Cultural heritage film discourse correlation mining and analysis system based on knowledge graph

Abstract

The application discloses a cultural heritage movie dialogue correlation mining and analysis system based on a knowledge graph, and relates to the technical field of cultural heritage digital analysis. The system is aimed at the problem that the follow-up object and the current main object in the local movie segment hit the local relationship, which easily causes the object to be unstable. The system first executes the prohibition of upgrading the main object for the candidate object that does not meet the independent confirmation condition, then generates an observation window according to the prequel of the current main object before the observation starting point, and compares the current main object regression evidence and the establishment of the independent confirmation condition of the candidate object in the observation window to output the current main object maintenance result or the main object switching result, and generate the cultural heritage object dialogue correlation result.

Description

Technical Field

[0001] This invention relates to the field of digital analysis technology for cultural heritage, and more specifically, to a knowledge graph-based system for mining and analyzing the associations of film and television discourse related to cultural heritage. Background Technology

[0002] Digital analysis technology for cultural heritage is frequently used in film and television content analysis. In this process, object clues are typically extracted from subtitles, narration, and visual content. These clues are then combined with object directories, alternative names, knowledge graphs, or rule bases to map, merge, and associate objects within the segment, thus generating object association results corresponding to the film and television segment. This type of technology is often used in scenarios where the subject of a story is identified from a local film or television clip, and where the attribution of subsequent shots and dialogue is continuously analyzed.

[0003] The existing technology has the following shortcomings: On the one hand, in cultural heritage film clips, after a relatively clear narrative center has been established around a particular cultural heritage object, subsequent shots or dialogue may still shift to the plaques, components, partial landscapes, or ancillary parts of that object. Such subsequent content may simply be a continuation of the aforementioned object, or it may gradually shift towards a new cultural heritage object, thus presenting two possible interpretations when determining object attribution. If existing techniques primarily update object attribution based on local information immediately following the appearance of subsequent content, it is easy to prematurely classify partial developments that should continue to belong to the aforementioned object as new object clues, or to continue incorporating subsequent content that already has an independent basis into the aforementioned object, thereby affecting the stability of object continuity within the clip.

[0004] On the other hand, another common approach in existing technologies is to observe subsequent content for a certain period before making an attribution judgment. However, this observation range often uses a fixed length or mainly relies on static evidence accumulation and empirical rules, making it difficult to adaptively adjust the subsequent observation range and the attribution judgment criteria by combining the strength of the connection established by the narrative center before the observation starting point. As a result, under different narrative intensities and different local insertion forms, existing processing results are prone to problems such as switching too early and merging too late, leading to unstable attribution judgment criteria and further affecting the consistency of cultural heritage film and television clip association analysis results. To overcome the above-mentioned defects of existing technologies, embodiments of the present invention provide a knowledge graph-based cultural heritage film and television discourse association mining and analysis system to solve the problems mentioned in the background technology. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: A knowledge graph-based system for mining and analyzing the associations of cultural heritage film and television discourse includes: The main object determination unit is used to obtain the object representation pointing to the cultural heritage object in the local film and television clip, perform knowledge graph mapping on the object representation, determine the current main object according to the order of appearance, and determine the remaining objects that have completed knowledge graph mapping as candidate objects; The local candidate suppression unit is used to place the candidate object into a prohibited primacy state and record its corresponding observation starting position when the candidate object and the current primacy object hit a pre-registered local relationship in the knowledge graph and the candidate object does not meet the independent confirmation condition. The observation window generation unit is used to determine the preceding length of the current main object before the observation starting point based on the observation starting point position, and to determine the preceding continuous verification segment based on the preceding length and the preset preceding basic verification length; within the preceding continuous verification segment, it counts the number of evidence hits for the overall name of the current main object and the number of evidence hits for the overall screen of the current main object, generates an observation length adjustment value according to the preset observation length adjustment rules, and generates the observation window length based on the preset basic observation length and the observation length adjustment value, and determines the end point of the observation window; The window comparison and judgment unit is used to compare the order in which the regression evidence of the current main object and the independent confirmation conditions of the candidate object are met within the observation window. The regression evidence of the current main object includes the regression evidence of the overall name of the current main object and the regression evidence of the overall screen of the current main object. When the regression evidence of the current main object is met first, the current main object is maintained. When the independent confirmation conditions of the candidate object are met first, the prohibition on becoming the main object is lifted, and the candidate object is switched to the new current main object. When the end of the observation window is reached and the independent confirmation conditions of the candidate object are not met first, the current main object is maintained. The association result generation unit is used to generate cultural heritage object discourse association results corresponding to local film and television clips based on the current main object maintenance result or the main object switching result.

[0006] In a preferred embodiment, the main object determination unit extracts object representations from the text and image content of the local video clip, wherein the object representations include at least name representations, pointer representations, and visual object representations.

[0007] In a preferred embodiment, the main object determination unit combines the standardized object name library, alias library, and type constraint rules to perform name normalization, alias normalization, and object-level matching on object expressions; it retains only object expressions that can form a unique and valid mapping in the cultural heritage knowledge graph; and it determines the object that completes the knowledge graph mapping first as the current main object according to the order of appearance, and determines the remaining objects that complete the knowledge graph mapping as candidate objects.

[0008] In a preferred embodiment, the local candidate suppression unit maps the current main object and candidate objects to the cultural heritage knowledge graph, and checks whether there is a local relationship between them according to the pre-registered local relationship table; the local relationship includes at least local composition relationship, subordinate component relationship, local landscape relationship and subordinate display relationship.

[0009] In a preferred embodiment, the local candidate suppression unit performs an independent confirmation condition check on candidate objects that hit local relationships to determine whether the candidate object has met the independent confirmation condition at the current screening position. The independent confirmation condition check includes at least checking whether the candidate object has a clear name expression that can uniquely map to the candidate object as a whole at the current screening position, and whether there are continuous narrative clues or continuous visual clues surrounding the candidate object after the clear name expression. When the candidate object does not meet the independent confirmation condition at the current screening position, the corresponding judgment position is recorded as the observation starting point position, and the candidate object is placed in a prohibited promotion state.

