated until the target is met or S3.5 is triggered.
[0069] For example, the count field value of the second part out of the six parts is forty, which is lower than the preset minimum number of fifty. The list of missing parts only includes the second part. The shuttle bus scheduling period is from 22:30 to 22:40. It is determined that the supplementary sampling period is from 22:42 to 22:52. The frame rate of the fixed camera is increased from 25 frames per second to 50 frames per second. After ten minutes of supplementary sampling, the count field value of the second part increases to seventy-one. All parts meet the standard, and the fragment sequence is turned into a sequence that can be used for splicing.
[0070] Step S3.5: If there are still parts that have not reached the preset minimum number of collections after supplementary collection, generate an insufficient collection flag and retain the written fragments in the fragment sequence.
[0071] Furthermore, when the number of times a supplementary mining task is executed reaches the preset upper limit of the number of supplementary mining rounds, and there are still parts in the visible count list whose count field value is less than the preset minimum number of times, the generation of new supplementary mining tasks will stop.
[0072] Specifically, an insufficient acquisition flag is generated in the local database. This flag contains the location identifier of the non-compliant area, the latest cumulative visibility count, and the number of re-acquisition rounds. The flag is also written to the acquisition log of the current camera position and is used by S5.10 for anomaly type determination. All fragments already written in the fragment sequence are retained as is, without deleting any records. The fragment sequence status field remains "acquiring," and the retained fragments continue to serve as baseline data for subsequent re-acquisition comparisons. In this application, the preset upper limit for the number of re-acquisition rounds is obtained by calculation based on flight transit time and single-round re-acquisition duration; the preferred setting range is three to six rounds; here, four rounds are selected as the preset upper limit for the number of re-acquisition rounds and written into the re-acquisition termination logic.
[0073] Step S4: Obtain candidate numbers by splicing together the fragment sequence.
[0074] It should be noted that the fragments within the fragment sequence belong to different frames, and the degree of influence of the de-icing liquid film varies among different frames. If all fragments are directly synthesized and identified without screening, the liquid film trailing will cause the identification results to jump. Therefore, the purpose of this step is to select frames with controllable quality for each character part to participate in the identification, and then assemble the characters identified from each part into a complete number according to their positional order. The splicing process is closely coupled with the subsequent comparison process in terms of data flow. The specific actions include detecting the state of the liquid film on the de-icing screen, selecting frames to participate in the identification according to the state, backtracking to dry screens when there is a conflict, marking pending characters when there are no dry screens, and splicing character by character. The specific execution methods of the above operations are detailed in S5.5 to S5.9. The candidate number obtained by splicing is delivered to S5 for double comparison.
[0075] Step S5: Compare the candidate number with the first comparison benchmark and the candidate numbers of adjacent gates, and determine the binding status identifier between the fuselage number and the flight number in the flight data based on the comparison results.
[0076] It should be noted that the candidate number is obtained by splicing together fragments from multiple frames, and there is still a possibility of splicing together the number of the same aircraft model at adjacent aircraft positions. Therefore, the purpose of this step is to perform a double verification using the complete fuselage image of the aircraft and the candidate number of the adjacent aircraft position. Only those that pass the verification are written into the binding success status flag, and those that fail the verification are transferred to supplementary collection or anomaly handling. The operations under this step include extracting the first comparison benchmark and comparing segment by segment, querying the candidate number from the adjacent aircraft position, determining the validity of the candidate number or backtracking for a second judgment, as well as liquid film status detection, frame selection, dry image backtracking, pending character marking and character-by-character splicing, and also includes anomaly type determination and separate handling of the three types of anomalies. The specific execution methods of the above operations are detailed in S5.1 to S5.13.
[0077] Step S5.1: Extract the complete fuselage image of the aircraft taken when the corresponding flight taxis in as the first comparison benchmark, and compare the candidate number with the visible part of the number in the first comparison benchmark segment by segment.
[0078] The first comparison benchmark refers to the complete fuselage image captured by the fixed camera at the aircraft position during the flight's taxiing phase when the fuselage is unobstructed and without liquid film. The first comparison benchmark contains the taxiing time field and the image of the numbered area.
[0079] Specifically, the local database is searched for the image index table by camera position number and taxiing time. The search criteria are that the camera position number is equal to the current camera position number, the shooting time is earlier than the expected docking time, and the image integrity field is complete. The retrieved images are extracted, and the numbered areas of the extracted images are marked as the first comparison benchmark. Each part of the candidate number is compared segment by segment with the corresponding visible part of the number in the first comparison benchmark. The segment comparison method is to perform character recognition on the corresponding part of the first comparison benchmark, and compare the recognized characters with the characters at the same position of the candidate number. The comparison result of each part is recorded as a match or a mismatch. The comparison result list is used by S5.3 for judgment.
[0080] Step S5.2: Send a query request to the adjacent camera acquisition unit and receive the candidate numbers of the adjacent camera positions.
[0081] Furthermore, read the two camera positions that are physically adjacent to the current camera position number from the preset camera position acquisition unit registration table, and perform the following actions for each adjacent camera position number.
[0082] Specifically, a query request message is generated, containing the current gate number, current flight number, and query time. This message is then sent via the airport intranet to the collection unit of the adjacent gate. The collection unit of the adjacent gate forwards the query request message to the processing terminal of the adjacent gate executing the identification process. If its own candidate number has been generated, the processing terminal of the adjacent gate encapsulates the candidate number along with the adjacent gate number and adjacent flight number into a query response message and sends it back. If its own candidate number has not yet been generated, it sends back a response message carrying an incomplete field. Upon receiving the query response message, the adjacent gate candidate number carried within it is written to the adjacent comparison table in the local database. This adjacent comparison table is available for S5.3 and S5.4 to read. If the adjacent gate does not respond within the preset response time limit, a "not responded" field is written to the corresponding row in the adjacent comparison table.
[0083] Step S5.3: When the candidate number matches the first comparison benchmark and is different from the candidate number of the adjacent station, the candidate number is determined to be valid, and a binding status identifier indicating successful binding is written.
[0084] Understandably, the core idea of dual comparison includes proving that the first comparison benchmark number does indeed belong to the aircraft parked at this position, and proving that the adjacent difference number has not been misspelled as the number of the adjacent aircraft of the same model. Both conditions must be met simultaneously to confirm the binding.
[0085] Specifically, the comparison result list obtained in S5.1 and the adjacent comparison table obtained in S5.2 are read. When all parts of the comparison result list match, and the adjacent gate candidate number exists in the adjacent comparison table, and the candidate number differs from the adjacent gate candidate number by at least one character in a character-by-character comparison, the candidate number is deemed valid. The candidate number, the current flight number, and the "binding successful" message are assembled into a binding status identifier. This binding status identifier is written to the row corresponding to the current gate number in the binding table of the local database. The writing time and the written candidate number are recorded together in the binding table. The binding status identifier for successful binding is read by the docking control in S6 to S10.
[0086] For example, if the candidate number is B1234, the first comparison benchmark identification result matches segment by segment. The candidate number of the adjacent camera on the left is B5678. The character-by-character comparison shows a difference at the third character. Both conditions are met at the same time, and B1234 is written into the binding table and the binding is successful.
[0087] Step S5.4: When the candidate number is the same as or has a similarity higher than the preset similarity condition with the candidate number of the adjacent camera position, backtrack to query the earliest shooting time of the fragment sequence for secondary judgment. If the secondary judgment fails, the splicing is determined to be unsuccessful and the process is switched to the supplementary acquisition of the fragment sequence. Write the binding status identifier of the suspended binding.
[0088] It should be noted that the fuselage numbers of two aircraft of the same model from the same airline often differ by only one or two characters. Identification errors can easily lead to different characters being combined into the same characters. Therefore, when a candidate number is the same as or highly similar to its neighboring number, it cannot be directly rejected. It is necessary to go back to the earliest fragment for cross-verification.
[0089] Specifically, the candidate number is compared character by character with the candidate numbers of adjacent camera positions in the adjacent comparison table. The proportion of identical characters to the total number of characters in the number is counted. When a candidate number is completely identical to the candidate number of an adjacent camera position, or when the proportion of identical characters is higher than the threshold of the preset similarity condition, backtracking is triggered. The backtracking method is to read the fragment with the smallest shooting time value in the fragment sequence, perform character recognition on that fragment, and compare the recognized character with the character at the same position in the candidate number. If there is a mismatch in the character comparison result, the second judgment is determined to fail. When the second judgment fails, the splicing is determined to fail, a supplementary acquisition trigger message is generated, and the supplementary acquisition process in S3.4 is transferred. At the same time, the binding status identifier of the binding suspension is written into the binding table. The binding status identifier of the binding suspension records the adjacent number and the position identifier of the difference that triggered the backtracking. In this application, the preset similarity condition is obtained by statistically obtaining it based on the smallest character difference between adjacent registration numbers of the same model; the preferred setting range is 80% to 90% of the identical characters; in this embodiment, the threshold of the preset similarity condition, which exceeds 83%, is selected and written into the similarity judgment logic.
[0090] Figure 3 This demonstrates a dual verification logic for candidate numbers and comparison benchmarks, as well as adjacent aircraft position numbers. After extracting the complete fuselage image of the aircraft as the first comparison benchmark and receiving candidate numbers from adjacent aircraft positions, it determines whether the candidate number matches the first comparison benchmark and is different from the adjacent numbers. If the conditions are met, the candidate number is deemed valid and a successful binding flag is written. If the conditions are not met, a secondary judgment is made by backtracking to the earliest captured fragment in the fragment sequence. If the secondary judgment passes, the candidate number is deemed valid; if the secondary judgment fails, the stitching is deemed unsuccessful, and the process switches to fragment sequence re-acquisition, while a pending binding flag is written. This mechanism effectively prevents the stitching of adjacent aircraft numbers.
[0091] Step S5.5: Detect the liquid film coverage status frame by frame in the numbered area of the de-icing screen to obtain the liquid film status identifiers of liquid film flow status, liquid film stillness status, or liquid film receding status.
[0092] It should be noted that after the aircraft is sprayed with de-icing fluid on the de-icing pad, the liquid film flows downward along the fuselage. The degree of interference with the edges of the numbered strokes varies in the three stages of the liquid film's flow, stillness, and dissipation. The state of the liquid film must be determined frame by frame before identification to determine which frames are reliable.
[0093] Specifically, for each frame of the de-icing scene, a numbered region image is cropped, and the absolute value of the difference in the average brightness of pixels in adjacent rows within the numbered region image is calculated. The absolute values of the differences in each row are arranged vertically to obtain a brightness distribution sequence. The brightness distribution sequences of adjacent frames are compared row by row. When the peak position of the brightness distribution sequence between two frames moves vertically and the amount of movement exceeds a preset movement threshold, the liquid film flow state is determined. When the peak position of the brightness distribution sequence remains unchanged for several consecutive frames and the peak value is higher than the dryness reference value, the liquid film static state is determined. When the peak value of the brightness distribution sequence falls back to near the dryness reference value and the difference is less than a preset fallback threshold, the liquid film receding state is determined. The determined state, along with the frame's shooting time, is written into a liquid film state identifier. The liquid film state identifier is stored in the local database frame by frame for selection in S5.6 and for retrospective use in S5.7. In this application, the preset movement threshold and preset fall-off threshold are obtained by continuously shooting and calibrating the body after spraying on the de-icing pad; the preferred setting range is two to five lines of movement threshold and 10% to 20% of the fall-off threshold reference value; here, three lines of movement threshold and 15% of the fall-off threshold reference value are selected and written into the liquid film determination logic.
[0094] Step S5.6: Select frames for identification from each frame in order of priority for liquid film receding state and second for liquid film stationary state. Frames with liquid film flow state are only used to provide position information in the residual sequence.
[0095] Among them, the frames involved in recognition refer to the de-icing screen frames selected for character recognition.
[0096] Further, the system reads the liquid film state identifiers stored in the local database frame by frame, dividing all frames into three sets: liquid film receding state frames, liquid film stationary state frames, and liquid film flowing state frames. When the liquid film receding state frame set is not empty, frames for recognition are selected from it in ascending order of shooting time. When the liquid film receding state frame set is empty but the liquid film stationary state frame set is not empty, frames for recognition are selected from it in ascending order of shooting time. Frames in the liquid film flowing state frame set do not participate in character recognition; only the horizontal position coordinates of each fragment are extracted from the liquid film flowing state frames. The extracted horizontal position coordinates are written into the position information field of the corresponding record in the fragment sequence to confirm that the relative position of each part in the image has not shifted. The selection results are written into the selected frame list, which is read by S5.9 during stitching.
[0097] Step S5.7: When the characters identified at the same number location in the liquid film receding state frame and the liquid film stationary state frame are different, backtrack to query the dry image before the de-icing started, and use the identification result of the dry image to participate in the stitching.