[0010] In a preferred embodiment, the observation window generation unit organizes the text content and screen content in the local video clip into a unified discrete judgment position sequence; for judgment positions that simultaneously contain both text content and screen content, they are merged and registered as the same judgment position according to a unified time progression order; for judgment positions that contain only text content or only screen content, they are separately registered as corresponding judgment positions according to their actual appearance order; both the observation window generation unit and the window comparison judgment unit perform position-by-position checks according to the unified discrete judgment position sequence.

[0011] In a preferred embodiment, the observation window generation unit traces backward from the previous judgment position before the observation starting point, performing a succession check on each of the preceding consecutive judgment positions before the observation starting point. When a certain preceding consecutive judgment position has a clear name expression that can uniquely map to the current main object as a whole, or has a continuous screen clue that can stably point back to the current main object as a whole, the succession is recorded as successful and the backtracking continues. When a certain preceding consecutive judgment position has neither formed a name-side succession nor a screen-side succession, the backtracking stops. Among them, weak references, local component names, close-up views of local components, and local screens that cannot uniquely point back to the current main object as a whole are not counted as successful succession.

[0012] In a preferred embodiment, the observation window generation unit compares the preceding continuity length with a preset continuous continuity lower limit, compares the number of evidence hits for the current main object's overall name with a preset name continuity lower limit, and compares the number of evidence hits for the current main object's overall screen with a preset screen continuity lower limit. When at least two of the three comparison results reach the corresponding lower limit, a positive observation length adjustment value is generated; when at least two of the three comparison results do not reach the corresponding lower limit, a negative observation length adjustment value is generated; in other cases, the preset basic observation length remains unchanged; and the observation window length is generated based on the preset basic observation length and the observation length adjustment value to determine the end point of the observation window.

[0013] In a preferred embodiment, the window comparison and determination unit checks the independent confirmation conditions of candidate objects in the observation window in the order of self-confirmation followed by expansion and continuation. When a certain continuous determination position has a clear name expression that can be uniquely mapped to the whole candidate object, the self-confirmation is recorded as successful. After the self-confirmation is successful, the unit continues to check whether the subsequent continuous determination positions form a continuous narrative or continuous screen expansion around the candidate object. When the self-confirmation is successful and expansion and continuation are subsequently obtained, the independent confirmation condition of the candidate object is recorded as successful.

[0014] In a preferred embodiment, the window comparison and determination unit records the first established position of the regression evidence of the current master object and the first established position of the independent confirmation condition of the candidate object, and compares their order. When the first established position of the regression evidence of the current master object is earlier, the current master object is maintained. When the first established position of the independent confirmation condition of the candidate object is earlier, the prohibition on becoming the master object is lifted, and the candidate object is switched to the new current master object. When both are established at the same consecutive determination position, or when the independent confirmation condition of the candidate object is not established before the end of the observation window, the current master object is maintained.

[0015] The effects and advantages of the knowledge graph-based cultural heritage film and television discourse association mining and analysis system provided by this invention are as follows: This invention provides a knowledge graph-based system for mining and analyzing the association of cultural heritage film and television discourse. Addressing the issue of object continuation control in local film and television clips, it prevents subsequent objects from immediately seizing control of the main narrative when a candidate object hits a local relationship but hasn't yet met the independent confirmation criteria. This ensures a more stable entry point for local expanded content and independent object content when entering the judgment chain. Furthermore, this invention transforms the preceding continuation length, overall name continuation evidence, and overall scene continuation evidence of the current main object before the observation starting point into the basis for generating the observation window. This allows the subsequent comparison range to be adjusted according to the strength of the preceding continuation, avoiding premature switching or delayed merging under different continuation intensities and local insertion forms under fixed-length observation. Moreover, this invention compares the order in which the regression evidence of the current main object and the independent confirmation criteria of candidate objects are met within the observation window, rather than relying on static accumulated evidence or empirical total scores. This compresses the maintenance and switching of the main narrative into a unified comparison criterion, improving the consistency and stability of the discourse association results for local film and television clips. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a flowchart of the dynamic adjustment of the observation window length based on prior stability. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides a knowledge graph-based system for mining and analyzing the discourse associations of cultural heritage film and television clips. The analysis of cultural heritage film and television clips is commonly used to continuously track subsequent shots and discourse content around a cultural heritage object that has formed a narrative center within a local film or television clip, and outputs the corresponding discourse association results of the cultural heritage object after the clip ends. Existing solutions typically perform object succession judgments based on subsequently appearing mappable objects, and use fixed-length delayed observation or subsequent evidence accumulation to correct and maintain or switch the current main object. The default premise is that once subsequent content gains visible support, it can be relatively stably interpreted as a local expansion of the current main object, or as a new independent narrative object.

[0019] However, in scenarios where local relationships interfere, subsequent shots or dialogue often shift to the plaques, components, local landscapes, or ancillary parts of the current main object. This results in subsequent content having two possible interpretations after object mapping: either a local expansion or an independent object. In such cases, without entry control for candidate objects of local relationships, the generation of observation ranges, and a unified comparison standard, the system may prematurely switch the expanded local content to a new narrative thread, or continue to merge subsequent objects that already have an independent basis into the current main object. This leads to unstable object continuity and shifts in the discourse association results of local film and television segments.

[0020] To this end, this invention uses the local relationship hit result between the current main object and candidate objects as the entry point. First, it performs a process to prevent candidate objects from becoming the main object and registers the observation starting point position. Then, based on the preceding inheritance length of the current main object before the observation starting point position, and the overall name inheritance evidence and overall image inheritance evidence within the preceding continuous verification segment, it generates the observation window length and observation window. Next, within the generated observation window, it compares the order in which the regression evidence of the current main object and the independent confirmation conditions of the candidate objects are met, outputting either the current main object maintenance result or the main object switching result, and further generating the discourse association result of the cultural heritage object corresponding to the local film and television clip.