[0098] It is understandable that during the static phase, the surface tension of the liquid film will cause the edges of the numbered strokes to become rough. Individual characters with similar shapes may be identified differently in static and fading frames. At this time, the only basis for judgment is the image where the numbering is still dry before de-icing begins.
[0099] Specifically, for each numbered location, character recognition is performed on selected frames in both the liquid film receding state frame and the liquid film stationary state frame, and the two recognition results are compared. When the comparison results are different, a dry image is retrieved from the local database based on the condition that the shooting time is earlier than the de-icing start time. The de-icing start time of the dry image is read from the de-icing operation record table. After a dry image is retrieved, the area corresponding to that location in the dry image is cropped, and character recognition is performed on the cropped area. The recognition result of the dry image is used as the character for that location in the stitching, and the dry backtracking event and all values of the three recognition results are recorded in the stitching log. For example, if the third location is recognized as character 8 in the receding frame and character 6 in the stationary frame, and the recognition result is character 8 in the dry image 20 minutes before the start of de-icing, then character 8 is used as the character for the third location in the stitching log. If no dry image is retrieved, the process proceeds to step S5.8.
[0100] Step S5.8: If no dry screen is available before de-icing, mark the characters in the corresponding area as pending characters.
[0101] It should be noted that if the local database cannot retrieve dry footage taken before the start of de-icing, or if the corresponding part in the retrieved dry footage is obscured by the camera's shadow or ground equipment and cannot be identified, then that part loses its credible basis for judgment.
[0102] Specifically, the part numbers of undetermined areas are written into a pending character marker. This marker records two values: the part's location identifier, the liquid film fading state frame recognition result, and the liquid film stationary state frame recognition result. The pending character marker is then written into a pending character table in the local database, which is read during the generation of the reshoot request in S5.81. Parts with pending character markers participate as placeholders during stitching in S5.9; this placement does not affect the normal stitching of other parts.
[0103] Step S5.9: Concatenate the characters obtained from the recognition of each part to obtain the candidate number.
[0104] Specifically, the system reads the final determined character sources for each numbered segment. These sources include four types: liquid film fading state frame recognition results, liquid film still state frame recognition results, dry image recognition results, and comparison results with reshot images. Starting from the first segment, characters are extracted sequentially according to their positional identifiers, and then concatenated into a string. When a pending character marker exists, a placeholder is inserted at the corresponding position before concatenating the remaining characters to obtain a candidate number with placeholders. The concatenated string is written to the candidate number field in the local database. The candidate number, along with the character source record in the concatenation log, is saved. The candidate number is then submitted to S5.1 for segment-by-segment comparison, and S5.83 performs character-by-character comparison by reading the candidate number field.
[0105] Step S5.81: When there are pending characters, send a first reshoot request with the specified camera serial number location and suggested shooting angle to the handheld shooting device of the on-duty inspection personnel.
[0106] It should be pointed out that when inspectors use handheld cameras to take pictures closer to the camera body and the shooting angle is adjustable, the clarity of the handheld footage is higher than that of the footage taken from a distance. By entrusting the determination of pending characters to handheld footage, the basis for judgment can be supplemented when dry footage is missing.
[0107] Specifically, the location identifiers in the pending character table of the local database are read, and these identifiers are converted into physical location descriptions on the fuselage surface. The conversion method is based on the fixed dimensional relationship between the numbered area and the nose and porthole in the aircraft model's paint scheme, resulting in the fuselage position interval for the fuselage number. Based on the position interval and the photographer's standing direction, a suggested shooting angle is read from a preset shooting angle lookup table. This table records the correspondence between each position interval and the pitch and yaw angles. The fuselage position interval, suggested shooting angle, and pending part number are assembled into a first reshoot request. This first reshoot request is sent to the handheld shooting device registered by the on-duty aircraft inspection personnel, and the time of sending the request is recorded simultaneously for reference during S5.86 retrospective. In this application, the preset shooting angle comparison table is obtained by calibrating each model by shooting at the stopping position; the preferred setting range is pitch angle from -10 degrees to +10 degrees and yaw angle from ±30 degrees; here, a pitch angle of 0 degrees and a yaw angle from ±20 degrees are selected and written into the angle comparison logic.
[0108] Step S5.82: Receive the reshot footage and shooting location markers transmitted back from the handheld shooting device.
[0109] Among them, the shooting location identifier refers to the positioning field that is transmitted back by the handheld shooting device along with the reshot footage. The positioning field contains the shooting time, device number and shooting direction record.
[0110] Specifically, within a preset reception time limit, the system listens for data packets transmitted from the handheld shooting device. Upon receiving a transmitted data packet, it reads the shooting location identifier and image frame data from the packet header. It then compares the device number in the shooting location identifier with the registered on-duty device number. If they match, the transmitted data packet is registered as a valid reshoot image and written to the reshoot image table in the local database. If they do not match, the image frame data is discarded, and a retransmission request is sent back to the handheld shooting device. If no transmitted data packet is received within the preset reception time limit, the first reshoot request is retransmitted to the handheld shooting device. After the number of retransmissions reaches the preset retransmission limit, a "no response" field is written to the corresponding row of the pending character table. In this application, the preset reception time limit is obtained by statistical analysis based on the average time consumed per single inspection operation; the preferred setting range is 5 to 15 minutes; here, 10 minutes is selected as the preset reception time limit and written into the monitoring logic.
[0111] Step S5.83: Compare the identification result of the number in the reshot image with the splicing result of the fragment sequence character by character.
[0112] Furthermore, valid re-shot images are read from the local database re-shot image table, the numbered areas in the re-shot images are cropped, and character recognition is performed on the numbered areas to obtain the re-shot recognition string.
[0113] Specifically, the candidate number string with placeholders is read from the candidate number field. The supplementary image recognition string is compared with the candidate number string with placeholders character by character position. The comparison rules are: if the characters at the corresponding positions are the same, it is considered a match; if they are different, it is considered a mismatch; if the candidate number has a placeholder character at the corresponding position but the supplementary image recognition string contains the character, it is considered a completion. The number of matches, mismatches, and completions are counted, and the comparison difference records are written to the local database. When the number of mismatches is zero, proceed to S5.84; when the number of mismatches is greater than zero and less than or equal to the preset difference upper limit, proceed to S5.85; when the number of mismatches is greater than the preset difference upper limit, proceed to S5.86. In this application, the preset difference upper limit is obtained by statistically analyzing the historical recognition discrepancy ratio between the handheld image and the camera position image; the preferred setting range is 1 to 2 characters; here, 1 character is selected as the preset difference upper limit and written into the difference judgment logic.
[0114] Step S5.84: When all matches, remove the pending characters to obtain the determined candidate number. Step S5.85: When there is a partial match, update the fragment sequence based on the reshot image and then reassemble it to obtain the updated candidate number.
[0115] It should be noted that the comparison between the reshot footage and the spliced fragment sequence results can result in three outcomes: complete match, partial match, and complete mismatch. The handling of the first two outcomes is closely linked in the data flow. The comparison record serves as both the basis for removing pending characters and the input for updating the fragment sequence, forming an inseparable decision chain.
[0116] Specifically, when the number of mismatches in the comparison difference record of S5.83 is zero, all pending character markers in the pending character table are read, the character corresponding to the placeholder position in the supplementary identification string is extracted, and it is filled into the placeholder position of the candidate number string. The record of the corresponding row in the pending character table is deleted, and the filled string is rewritten back to the candidate number field. The candidate number changes from the placeholder state to the determined state, and the determined candidate number is obtained. The determined candidate number is then transferred to the double comparison of S5.3.
[0117] Furthermore, when the number of mismatches in the comparison difference record is greater than zero and less than or equal to the preset difference upper limit, the image of the numbered region cropped from the reshot image is segmented by part. The images of each segmented part are written into the fragment sequence along with the shooting time of the reshot image, and the reshot source field is marked in the fragment record during writing. The recognition and splicing of S5.5 to S5.9 are re-executed on the updated fragment sequence. During the splicing process, the fragment from the reshot source has the highest priority when there is a character conflict. After the re-sponsoring is completed, the new string is rewritten back to the candidate number field to obtain the updated candidate number. The updated candidate number is also transferred to the double comparison in S5.3.
[0118] For example, if the candidate ID with placeholder is B12_45, and the re-shot identification string is B12345, then a character-by-character comparison of five positions shows four matches, one completes the comparison, and there are no mismatches. The character 3 is then inserted into the placeholder position, and the candidate ID is determined to be B12345, proceeding to the double comparison. If a mismatch occurs at the fourth position in the comparison results, the fourth and fifth parts of the re-shot image are used as the source fragment for the re-shot and written into the fragment sequence. This is then reassembled to obtain the updated candidate ID, which is then returned to the double comparison.
[0119] Step S5.86: If there is a complete mismatch, backtrack the verification record of the liquid film status identifier and the order of shooting and reject the candidate number.
[0120] Furthermore, when the number of discrepancies in the comparison difference records of S5.83 exceeds the preset difference limit, it indicates that there is an overall contradiction between the camera position image recognition chain and the handheld image recognition chain, and there is no basis to continue taking the result of either one.
[0121] Specifically, the system reads the frame-by-frame records of the liquid film status identifiers from the local database, along with the shooting sequence verification records reconstructed in S1.1. The system then compares the judgment frame order of the liquid film status identifiers with the shooting sequence frame by frame. The verification method involves checking whether the liquid film status transitions over time according to a unidirectional sequence of flow, stillness, and decay, while simultaneously checking whether the shooting times of each frame monotonically increase and have no inverted records. Abnormal frame records discovered during the verification, along with comparison difference records, are written to the rejection log. A null value is written to the candidate number field, and a rejection candidate mark is added. The rejection candidate mark is then used for anomaly type determination in S5.10. The identification results of the current round for this location are no longer passed down.
[0122] Step S5.10: In response to the insufficient data collection flag, conflict with the same flag image, or difference in re-shot comparison exceeding the preset difference limit, determine the abnormality type of insufficient data collection, questionable attribution, and contradictory identification respectively, and process them in the order of questionable attribution, contradictory identification, and insufficient data collection.
[0123] It should be noted that multiple anomalies may occur simultaneously at the same camera position. If the order of handling is not appropriate, re-capturing footage from a camera position whose ownership has not yet been determined will result in a waste of resources. Therefore, it is necessary to specify the order of handling for the three types of anomalies.
[0124] Specifically, an anomaly registration table is established in the local database. Three types of triggering conditions are checked one by one: if an insufficient collection marker is found in the local database, an insufficient collection anomaly is registered; if a questionable attribution marker is generated in S1.4 or there are unprocessed conflicting records with the same marker, a questionable attribution anomaly is registered; if the number of discrepancies in the re-enhancing comparison difference records exceeds the preset difference limit or there are candidate rejection markers, an identification contradiction anomaly is registered. All anomaly types registered in the anomaly registration table are read and sorted in the order of questionable attribution, identification contradiction, and insufficient collection to generate a processing queue. Each anomaly in the processing queue is sequentially transferred to the corresponding processing procedure in S5.11, S5.12, or S5.13. After processing, the processing result is written back to the corresponding row in the anomaly registration table.
[0125] Step S5.11: When the classification is questionable, freeze all the images of the corresponding machine position that have been entered into the database, send a manual assignment request to the apron duty terminal, rebind according to the response result and write the binding status mark of successful binding or binding invalidation.
[0126] It should be noted that ownership in doubt means that the location of the camera in the footage has not yet been confirmed. If ownership is not confirmed, any identification and binding actions will lose their object basis. Therefore, cases with ownership in doubt are placed at the top of the processing queue.
[0127] Specifically, in the local database, the status field of all images already in the database associated with the corresponding gate number is changed from "normal" to "frozen." Frozen images cannot be read by any associated operations. A manual assignment request message is generated, containing the gate number, conflict acquisition unit identifier, conflict capture time, and involved flight number. This request is then sent to the apron duty terminal. The response from the apron duty terminal is received. If the response confirms the gate, the frozen status of the corresponding gate image is released, and the S2 entry verification is re-executed according to the confirmed gate number. If the verification passes, a successful binding status flag is written. If the response is negative, the binding status flag of the corresponding gate's already in the database is changed to "binding invalid," the frozen status is changed to "invalid archive," and the invalid image is retained in the local database for post-event verification.
[0128] Step S5.12: When identifying a contradictory type, send a second reshoot request containing the positions of the differing characters to the handheld shooting device, and write the reshoot comparison result into the binding status identifier.
[0129] Furthermore, the location identifiers of all mismatched characters are read from the comparison difference record, and the camera position interval corresponding to the location identifier is written into the second reshoot request. The second reshoot request also includes the suggested shooting angle and the candidate values of the mismatched characters.