[0021] Based on the above design, this invention constructs a knowledge graph-based cultural heritage film and television discourse association mining and analysis system, which consists of a main object determination unit, a local candidate suppression unit, an observation window generation unit, a window comparison and judgment unit, and an association result generation unit, as follows: The main object determination unit, located at the starting point of the system link, is used to extract object representations pointing to cultural heritage objects from local video clips and determine the current main object, candidate objects, and the positions where candidate objects appear. The main object determination unit reads the local video clip X101 and the cultural heritage knowledge graph X102, and performs object representation extraction and knowledge graph mapping by combining the standardized object name library, alias library, and type constraint rules, outputting the object entry result R101 for the local candidate suppression unit to read. R101 is used to hold the current main object, the set of candidate objects, and the positions where candidate objects appear, and serves as the unified object entry point for subsequent local relation matching and entry control path judgment.

[0022] The local candidate suppression unit, located after the main object determination unit, is used to screen the local relationships between the current main object and candidate objects, and to import candidate objects that do not yet have an independent basis for establishment into the prohibition of promotion to main object control path. The local candidate suppression unit reads the object entry result R101 and the cultural heritage knowledge graph X102, and, combined with the pre-registered local relationship table and the initial rules for independent confirmation conditions, performs local relationship hit judgment, prohibition of promotion to main object registration, and observation start point position registration on the candidate objects, outputting the object inheritance control record C102 for the observation window generation unit to read. Specifically, R101 provides the current main object, the set of candidate objects, and the location of candidate object occurrences; C102 inherits the candidate objects entering the prohibition of promotion to main object state, the candidate object suppression flag, the observation start point position, and the object inheritance control entry item, and serves as the unified control chain entry for subsequent observation window generation and window comparison judgment.

[0023] The observation window generation unit is used to generate an observation window around the observation starting point position, taking into account the preceding context of the current main object before that position, for subsequent comparison. The observation window generation unit reads the object entry result R101, the object connection control record C102, and the local video clip X101, generates the observation window generation result R103, and writes the preceding connection quantification result, preceding evidence statistics result, and observation window boundary result into C102 for the window comparison and judgment unit to read. Specifically, R101 provides the object-level information required by the current main object and its control entry point; C102 carries the observation starting point position and the preceding connection quantification result, preceding evidence statistics result, and observation window boundary result written back by this unit; and R103 carries the preceding connection length, preceding continuous verification segment, overall name connection evidence statistics result, overall image connection evidence statistics result, observation length adjustment value, and observation window.

[0024] The window comparison and determination unit is used to detect the regression evidence of the current main object and the independent confirmation conditions of the candidate objects within the observation window, compare the order of their establishment, and update the current main object. The window comparison and determination unit reads the observation window generation result R103, the object inheritance control record C102, and the local video clip X101, outputs the window comparison and determination result R104, and writes back the main object regression evidence status, the candidate object independent confirmation status, the order of establishment comparison result, and the updated current main object to C102 for the association result generation unit to read. Specifically, R103 provides the observation window, the observation window endpoint, and the comparison range; C102 carries the comparison status and update result written back by this unit; and R104 carries the main object regression evidence status, the candidate object independent confirmation status, the order of establishment comparison result, and the updated current main object.

[0025] The association result generation unit, located after the window comparison and determination unit, is used to determine the final associated object of the current local film and television segment and generate the corresponding cultural heritage object discourse association result. The association result generation unit reads the window comparison and determination result R104 and the object inheritance control record C102, generates the cultural heritage object discourse association result R105 based on the establishment order comparison result and the updated current main object, and writes the result as the final output of the current local film and television segment. Specifically, R104 provides the establishment order comparison result, the updated current main object, and the main result status; C102 carries the control chain status after window processing; and R105 inherits the final main object and its discourse association result corresponding to the current local film and television segment.

[0026] The implementation process and operating effects of the system of the present invention will be described in detail below with reference to specific embodiments. It should be understood that the embodiments are only used to illustrate the technical solution of the present invention, and not to limit it. Without changing the essence of the invention, the relevant steps, parameters and module divisions can be appropriately adjusted.

[0027] In this specification, the current primary object refers to the cultural heritage object that serves as the subject of the current narrative thread within the current local video clip. Candidate objects refer to other cultural heritage objects that have completed knowledge graph mapping within the same local video clip but have not yet been identified as the current primary object. The observation starting point position refers to the judgment position corresponding to when a candidate object hits a local relationship but does not meet the independent confirmation entry requirements and is registered as prohibited from becoming the primary object. Evidence of the current primary object's regression refers to name expressions or visual clues that can directly point back to the overall current primary object. The discourse association result of cultural heritage objects refers to the object attribution result formed between the local video clip and the final primary object.

[0028] In an optional implementation, the main object determination unit's implementation process includes fragment reading and object representation extraction, object representation normalization and effective mapping screening, current main object determination, candidate object retention and candidate object occurrence location registration, and object entry result encapsulation and writing.

[0029] In the fragment reading and object representation extraction process, this unit first reads the text and visual content of a local film and television fragment X101, and then extracts object representations based on the cultural heritage-related information within it. In some implementations, object representations include at least name representations, indicative representations, and visual object representations. Name representations are used to receive object names that appear directly in the text content; indicative representations are used to receive alternative names or further references to the aforementioned objects in the text content; and visual object representations are used to receive identifiable object clues in the visual content that point to cultural heritage objects. This forms a set of object representations, which serves as input for subsequent knowledge graph mapping.

[0030] In the object representation normalization and effective mapping selection, this unit performs name normalization, alias normalization, and object-level matching on each object representation in the object representation set, and performs knowledge graph mapping in conjunction with the cultural heritage knowledge graph X102. In some implementations, a standardized object name library is used to provide standard object names, an alias library is used to provide the correspondence between object aliases and standardized object names, and type constraint rules are used to restrict the mapping results to the scope of cultural heritage objects. If an object representation can form a unique effective mapping in the knowledge graph, it is retained as a valid mapping object. If an object representation cannot form a valid mapping, or forms multiple conflicting mapping results, it is not written into the subsequent object entry results. This forms a set of valid mapping objects.

[0031] In determining the current master object, this unit identifies the current master object based on the order in which valid mapped objects appear in the local video clip. In some implementations, determining the current master object by appearance order means identifying the object that first completes a valid knowledge graph mapping in the local video clip as the current master object. This step only establishes the entry points for the current master object and candidate objects; it does not change the attribution of the current master object within this step.

[0032] In the process of candidate object retention and candidate object occurrence position registration, this unit retains all valid mapped objects other than the current master object as candidate objects and registers the occurrence position of each candidate object in the local video clip. In some implementations, the candidate object occurrence position is used to characterize the determination position where the candidate object first completes a valid mapping of the knowledge graph. This position only serves as the upstream position basis for subsequent local candidate suppression processing and is not directly used as the observation starting point position. This results in the candidate object set and the candidate object occurrence position result.

[0033] In the object entry result encapsulation and writing process, this unit uniformly writes the current main object, the candidate object set, and the candidate object occurrence positions into the object entry result R101. In some implementations, R101 organizes the object entry content according to local video clip identifiers, enabling subsequent local candidate suppression units to read the current main object, candidate object set, and candidate object occurrence positions one by one within the same segment range. After writing is complete, R101 is written out as a whole for the local candidate suppression unit to perform local relationship matching and entry control path judgment.

[0034] In one optional implementation, the local candidate suppression unit includes reading the entry object and expanding the candidate object, judging the local relationship hit, initial screening based on independent establishment, registering the prohibition of ascending to master status, generating the observation starting position, and encapsulating and writing the object inheritance control record.

[0035] In the entry object reading and candidate object expansion, this unit first reads the current main object, the candidate object set, and the candidate object occurrence position from the object entry result R101, and uses the current main object as the basis for the main object of this round of candidate screening. In some implementations, the candidate object set only includes objects that have completed effective mapping of the knowledge graph and have not been determined as the current main object, and the candidate object occurrence position is used to characterize the determination position where the candidate object first completes effective mapping of the knowledge graph. This establishes the object basis and position basis required for candidate-by-candidate screening.

[0036] In the local relationship matching process, this unit maps the current main object and candidate objects one by one to the cultural heritage knowledge graph X102, and checks whether there is a local relationship between them based on the pre-registered local relationship table. A local relationship refers to an object relationship pre-registered in the cultural heritage knowledge graph that represents a local composition, subordinate component, local landscape, or subordinate display relationship between the candidate object and the current main object. In some implementations, if the candidate object and the current main object match the above-mentioned local relationship in the knowledge graph, then the candidate object is considered to have met the local relationship matching condition. If no local relationship is matched, the candidate object is not imported into the prohibited promotion control path.

[0037] In the initial screening of independent confirmation conditions, this unit performs an independent confirmation condition entry check on candidate objects that meet the local relation hit condition. The independent confirmation condition entry check is used to determine whether a candidate object already clearly possesses an independent object pointer and a basis for subsequent expansion when entering the prohibited primacy processing stage. In some implementations, the initial independent confirmation condition rule is used to check whether the candidate object currently has a clear object name expression, whether it has formed an independent narrative starting point, and whether it already has initial clues for further expansion. If the candidate object still does not meet the independent confirmation condition entry requirements at the current screening position, then the candidate object is recorded as meeting the conditions for entering the prohibited primacy processing stage. If the candidate object already clearly meets the independent confirmation condition entry requirements, then the prohibited primacy state is not applied to it in this step. The independent confirmation condition entry check in this step is only used to determine whether the candidate object has not yet met the entry requirements for independent confirmation conditions when entering the prohibited primacy control path. Whether the candidate object ultimately meets the independent confirmation conditions is further determined by the window comparison and judgment unit within the observation window.

[0038] In the registration of prohibited master status, this unit performs prohibited master status writing for candidate objects that do not meet both the local relation hit condition and the independent confirmation condition entry requirement. For candidate objects that enter the prohibited master status process, a candidate object suppression flag is written, and the candidate object is registered as a candidate object entering the prohibited master status. In some implementations, the candidate object suppression flag is used to indicate that the candidate object does not directly participate in the current master object switching judgment before entering the subsequent observation window processing, but instead enters the establishment order comparison process in the subsequent observation window first.

[0039] In the observation start position generation process, this unit registers the observation start position for candidate objects that enter the prohibited promotion state. The observation start position refers to the determination position where a candidate object first satisfies the local relation hit condition and the independent confirmation condition entry requirement is not met, and is written into the prohibited promotion state. In some implementations, the observation start position is directly generated from the determination position where the candidate object enters the prohibited promotion state in this step, rather than being directly equivalent to the candidate object's appearance position.

[0040] In the object succession control record encapsulation and writing process, this unit uniformly writes the candidate objects entering the prohibited master state, the candidate object suppression flag, the observation start position, and the object succession control entry item into the object succession control record C102. In some implementations, C102 organizes the control chain content according to the current master object, candidate objects, and observation start position, enabling subsequent observation window generation units to read the candidate object suppression state and observation start position along the same control chain. After writing is completed, C102 is written out as a whole, allowing the observation window generation unit to continue executing the preceding succession backtracking and observation window generation.

[0041] In one optional implementation, the observation window generation unit includes entry reading and reverse starting point determination, unified sequence benchmark organization, preceding continuous inheritance and backtracking, preceding continuous verification and dual evidence statistics, observation length adjustment and observation window generation, and result writing back and output.

[0042] In the entry point reading and reverse start point determination, this unit first reads the observation start point position, candidate object suppression status, and existing entry field from the object acceptance control record C102. Then, combined with the current main object information in the object entry result R101, it determines the object generated in this round of window generation. Subsequently, using the previous judgment position before the observation start point position as the reverse backtracking start position, the starting entry point for subsequent preceding acceptance checks is established.

[0043] In the unified sequence benchmark organization, this unit organizes the text content and image content in the local video clip X101 into a unified discrete decision position sequence, so that the preceding backtracking and subsequent window comparisons share the same sequence caliber. In some implementations, the unified discrete decision position sequence uses the sequential advancement unit in the local video clip as the basic organizational granularity. For decision positions that contain both text content and image content, the text expression appearance position and the corresponding image display position are merged and registered as the same decision position according to the unified time advancement order. For decision positions that contain only text content or only image content, they are registered separately as corresponding decision positions according to their actual appearance order in the local video clip. Both the observation window generation unit and the window comparison decision unit perform position-by-position checks according to this unified discrete decision position sequence, thereby ensuring that the preceding continuation length generation and the comparison of the establishment order within the window use the same position granularity.