[0130] Specifically, the second retake request is sent to the handheld shooting device. The handheld shooting device returns the retake image and shooting location identifier within a preset reception time limit. Character recognition is performed on the positions of the different characters in the retake image to obtain the retake identification character. The retake identification character is compared with the character at the same position in the reassembled fragment sequence. If the comparison result is consistent, the candidate number is corrected with the retake identification character, and the corrected candidate number is transferred to S5.3 for double comparison. If the comparison is successful, a binding status identifier is written to the binding success status. If the comparison result is still inconsistent, the binding status identifier is rewritten to binding suspended and a manual review record is generated. In this application, the preset reception time limit is marked at S5.82, and the preferred setting range is 5 minutes to 15 minutes. Here, 10 minutes is used and written into the retake monitoring logic.
[0131] Step S5.13: When there is insufficient data collection, send a channel control query to the apron dispatch terminal. After receiving a positive response, collect data centrally in the clear window and write the binding status flag indicating successful binding. When receiving a negative response, switch to manual binding and write the binding status flag indicating pending binding.
[0132] Understandably, the root cause of insufficient data collection is the continuous obstruction of the shuttle bus lane. This obstruction is a conflict between ground operation resources and recognition requirements, which cannot be solved by unilaterally increasing the frame rate. The scheduling side must provide an unobstructed clearing window. As an anomaly that relies on external resource coordination, insufficient data collection is placed at the end of the processing queue and is the final point of all recognition and binding actions.
[0133] Specifically, the system reads the location identifiers and latest cumulative visibility counts of the non-compliant areas written in the insufficient data collection flag, calculates the number of data collection frames required to reach the preset minimum number of frames, calculates the required clearing time based on the increased target frame rate, assembles the required clearing time, the gate number, and the list of non-compliant areas into a channel control query message, and sends the channel control query message to the apron dispatch terminal. The apron dispatch terminal queries the shuttle bus schedule and sends back a response result, which is either a positive response or a negative response.
[0134] Furthermore, upon receiving a positive response, the apron dispatch terminal specifies the start and end times of the clearing window in the response result. Before the start time of the clearing window arrives, it sends a window acquisition command to the fixed camera at the camera position. The fixed camera at the camera position captures images of the non-compliant parts at the target frame rate within the clearing window. After the acquired images undergo the same occlusion detection as in S3.1, the newly added fragments are written into the fragment sequence and the cumulative visibility count is recalculated. When the cumulative visibility count of all parts reaches the preset minimum count, the fragment sequence status field is rewritten to be usable for splicing, and the double comparison in S4 splicing and S5.3 is re-executed. If the comparison matches and is different from the candidate number of the adjacent camera position, the candidate number is determined to be valid, and a binding status identifier of successful binding is written. The binding status identifier of successful binding, along with the start and end times of the clearing window, is recorded in the acquisition log. After the clearing window ends, a frame rate recovery command is sent to the fixed camera at the camera position to restore the shooting frame rate to a fixed frame rate.
[0135] Meanwhile, when a negative response is received, if the apron dispatch terminal states in the response that the clearing window schedule exceeds the flight's transit time, and the difference between the flight's estimated departure time and the current time is less than the remaining time required for data collection, then it is determined that there is no time left for supplementary data collection. The binding method is then switched to manual binding, and a manual binding work order is sent to the apron duty terminal. The work order includes the identified partial numbers of the residual sequence, a list of non-compliant parts, and records of pending characters. The duty personnel verify the fuselage number on-site according to the work order and send back a confirmation result. The candidate number is completed upon confirmation of the result. The completed candidate number is written into the binding table and a binding status flag indicating that the binding is pending is written into the binding status flag, which is read by the pending queue of S9. If the reason for the negative response is that the channel cannot be controlled on that day but the flight departure time is sufficient, the insufficient collection flag is retained, and the channel control query is sent to the apron dispatch terminal again at the preset re-query interval until a positive response is received or the flight enters the departure countdown. If the apron dispatch terminal does not send back any response result within the preset query time limit, it is treated as a negative response and transferred to the manual binding process. In this application, the preset query time limit is obtained by statistically analyzing the historical response time of the dispatch terminal to the query message; the preferred setting range is 3 minutes to 8 minutes; in this embodiment, 5 minutes is selected as the preset query time limit and written into the query waiting logic.
[0136] Step S6: Based on the successful binding status identifier, establish an image recording chain for the corresponding flight, and sequentially attach the aircraft position screen, de-icing screen, and aircraft inspection screen, and write the binding status identifier at the beginning of the chain.
[0137] It should be noted that the image recording chain is a data structure that is jointly attached to the camera position screen, de-icing screen, and walk-around inspection screen. The head of the chain serves as the access credential for the entire chain. Only the binding status identifier that has been successfully bound is eligible to establish an image recording chain.
[0138] Specifically, the system retrieves records from the binding table in the local database whose status field indicates successful binding, and reads the candidate number, flight number, and aircraft stand number from these records. Image recording chain entries are established using the flight number as an index. Each chain entry reserves three storage areas: an aircraft stand view attachment area, a de-icing view attachment area, and a walk-around inspection view attachment area. The successful binding status identifier is written to the chain head field of the chain entry, which also contains the writing time and candidate number. Aircraft stand views that have passed entry verification are written to the aircraft stand view attachment area in the order they were captured; de-icing views are written to the de-icing view attachment area in the order they were captured; and walk-around inspection views are written to the walk-around inspection view attachment area in the order they were received. The attachment action is recorded in the local database for each instance, and the resulting image recording chain is used for pre-entry retrieval by the S7.
[0139] Furthermore, when the shooting time of the image written in the camera position image attachment area is earlier than the writing time at the beginning of the chain but the image has not yet been stored, it is attached in the order of shooting. When the image recording chain entry for the same flight already exists, the chain entry will not be created again, and the newly arrived image will be directly added to the corresponding attachment area. The image recording chain and the actual flight establish a unique correspondence through the candidate number at the beginning of the chain and the flight number.
[0140] Step S7: Before each de-icing screen or machine inspection screen is put into storage, check the binding status indicator of the chain head.
[0141] It should be noted that the attachment error occurs at the entry point of the image into the database. If the verification is performed after the image is entered into the database, the erroneous image will be mixed with the correct image, and the rollback cost will be high. Therefore, a rollback action should be performed before each image is written into the attachment area.