[0044] In the preceding continuous inheritance backtracking, this unit starts from the local video clip X101 and the current main object information, and performs inheritance checks on each preceding continuous judgment position before the observation starting position according to the current main object inheritance judgment rules. If a preceding continuous judgment position has a clear name expression that can uniquely map to the current main object as a whole, or has a continuous screen clue that can stably point back to the current main object as a whole, then the inheritance at that position is recorded as successful, and the backtracking continues. In some implementations, name-side inheritance clues include the standard name, stable alias, and name combination expressions that clearly point to the current main object as a whole, while screen-side inheritance clues include the overall outline, overall structure display, and continuous screen clues that still serve the overall display of the current main object. If a position only has a weak representation, a local component name, a close-up of a local component, or a local screen that cannot uniquely point back to the current main object as a whole, then it is recorded as weak inheritance and is not counted as successful inheritance. If a position fails to establish a connection on either the name side or the screen side, that position is recorded as the first invalid position, and the backtracking process stops. After the backtracking is completed, the nearest continuous segment between the first invalid position and the previous judgment position before the observation starting point is determined as the preceding continuous connection segment, and the preceding connection length is generated based on the start and end boundaries of this segment.

[0045] In the preceding continuous verification and dual evidence statistics, this unit generates a preceding continuous verification segment based on the preceding inheritance length and the preset preceding basic verification length, and counts the name-side inheritance evidence and the image-side inheritance evidence within this verification segment. In some implementations, name-side inheritance evidence is formed by extracting name class expressions that can be directly mapped to the current main object as a whole, and image-side inheritance evidence is formed by extracting image clues that can point to the current main object as a whole. Based on the above two types of evidence, the overall number of name inheritance evidence hits and the overall number of image inheritance evidence hits are counted to obtain the preceding evidence statistics results. This result is used to characterize the preceding stability of the current main object before the observation starting position, and only serves as the input basis for the generation of the observation window, without rewriting it as the main object regression evidence in the window comparison and judgment stage.

[0046] The process for dynamically adjusting the observation window length based on the preceding succession length and the statistical results of preceding evidence can be found in [link to documentation]. Figure 2 .

[0047] In the observation length adjustment and observation window generation, such as Figure 2 As shown, this unit performs growth, contraction, or maintenance processing on the preset basic observation length based on the preceding acceptance length, the statistical results of overall name acceptance evidence, and the statistical results of overall image acceptance evidence. The preset basic observation length represents the default observation range length pre-set to complete the subsequent observation of a candidate object without considering factors that enhance or weaken preceding stability. In some implementations, the preset continuous acceptance lower limit, the preset name acceptance lower limit, and the preset image acceptance lower limit are all provided by preset system parameters and configured in conjunction with the granularity of the unified discrete judgment position sequence. Whether the preceding acceptance length reaches the preset continuous acceptance lower limit is used to determine whether the preceding acceptance length shows enhancement or weakening. Whether the number of overall name acceptance evidence hits reaches the preset name acceptance lower limit is used to determine whether the overall name acceptance evidence shows enhancement or weakening. Whether the number of overall image acceptance evidence hits reaches the preset image acceptance lower limit is used to determine whether the overall image acceptance evidence shows enhancement or weakening. When at least two of the three factors show enhancement, a positive observation length adjustment value is generated. A negative observation length adjustment value is generated when at least two of the three factors show a decrease. The preset base observation length remains unchanged when neither the growth nor contraction trigger conditions are met. In cases where the evidence supporting the overall name and the evidence supporting the overall image are inconsistent, a conservative approach is prioritized, triggering growth only when the preceding length simultaneously shows an increase; otherwise, it remains unchanged. After obtaining the observation length adjustment value, the observation window length is generated based on the preset base observation length and the observation length adjustment value.

[0048] In the observation window boundary correction and observation window formation process, this unit performs boundary correction on the generated observation window length and further determines the observation window endpoint. In some implementations, both the minimum and maximum window lengths are provided by system preset parameters and configured according to the granularity of a unified discrete judgment position sequence. The minimum window length is used to ensure that a candidate object has at least one complete subsequent observation range after entering the prohibited ascending path, so as to complete the continuous check of subsequent changes of the candidate object. The maximum window length is used to limit the observation range to remain within the object acceptance judgment range corresponding to the current local video clip, avoiding the observation range from exceeding the local acceptance judgment boundary. If the generated observation window length is less than the minimum window length, the final observation window length is corrected to the minimum window length. If the generated observation window length is greater than the maximum window length, the final observation window length is corrected to the maximum window length. After the length correction is completed, the observation window endpoint is determined with the observation start position as the starting point, and the observation window is formed accordingly.

[0049] In the result write-back and output process, this unit uniformly writes the preceding continuous acceptance segment, preceding acceptance length, preceding continuous verification segment, overall name acceptance evidence statistics, overall screen acceptance evidence statistics, observation length adjustment value, observation window length, observation window endpoint, and observation window into the observation window generation result R103, and simultaneously writes the preceding acceptance quantification field, preceding evidence statistics field, and observation window boundary field into the object acceptance control record C102. In some implementations, R103 organizes the window generation result according to the current main object, candidate object, and observation start point position, and C102 continuously carries the entry state, preceding quantification result, and window boundary result according to the same control chain. After the write-back is completed, R103 and the updated C102 are output to the window comparison and judgment unit for it to perform a succession comparison within the predetermined observation window range.

[0050] In an optional implementation, the window comparison determination unit includes reading the comparison range and expanding the position within the window, detecting the main object regression evidence, detecting the independent confirmation conditions of the candidate object, comparing the order of establishment, updating the current main object, and writing back and outputting the results.

[0051] In the comparison range reading and window position expansion, this unit first reads the observation window and its endpoint from the observation window generation result R103, and then reads the current master object, the candidate object entering the prohibited master state, the observation start position, and the prohibited master state from the object inheritance control record C102. Subsequently, it expands the continuous decision positions within the observation window along the unified discrete decision position sequence, and determines each continuous decision position between the observation start position and the observation window endpoint as the range of continuous decision positions within the current round of the observation window.

[0052] In the main object regression evidence detection, this unit starts from the local video clip X101 and the current main object information, and checks each continuous judgment position within the continuous judgment position range inside the observation window to see if the current main object regression evidence is valid. In some implementations, the current main object regression evidence includes the current main object overall name regression evidence and the current main object overall screen regression evidence. If a certain continuous judgment position has a clear name expression that can uniquely map to the current main object as a whole, then the overall name regression evidence is recorded as valid, and this position is recorded as the first valid position of overall name regression. If a certain continuous judgment position has a continuous screen clue that can stably point back to the current main object as a whole, then the overall screen regression evidence is recorded as valid, and this position is recorded as the first valid position of overall screen regression. If at least one of the overall name regression evidence and the overall screen regression evidence is valid, then the main object regression evidence is recorded as valid, and the earlier valid position is determined as the first valid position of main object regression.

[0053] In the independent confirmation condition detection of candidate objects, this unit starts from the local video clip X101, the candidate object that has entered the prohibited ascending state, and the continuous judgment position range within the observation window, and checks whether the independent confirmation condition of the candidate object is met position by position in the order of first self-referential confirmation and then expansion and continuation. In some implementations, self-referential confirmation means that a certain continuous judgment position has a clear name expression that can uniquely map to the whole candidate object, and expansion and continuation means that after the self-referential confirmation is established, the subsequent continuous judgment positions continue to form a continuous narrative or continuous screen development around the candidate object. If a certain continuous judgment position has a candidate object's standard name, stable alias, or name expression that clearly points to the whole candidate object, then the candidate object self-referential confirmation is recorded as established, and this position is recorded as the first established position of self-referential confirmation. After the self-referential confirmation is established, it continues to check whether there are expansion clues around the candidate object's attributes, position, composition, use, or continuous screen display in the subsequent continuous judgment positions. If the above expansion clues appear, then the candidate object expansion and continuation is recorded as established, and this position is recorded as the first established position of expansion and continuation. If a candidate object confirms itself and subsequently acquires expansion and continuation, then the independent confirmation condition of the candidate object is recorded as being met, and the position where the expansion and continuation first becomes met is determined as the position where the independent confirmation of the candidate object first becomes met.

[0054] In the order of establishment comparison, this unit continuously compares the initial establishment position of the primary object regression and the initial establishment position of the candidate object independent confirmation. If the primary object regression evidence is established first, the order of establishment comparison result is recorded as the maintenance result, and the current primary object remains unchanged. If the candidate object independent confirmation condition is established first, the order of establishment comparison result is recorded as the switching result, and the candidate object is determined as the new current primary object. In some implementations, if both are established concurrently at the same continuous determination position, the current primary object is maintained first, and the order of establishment comparison result is recorded as the maintenance result. If, when the observation window ends, the candidate object independent confirmation condition is still not established before the primary object regression evidence, the order of establishment comparison result is recorded as the maintenance result. The concurrent establishment processing result is a special formation result of the order of establishment comparison result and is written together with the maintenance result into the window comparison determination result R104.

[0055] During the update of the current master object, this unit updates the state of the current master object based on the result of the order of establishment comparison. If the result of the order of establishment comparison is to maintain, the original current master object is maintained, and the prohibition of promotion processing for candidate objects is retained. If the result of the order of establishment comparison is to switch, the prohibition of promotion for candidate objects is lifted, and the candidate object is updated to the new current master object. If the result of concurrent establishment processing or the processing result after reaching the end of the observation window points to the maintain result, the original current master object is maintained.

[0056] In the result write-back and output process, this unit uniformly writes the following data into the window comparison judgment result R104: overall name regression evidence establishment status, overall name regression first establishment position, overall image regression evidence establishment status, overall image regression first establishment position, main object regression evidence establishment status, main object regression first establishment position, candidate object self-referential confirmation status, candidate object self-referential confirmation first establishment position, candidate object expansion and inheritance status, candidate object expansion and inheritance first establishment position, candidate object independent confirmation condition establishment status, candidate object independent confirmation first establishment position, establishment order comparison result, concurrent establishment processing result, and updated current main object and main line result status. Simultaneously, it writes the object inheritance control record C102. In some implementations, R104 organizes the internal comparison results of the window according to the current main object, candidate object, observation start position, and observation window end point. C102 continuously carries the comparison status and updated main line attribution status according to the same control chain. After completing the write-back, R104 and the updated C102 are output as a whole to the association result generation unit for generating the cultural heritage object discourse association results corresponding to the local film and television clips.

[0057] In an optional implementation, the process of the association result generation unit includes result reading and status verification, determination of the current master object ownership, encapsulation of association results, and result writing back and output.

[0058] In the result reading and status verification, this unit first reads the establishment order comparison result, the updated current main object, and the main line result status from the window comparison judgment result R104. Then, it reads the control chain status after window processing and the current record status from the object inheritance control record C102. Subsequently, a consistency check is performed on the main line attribution information corresponding to the current local video clip in R104 and C102, forming the result input basis required for the generation of this round of association results. In some implementations, if there is a difference between the updated current main object in R104 and the control chain status in C102, the updated current main object driven by the establishment order comparison result is used as the basis for the generation of subsequent association results.

[0059] In determining the current master object, this unit determines the final master object of the current local video segment based on the result of the order of establishment comparison. If the order of establishment comparison result is a maintenance result, the current master object before the window comparison is retained as the final master object. If the order of establishment comparison result is a switching result, the updated current master object is determined as the final master object. If the order of establishment comparison result corresponds to the maintenance processing result after concurrent establishment, or corresponds to the maintenance processing result after reaching the end of the observation window, the original current master object is continued to be determined as the final master object.