[0142] Specifically, when the de-icing screen or the walk-around inspection screen reaches the warehousing stage, the camera position number and acquisition unit identifier in the screen data packet header are read. Based on the camera position number, the corresponding image record chain entry is retrieved from the local database. The binding status identifier in the chain head field of the chain entry is read, and the retrieved binding status identifier is compared with the latest binding status identifier in the binding table. If they match, the content of the chain head field is taken as the standard, and the process proceeds to the branch judgment in S8 to S10. When no image record chain entry for the corresponding camera position number is found, the screen is written to the suspended buffer. After the chain head is established, the screen in the suspended buffer will re-execute the back-check action. The back-check result, along with the screen number, is recorded in the warehousing audit log for reference during the S11 back-check. For example, if the walk-around inspection screen carries camera position number 17, and the image record chain for camera position 17 is retrieved with the chain head field indicating successful binding, the process proceeds to the automatic docking branch in S8 after the back-check is passed.
[0143] Step S8: When the binding status at the head of the chain is marked as successful, the chain is automatically attached.
[0144] Understandably, a successful binding status indicates that the image source, number identification, and dual comparison have all passed, meeting the conditions for automatic data entry. Automatic binding avoids the time overhead of manual frame-by-frame confirmation.
[0145] Specifically, when S7 retrieves the binding status flag of the chain head as successful, it reads the image type field of the image to be added to the database. If the image type field is a camera position image, it is written to the camera position image aggregator; if the image type field is a de-icing image, it is written to the de-icing image aggregator; and if the image type field is a walk-around inspection image, it is written to the walk-around inspection image aggregator. During writing, the image's capture time, acquisition unit identifier, and the candidate number of the chain head are all written to the aggregator record, and the aggregator records are sorted by the writing time.
[0146] Furthermore, when the capture time of the image to be added to the database is earlier than the earliest capture time of the image already added to the database in the image recording chain, the image to be added to the database is inserted into the correct position in the attachment area according to the capture time sorting. After insertion, the sequence number of all records in the attachment area is rearranged. When the capture time of the image to be added to the database is later than the write time at the beginning of the chain and exceeds the preset attachment time window, the image is marked as a timeout image. The timeout image is still automatically attached, but an additional timeout mark is written. The timeout mark is recorded separately in the audit log. The status field of the image after automatic attachment is completed is rewritten as "added to the database".
[0147] In this application, the preset docking time window is obtained by statistical analysis based on the total transit time of flights; the preferred setting range is 2 to 5 hours; here, 3 hours is selected as the preset docking time window and written into the timeout judgment logic.
[0148] Step S9: When binding is suspended, enter the suspension queue to wait.
[0149] The suspended queue refers to the first-in-first-out queue list in the local database used to temporarily store the images corresponding to the image record chains whose binding status is marked as suspended.
[0150] Specifically, when S7 retrieves the binding status identifier at the head of the chain as "binding suspended," it writes the image to be added to the database, along with the image type field, shooting time, and camera position number, to the end of the suspended queue list, and sets the image status field to "suspended and waiting." It reads the suspension reason field recorded in the binding status identifier of the suspended image. Suspension reasons include three types: re-capture triggered, manual binding in progress, and secondary judgment pending. The suspension reason is associated with and saved with the image number. The local database repeatedly searches the binding status identifier of the corresponding camera position number in the binding table at preset polling intervals. When the binding status identifier is changed from "binding suspended" to "binding successful," all images of that camera position number in the suspended queue list are removed from the suspended queue in chronological order of shooting time and transferred to S8 for automatic attachment. When the binding status identifier is changed to "binding invalid," the corresponding images in the suspended queue list are transferred to the rejection process in S10. When the suspension waiting time exceeds the preset suspension time limit, a queue backlog notification message is sent to the apron duty terminal.
[0151] For example, three de-icing shots from camera position 16 were put into the suspension queue due to re-capture. Forty minutes later, the binding status of camera position 16 in the binding table was changed to "binding successful". The three de-icing shots were removed from the suspension queue according to the shooting time and all were automatically attached.
[0152] In this application, the preset polling interval and the preset suspension duration upper limit are obtained by statistically analyzing the average completion time of the binding process; the preferred setting range is a polling interval of 10 seconds to 60 seconds and a suspension duration upper limit of 30 minutes to 90 minutes; here, a polling interval of 30 seconds and a suspension duration upper limit of 60 minutes are selected and written into the suspension waiting logic.
[0153] Step S10: When binding is invalidated, the attachment is rejected and a first stop acquisition command is sent to the acquisition side.
[0154] It should be noted that invalidating the binding means that the ownership of the image has been manually rejected. Continuing to collect images will only generate more images that cannot be bound and consume storage and transmission resources. Therefore, the rejection action and the stop collection action must be executed in conjunction with each other in the same process.
[0155] Specifically, when S7 checks and finds that the binding status of the chain head is invalid, it refuses to write the image to be entered into any affiliation area. A rejection record is created in the local database for the rejected image, containing the image number, shooting time, rejection time, and the binding status of the chain head. The system reads the acquisition unit identifier and fixed camera address of the corresponding camera position from the preset camera position acquisition unit registration table, generates a first stop acquisition command, and includes the invalidation reason field and stop range field. The first stop acquisition command is sent to the fixed camera position and the acquisition unit. Upon receiving the command, the fixed camera position stops transmitting images back to that camera position, and upon receiving the command, the acquisition unit freezes the transmission channel for that camera position.
[0156] Furthermore, when the binding status identifier of the corresponding camera position number in the binding table is changed from binding invalid to other statuses, a collection recovery command is sent to the fixed camera and the acquisition unit at the camera position. The fixed camera resumes transmission at a fixed frame rate, and the recovered image goes through the S2 entry verification process again. When the binding status identifier remains binding invalid for a preset invalidation retention time, the rejected record is archived and the occupant record of the corresponding camera position in the suspension queue is released.
[0157] In this application, the preset retention time for obsolescence is obtained by statistically analyzing the average processing time of obsolescence locations manually checked; the preferred setting range is 1 hour to 4 hours; here, 2 hours is selected as the preset retention time for obsolescence and written into the archiving judgment logic.
[0158] Step S11: On the flight with the incorrect binding, trace back along the image recording chain to check the liquid film status identifier, acquisition unit identifier and shooting time of each frame, locate the error level, and generate a correction record that retains the original incorrect binding content and the corrected content.
[0159] It should be noted that binding errors may occur at three levels: image source, identification splicing, or attachment control. The purpose of frame-by-frame review is to locate the error to the specific level, provide a precise entry point for manual correction, and avoid the entire image recording chain being overturned and rebuilt.