[0060] In the encapsulation of association results, this unit generates a discourse association result R105 for cultural heritage objects based on the final main object attribution result and the main result status. In some implementations, R105 includes at least the current local video clip identifier, the final main object, the main result status, and the result output status. If the final main object attribution result corresponds to a maintenance result, a discourse association is established between the current local video clip and the maintained current main object. If the final main object attribution result corresponds to a switching result, a discourse association is established between the current local video clip and the switched new current main object.

[0061] In the write-back and output process, this unit writes the current local video clip identifier, the final main object, the main result status, and the result output status into the cultural heritage object discourse association result R105, and simultaneously writes the final main object attribution result and the result output status into the object inheritance control record C102. In some implementations, R105 organizes the result content according to the current local video clip identifier and the final main object, while C102 continuously carries the final main object attribution result and the result output status according to the same control chain. After the write-back is completed, R105 is written out as the final system output result corresponding to the current local video clip.

[0062] In a specific embodiment, taking the object acceptance determination scene in a certain ancient building narration segment as an example, the linkage processing process of this system is explained. The first part of this partial video segment continuously revolves around the entire Hall of Supreme Harmony, and the latter part of the shot and narration briefly shifts to the plaque of the Hall of Supreme Harmony. For ease of explanation, the continuous processing positions in this segment are sequentially denoted as T1 to T8. Among them, T1 is the position where the overall name of the Hall of Supreme Harmony appears, T2 is the position where the overall picture of the Hall of Supreme Harmony appears, T3 is the position where the overall name of the Hall of Supreme Harmony appears again, T4 is the position where the overall picture of the Hall of Supreme Harmony appears again, T5 is the position where the plaque of the Hall of Supreme Harmony is first mapped to the knowledge graph, T6 is the position where the close-up of the plaque continues to appear, T7 is the position where the overall name of the Hall of Supreme Harmony returns to the narration content, and T8 is the position where this round of comparison ends. As can be seen from T1 to T4, before the plaque appeared, the narrative of the fragment had been continuously advancing around the Hall of Supreme Harmony. Therefore, in this case, T1 to T4 is taken as the preceding section before the starting point of observation, and based on this, the length of the preceding section is determined to be 4, the number of evidences for the overall name is 2, and the number of evidences for the overall picture is 2.

[0063] In this example, the main object determination unit first performs object representation extraction and knowledge graph mapping on the text and image content of the segment, and determines the Hall of Supreme Harmony as the current main object in the object entry result R101, and the plaque of the Hall of Supreme Harmony as a candidate object. Subsequently, the local candidate suppression unit detects that the plaque of the Hall of Supreme Harmony and the Hall of Supreme Harmony have a local compositional relationship in the knowledge graph, and at the same time confirms that the plaque at position T5 is only a local object that has been brought in, and has not yet formed a new narrative center that can be independently developed. Therefore, the plaque is registered as prohibited from becoming the main object in the object inheritance control record C102, and T5 is determined as the observation starting point position. The observation window generation unit then combines the preceding inheritance length and the preceding dual evidence statistical results to generate an observation window covering T5 to T8 in the observation window generation result R103. That is to say, the system does not immediately switch the main object when the plaque just appears at T5, but continues to look at the short segment of content after T5, and then decides whether this transition is just the insertion of local content of the Hall of Supreme Harmony, or has formed a new independent main object.

[0064] The window comparison judgment unit then performs a precedence comparison within the observation windows T5 to T8. Specifically, at position T5, the object expression of the plaque of the Hall of Supreme Harmony appears, and at position T6, a close-up view of the plaque continues. However, both positions only offer short-term displays of the plaque itself; the plaque's name is not explicitly mentioned again, nor is there any subsequent continuous explanation surrounding the plaque. Therefore, no independent confirmation condition for the plaque is formed within these windows. Conversely, at position T7, the narration returns to the Hall of Supreme Harmony as a whole, with a name expression that directly points to the Hall of Supreme Harmony as a whole. Therefore, the system detects the current main object regression evidence as established at position T7. Since the overall regression of the Hall of Supreme Harmony precedes the independent confirmation of the plaque, the window comparison judgment result R104 is recorded as a maintenance result, and the current main object remains unchanged as the Hall of Supreme Harmony. The association result generation unit finally generates the cultural heritage object discourse association result R105 based on R104, establishing a final discourse association between this partial film clip and the Hall of Supreme Harmony, and recording the main line status as main line maintenance.

[0065] Therefore, in this specific scenario, although a plaque object capable of completing the knowledge graph mapping has appeared in the fragment, because this object has a local relationship with the Hall of Supreme Harmony, and no independent confirmation condition has been formed within the observation window, while the overall name of the Hall of Supreme Harmony has returned to the main narrative line, the system ultimately does not confirm the plaque as the new main object, but continues to maintain the Hall of Supreme Harmony as the final main object of this fragment. This example illustrates that this system does not immediately switch the object's main line just because a local component appears briefly, but first distinguishes between the local insertion scene and the actual main line switching scene by prohibiting promotion to main line and subsequent observation.

[0066] In another specific scenario, the first segment focuses on a main garden, while the second segment shifts to an auxiliary building within that garden. Upon its initial appearance, this auxiliary building, due to its local connection to the current main object, initially enters a state where it is prohibited from becoming the main object. The system also initially generates an observation window around the location where it enters this state. However, within this observation window, subsequent narration explicitly identifies the auxiliary building and continues to describe its location, purpose, and composition. The visual content also consistently revolves around this auxiliary building. Therefore, the system determines that the auxiliary building has already completed self-identification and gained continuity, and its independent identification condition precedes the evidence of its return to the original main object. Consequently, the system removes the prohibition on the auxiliary building from becoming the main object and switches it to the new main object, ultimately generating the corresponding cultural heritage object discourse association result. This scenario illustrates that the system does not indiscriminately suppress candidate objects from becoming the main object; as long as the candidate object truly forms an independent narrative thread in subsequent segments, the system can still switch the main object to the new cultural heritage object.

[0067] The above specific scenario embodiments are only used to illustrate the linkage processing process of this system in the determination of the acceptance of local film and television clip objects, and do not constitute a limitation on the type of cultural heritage object, the type of local relationship, the number of sequential positions, or the specific content of the clip.

[0068] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0069] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0070] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0071] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0072] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A knowledge graph-based system for mining and analyzing the associations of cultural heritage film and television discourse, characterized in that, include: The main object determination unit is used to obtain the object representation pointing to the cultural heritage object in the local film and television clip, perform knowledge graph mapping on the object representation, determine the current main object according to the order of appearance, and determine the remaining objects that have completed knowledge graph mapping as candidate objects; The local candidate suppression unit is used to put the candidate object into a prohibited primacy state and record its corresponding observation starting position when the candidate object and the current primacy object hit a pre-registered local relationship in the knowledge graph and the candidate object does not meet the independent confirmation condition. The observation window generation unit is used to determine the preceding length of the current main object before the observation starting point based on the observation starting point position, and to determine the preceding continuous verification segment based on the preceding length and the preset preceding basic verification length; within the preceding continuous verification segment, it counts the number of evidence hits for the overall name of the current main object and the number of evidence hits for the overall screen of the current main object, generates an observation length adjustment value according to the preset observation length adjustment rules, and generates the observation window length based on the preset basic observation length and the observation length adjustment value, and determines the end point of the observation window; The window comparison and judgment unit is used to compare the order in which the regression evidence of the current main object and the independent confirmation conditions of the candidate object are met within the observation window. The regression evidence of the current main object includes the regression evidence of the overall name of the current main object and the regression evidence of the overall screen of the current main object. When the regression evidence of the current main object is met first, the current main object is maintained. When the independent confirmation conditions of the candidate object are met first, the prohibition on becoming the main object is lifted, and the candidate object is switched to the new current main object. When the end of the observation window is reached and the independent confirmation conditions of the candidate object are not met first, the current main object is maintained. The association result generation unit is used to generate cultural heritage object discourse association results corresponding to local film and television clips based on the current main object maintenance result or the main object switching result.

2. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 1, characterized in that, The main object determination unit extracts object representations from the text and image content of local film and television clips. The object representations include at least name representations, descriptive representations, and visual object representations.

3. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 2, characterized in that, The main object determination unit combines the standard object name library, alias library and type constraint rules to perform name normalization, alias normalization and object-level matching on object expressions; it retains only object expressions that can form a unique and valid mapping in the cultural heritage knowledge graph; and determines the object that completes the knowledge graph mapping first as the current main object according to the order of appearance, and determines the remaining objects that complete the knowledge graph mapping as candidate objects.

4. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 1, characterized in that, The local candidate suppression unit maps the current main object and candidate objects to the cultural heritage knowledge graph, and checks whether there is a local relationship between them based on the pre-registered local relationship table; local relationships include at least local composition relationships, subordinate component relationships, local landscape relationships, and subordinate display relationships.

5. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 4, characterized in that, The local candidate suppression unit performs an independent confirmation condition check on the candidate objects that hit local relationships to determine whether the candidate objects have met the independent confirmation conditions at the current screening position. The independent confirmation condition check includes at least checking whether the candidate objects have a clear name expression that can be uniquely mapped to the candidate objects as a whole at the current screening position, and whether there are continuous narrative clues or continuous visual clues surrounding the candidate objects after the clear name expression. When a candidate does not meet the independent confirmation criteria at the current screening position, the corresponding judgment position is recorded as the observation starting point position, and the candidate is placed in a state where it is prohibited from becoming the main candidate.

6. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 1, characterized in that, The observation window generation unit organizes the text content and image content in a local video clip into a unified discrete judgment position sequence; for judgment positions that contain both text content and image content, they are merged and registered as the same judgment position according to a unified time progression order; for judgment positions that contain only text content or only image content, they are registered separately as the corresponding judgment positions according to their actual appearance order. Both the observation window generation unit and the window comparison and determination unit perform position-by-position checks according to a unified discrete determination position sequence.

7. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 6, characterized in that, The observation window generation unit traces backward from the previous judgment position before the observation starting point, performing a succession check on each of the preceding consecutive judgment positions. When a preceding consecutive judgment position has a clear name expression that can uniquely map to the current main object as a whole, or a continuous screen clue that can stably point back to the current main object as a whole, the succession is recorded as successful and the backward tracing continues. When a preceding consecutive judgment position neither forms a name-side succession nor a screen-side succession, the backward tracing stops. Among these, weak representations, local component names, close-up views of local components, and local screens that cannot uniquely point back to the current main object as a whole are not counted as successful successions.

8. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 1, characterized in that, The observation window generation unit compares the preceding continuity length with the preset continuous continuity lower limit, the number of evidence hits for the current main object's overall name with the preset name continuity lower limit, and the number of evidence hits for the current main object's overall screen with the preset screen continuity lower limit. When at least two of the three comparison results reach the corresponding lower limit, a positive observation length adjustment value is generated; when at least two of the three comparison results do not reach the corresponding lower limit, a negative observation length adjustment value is generated; in other cases, the preset basic observation length remains unchanged; and the observation window length is generated based on the preset basic observation length and the observation length adjustment value to determine the end point of the observation window.

9. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 1, characterized in that, The window comparison and judgment unit checks the independent confirmation conditions of candidate objects in the observation window in the order of self-confirmation followed by expansion and continuation. When a certain continuous judgment position has a clear name expression that can be uniquely mapped to the whole candidate object, the self-confirmation is recorded as successful. After the self-confirmation is successful, it continues to check whether the subsequent continuous judgment positions form a continuous narrative or continuous screen expansion around the candidate object. When the self-confirmation is successful and expansion and continuation are subsequently obtained, the independent confirmation condition of the candidate object is recorded as successful.

10. The knowledge graph-based cultural heritage film and television discourse association mining and analysis system according to claim 9, characterized in that, The window comparison and judgment unit records the first established position of the regression evidence of the current master object and the first established position of the independent confirmation condition of the candidate object, and compares the order of the two. When the first established position of the regression evidence of the current master object is earlier, the current master object is maintained. When the first established position of the independent confirmation condition of the candidate object is earlier, the prohibition on becoming the master object is lifted, and the candidate object is switched to the new current master object. When both are established at the same consecutive judgment position, or when the independent confirmation condition of the candidate object is not established before the end of the observation window, the current master object is maintained.

Images