[0160] Specifically, the system receives a binding error confirmation message from the apron duty terminal, which contains the incorrect flight number and the corrected flight number. Using the incorrect flight number as an index, the system retrieves the image recording chain from the local database, reading each frame sequentially from the beginning of the chain, including the liquid film status identifier, acquisition unit identifier, and capture time.
[0161] Specifically, the acquisition unit identifier of each frame is compared with the registration identifier of the corresponding gate for the corrected flight number in the preset gate acquisition unit registration table. Frames that do not match are recorded as source-level errors. The shooting time of each frame is compared with the preset time range corresponding to the expected docking time of the corrected flight. Frames falling outside this range are recorded as timing-level errors. The liquid film status identifier is checked against the unidirectional sequence of flow, stillness, and dissipation according to the transfer order of the shooting time. Frames that fail the check are recorded as identification-level errors. The number of erroneous frames at each of the three levels is counted, and the level with the largest number of erroneous frames is identified as the primary error level. The identification result is written into the correction work order. Generate a correction record. The correction record retains the original error binding content field and the corrected content field. The original error binding content field is written with the error flight number, the original candidate number, and the binding status identifier of the original chain head. The corrected content field is written with the corrected flight number, the corrected candidate number, and the binding status identifier of the new chain head. The correction record, along with the main error level and a complete list of error frames, is written to the correction archive table in the local database. The records in the correction archive table are available for post-audit reading and cannot be overwritten or modified.
[0162] Furthermore, when two or more layers of error frames have the same number of errors within the same image recording chain, the layer that appears first in the order of source layer, recognition layer, and time sequence layer is taken as the primary error layer. All images in the attachment area corresponding to the primary error layer are transferred to the re-verification process. After the re-verification is successful, the attachment order is reconstructed with the new chain head.
[0163] In this application, the preset time range is marked at S2.1. The preferred setting range is 30 minutes before and after each time step to 60 minutes before and after each time step. Here, 45 minutes before and after each time step is used and written into the time sequence comparison logic.
[0164] Example 2 is an embodiment of the present invention, which provides a civil aviation navigation data information processing system, including a data packet acquisition module, an entry verification module, a fragment sequence generation module, and a number binding module. The data packet acquisition module is used to acquire fuselage image data packets transmitted from fixed camera positions, and the fuselage image data packets contain the shooting time, aircraft position number, and acquisition unit identifier. The entry verification module is used to perform entry verification on the fuselage image data packets based on the shooting time, aircraft position number, acquisition unit identifier, and navigation data to obtain the entry verification result. The fragment sequence generation module is used to perform occlusion area detection on the fuselage images that have passed the entry verification, and write the fragments of the numbered areas in each frame that are not occluded and the shooting time into a fragment sequence arranged in chronological order of shooting time. The number binding module is used to obtain candidate numbers by splicing the fragment sequences, compare the candidate numbers with a first comparison benchmark and candidate numbers of adjacent aircraft positions, and determine the binding status identifier between the fuselage number and the flight number in the navigation data based on the comparison result.
[0165] The system proposed in this embodiment belongs to the same inventive concept as the method for processing civil aviation navigation data information proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.
[0166] See Figure 4 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. The electronic device 40 includes: a processor 41, a memory 42, and a computer program; wherein, The memory 42 is used to store computer programs, and the memory may also be flash memory. Computer programs may be, for example, application programs or functional modules that implement the methods described above.
[0167] Processor 41 is used to execute computer programs stored in memory to implement the various steps performed by the device in the above method.
[0168] For details, please refer to the relevant descriptions in the preceding method embodiments.
[0169] Alternatively, the memory 42 can be either standalone or integrated with the processor 41.
[0170] When the memory 42 is a device independent of the processor 41, the device may also include: Bus 43 is used to connect memory 42 and processor 41.
[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for processing civil aviation navigation data information, characterized in that, include: Acquire the camera body image data packet transmitted back by the fixed camera position, and write the shooting time, camera position number and acquisition unit identifier into the camera body image data packet; Based on the shooting time, the camera position number, the acquisition unit identifier, and the flight data, an entry verification is performed on the fuselage image data packet to obtain the entry verification result. For the camera body images that have passed the entry verification, perform occlusion area detection, and write the unoccluded fragments in the numbered areas of each frame and the shooting time into a fragment sequence arranged in chronological order of shooting time; Candidate numbers are obtained by splicing together the fragment sequence; The candidate number is compared with the first comparison benchmark and the candidate numbers of adjacent aircraft positions, and the binding status identifier between the fuselage number and the flight number in the flight data is determined based on the comparison results.
2. The civil aviation navigation data information processing method according to claim 1, characterized in that, Based on the shooting time, the camera position number, the acquisition unit identifier, and the flight data, an entry verification is performed on the aircraft image data packet to obtain the entry verification result, including: The shooting time is compared with the preset time range corresponding to the expected docking time in the navigation data; the gate number is compared with the gate number in the navigation data; and the acquisition unit identifier is compared with the preset gate acquisition unit registration form. When all comparisons match, a confirmation message is returned to the fixed camera at the location to obtain the entry verification result that allows recognition; If any comparison fails, the aircraft image data packet is placed in the verification queue, and a first prompt message containing the aircraft position number and shooting time is sent to the apron duty terminal to obtain the entry verification result of the suspension identification.
3. The civil aviation navigation data information processing method according to claim 1, characterized in that, For the camera body images that pass the entry verification, occlusion area detection is performed. The unoccluded segments in each frame, numbered regions, are written into a segment sequence arranged chronologically by shooting time, along with the shooting time. This includes: For each frame of the camera body image, the location marker of the obscured part of the detected number area is written into the fragment sequence along with the shooting time; Count the cumulative number of times each part of the fragment sequence is visible in the sequence; When the cumulative number of visible times in each part reaches the preset minimum number, a fragment sequence that can be used for splicing is obtained; When there are parts where the preset minimum number of times has not been reached, an instruction to increase the frame rate is sent to the fixed camera. A supplementary acquisition period is arranged outside the time period when the shuttle bus passes through the channel. After the supplementary acquisition, the fragment obtained by the supplementary acquisition is written into the fragment sequence and the cumulative number of times can be counted again until the cumulative number of times can be counted for each part reaches the preset minimum number of times, so as to obtain the fragment sequence that can be used for splicing. If, after supplementary sampling, there are still areas where the preset minimum number of samplings has not been reached, an insufficient sampling flag is generated, and the written fragments are retained in the fragment sequence.
4. The civil aviation navigation data information processing method according to claim 1, characterized in that, The candidate number is compared with the first comparison benchmark and the candidate numbers of adjacent gates. Based on the comparison results, the binding status identifier between the fuselage number and the flight number in the flight data is determined, including: The complete fuselage image of the aircraft captured when the corresponding flight taxis in is extracted as the first comparison benchmark. The candidate number is compared segment by segment with the visible part of the number in the first comparison benchmark. Send a query request to the adjacent camera acquisition unit and receive the candidate numbers of the adjacent camera positions; When the candidate number matches the first comparison benchmark and is different from the candidate number of the adjacent station, the candidate number is determined to be valid, and a binding status flag indicating successful binding is written. When a candidate number is the same as or has a similarity greater than the preset similarity condition with an adjacent candidate number, the earliest captured fragment in the fragment sequence is backtracked for a second judgment. If the second judgment fails, the splicing is determined to be unsuccessful and the process is switched to the supplementary acquisition of the fragment sequence. A binding status identifier for the suspended segment is written.
5. The civil aviation navigation data information processing method according to claim 1, characterized in that, Candidate numbers are obtained by splicing the fragment sequence, including: The liquid film coverage status is detected frame by frame in the numbered area of the de-icing image to obtain the liquid film status identifiers of liquid film flow status, liquid film still status, or liquid film receding status. Frames for identification are selected from each frame in the order of liquid film receding state first, followed by liquid film stationary state. Frames in the liquid film flowing state are only used to provide position information in the residual sequence. When the characters identified at the same number location in the liquid film receding state frame and the liquid film stationary state frame are different, backtrack to query the dry image before the de-icing started, and use the identification result of the dry image to participate in the stitching. If no dry screen is available before de-icing, mark the characters in the corresponding areas as pending characters; The candidate number is obtained by concatenating the characters identified from each part.
6. The civil aviation navigation data information processing method according to claim 1, characterized in that, Obtain the data packet of the camera's onboard view transmitted from the fixed camera position, including: After receiving the camera image data packet, the shooting time and acquisition unit identifier in the camera image data packet are extracted, and the shooting sequence of each camera position is reconstructed according to the shooting time; Determine whether the time interval between the capture of adjacent data packets in the same acquisition unit exceeds the preset multiple range of the interval corresponding to the fixed frame rate; If the time limit is exceeded, check the acquisition unit identifier of all data packets that have arrived within the corresponding time period and return the screen belonging to the local unit to its original position. When images with the same identifier appear at the same shooting time but belonging to different camera positions, a first self-check request is sent to the corresponding acquisition unit. The image with the earlier shooting time in the response is used as the standard, and the reconstructed image is delivered as the camera image data packet to be verified to the entry verification.
7. A civil aviation navigation data information processing method according to claim 5, characterized in that, If no dry screen is available before de-icing, mark the characters in the corresponding areas as pending characters; The characters identified from each part are concatenated to obtain the candidate number, including: When the pending character exists, send a first reshoot request with the specified location of the machine serial number and the suggested shooting angle to the handheld shooting device of the on-duty machine inspection personnel; Receive the reshot footage and shooting location marker transmitted back by the handheld shooting device; The identification results of the numbers in the reshot images are compared character by character with the splicing results of the fragment sequence; When all matches, the pending characters are removed, and a determined candidate number is obtained; When there is a partial match, the fragment sequence is updated based on the reshot image and then reassembled to obtain the updated candidate number. If there is a complete mismatch, the liquid film state identifier and the order of shooting are checked back and the candidate number is rejected.
8. A method for processing civil aviation navigation data information according to claim 1, characterized in that, The process of comparing the candidate number with the first comparison benchmark and the candidate numbers of adjacent aircraft stands, and determining the binding status identifier between the fuselage number and the flight number in the flight data based on the comparison results, further includes: In response to insufficient data collection, conflicting images with the same identifier, or differences in re-enhanced images exceeding the preset difference limit, the abnormality types of insufficient data collection, questionable attribution, and contradictory identification are determined respectively, and the abnormality types are handled in the following order: questionable attribution, contradictory identification, and insufficient data collection. When the attribution is in doubt, freeze all the images of the corresponding machine position that have been entered into the database, send a manual assignment request to the apron duty terminal, rebind according to the response result and write a binding status mark of successful binding or binding invalidation. When identifying contradictory types, a second retake request containing the positions of the differing characters is sent to the handheld shooting device, and the binding status identifier is written according to the retake comparison result. When data collection is insufficient, a channel control query is sent to the apron dispatch terminal. After receiving a positive response, the data is collected centrally in the clear window and the binding status flag of successful binding is written. When a negative response is received, the process is switched to manual binding and the binding status flag of pending binding is written.
9. A method for processing civil aviation navigation data information according to claim 8, characterized in that, Also includes: Based on the successful binding status identifier, an image recording chain is established for the corresponding flight, and the stand view, de-icing view and around-the-aircraft inspection view are sequentially linked, with the binding status identifier written at the beginning of the chain. Before each de-icing screen or machine inspection screen is put into storage, the binding status identifier of the chain head is checked again; When the binding status indicator at the head of the chain is "binding successful", it will automatically attach; When the binding is suspended, it enters the suspension queue to wait; When binding is invalidated, the attachment is rejected and a first stop collection command is sent to the collection side; On flights with incorrect bindings, the liquid film status identifier, acquisition unit identifier, and shooting time of each frame are checked back along the image recording chain to locate the error level and generate a correction record that retains both the original incorrect binding content and the corrected content.
10. A civil aviation navigation data information processing system, employing a civil aviation navigation data information processing method as described in any one of claims 1 to 9, characterized in that, It includes a data packet acquisition module, an entry verification module, a fragment sequence generation module, and a number binding module; The data packet acquisition module is used to acquire the camera body image data packet transmitted back by the fixed camera position. The shooting time, camera position number and acquisition unit identifier are written into the camera body image data packet. The entry verification module is used to perform entry verification on the fuselage image data packet based on the shooting time, the camera position number and the acquisition unit identifier and the flight data, and obtain the entry verification result. The fragment sequence generation module is used to perform occlusion area detection on the camera body image that has passed the entry verification, and write the fragments in the numbered areas of each frame that are not occluded and the shooting time into the fragment sequence arranged in chronological order of shooting time. The number binding module is used to obtain candidate numbers by splicing the fragment sequence, compare the candidate numbers with the first comparison benchmark and the candidate numbers of adjacent aircraft positions, and determine the binding status identifier of the fuselage number and the flight number in the flight data based on the comparison results.
