Intelligent data acquisition method and system based on multi-modal fusion RFID

Abstract

The application discloses a multi-modal fusion RFID intelligent data acquisition method and system, relates to the technical field of data acquisition, and comprises the following steps: in a recognition area, a same RFID tag is read multiple times in a continuous time slice, and packaging image data in a corresponding time slice is synchronously acquired; batch information is obtained by performing visual analysis on the packaging image data, multiple reading information corresponding to the same RFID tag and visual analysis results are recorded in time sequence, and original acquisition records with time pointing characteristics are formed; and a change amplitude between adjacent time slices is calculated for the visual analysis results corresponding to the same RFID tag in the original acquisition records. According to the application, the same identifier corresponds to single batch information through time pointing type reservation and convergence section extraction, the consistency of warehouse data is improved, and unique attribution information segments are generated through delay suspension and section-by-section collection, so that fluctuation data is avoided to be stored in the warehouse, and the stability of traceability is ensured.

Description

Technical Field

[0001] This application relates to the field of data acquisition technology, specifically to a smart data acquisition method and system for multimodal fusion RFID. Background Technology

[0002] Multimodal processing refers to the unified perception and collaborative analysis of different types of data (such as images, text, coded information, and radio frequency signals) within the same processing framework. By performing structured extraction and semantic alignment of multi-source information within the same time slice, various types of data can form a semantically related expression. Based on this, multimodal RFID-integrated intelligent data collection refers to the simultaneous acquisition of textual and UDI encoded information from the packaging image after an item enters the identification area, and the reading of the corresponding RFID tag. During this process, the visual analysis results and RFID identification results are bound, matched, and verified for consistency, thereby forming a unified data unit with multi-source mutual verification characteristics. Subsequently, this data unit directly drives the warehousing process and simultaneously completes compliance verification and data archiving, thus achieving high-precision, traceable data collection and closed-loop business processing without human intervention.

[0003] The existing technology has the following shortcomings:

[0004] In scenarios involving repeated warehousing, intelligent data acquisition devices continuously read the same RFID tag within a short period, acquiring consistent identification information. However, due to changes in angle during placement or movement, the visual recognition's analysis of packaging surface information fluctuates, causing the same item to be analyzed for different batch numbers at different times. During this process, the system binds the RFID tag to the visual analysis results based on temporal proximity, creating records of multiple batches for the same tag. These records are then considered valid data and written into inventory during subsequent warehousing processes, leading to batch-level conflicts in inventory data. This affects data consistency and reduces the accuracy of subsequent traceability.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this application is to provide an intelligent data acquisition method and system for multimodal fusion RFID to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, this application provides the following technical solution: a smart data acquisition method for multimodal fusion RFID, comprising the following steps:

[0008] Within the identification area, the same RFID tag is read multiple times within a continuous time slice, and packaging image data within the corresponding time slice is acquired simultaneously. Batch information is obtained by performing visual analysis on the packaging image data. The multiple reading information and visual analysis results corresponding to the same RFID tag are recorded in chronological order to form an original collection record with time-pointing characteristics.

[0009] For the visual analysis results corresponding to the same RFID tag in the original collection records, the change amplitude between adjacent time slices is calculated, and the change amplitude is segmented and merged along the time series. Continuous segments with the change amplitude in a convergent state are selected to generate stable interval marking information corresponding to time continuity.

[0010] Based on the stable interval marking information, the radio frequency reading information in the corresponding time slice of the original acquisition record is rearranged in order, and only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result is retained. The mapping relationship in the unstable interval is delayed and suspended to obtain the controlled mapping set.

[0011] For multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, the content with the convergence of change magnitude is compressed and aggregated segment by segment according to the temporal adjacency relationship to generate a unique attribution information segment.

[0012] The write operation is delayed based on the unique attribution information fragment, and the write process is only performed on the converged visual analysis result, thereby eliminating the conflict of multiple batches of information corresponding to the same RFID tag.

[0013] Preferably, forming the original acquisition record with time-pointing characteristics includes the following steps:

[0014] The same RFID tag is continuously observed and divided into multiple time slices at fixed time intervals, and multiple reads are performed to obtain RFID tag reading information.

[0015] Packaging image data is collected synchronously for each time slice, and RFID tag reading information is bound to packaging image data through a unified time identifier to form a set of raw data pairs arranged by time slice;

[0016] Visual analysis processing is performed on the packaging image data of each time slice in the raw data set to extract batch information and match it with the RFID tag reading information of the corresponding time slice to obtain a record sequence containing batch information;

[0017] The recorded sequences are sorted by time identifiers and the batch information differences between adjacent time slices are preserved. These are then integrated to form the original acquisition records indexed by time slices, thereby constructing original acquisition records with time-pointing characteristics.

[0018] Preferably, generating stable interval marker information corresponding to time continuity includes the following steps:

[0019] Extract the visual analysis results corresponding to the same RFID tag in the original collection records, expand the batch information according to the character position and content, and perform character position comparison on adjacent time slices to record the number of inconsistent characters, the index of the difference position and the length change, forming a change range record;

[0020] Organize the change range records and traverse them in time slice order. Use the first change range record as the starting reference and group change range records with the same number of inconsistent characters, difference position index and length change into the same segment interval.

[0021] Process segmented intervals and count the number of time slices. Filter segmented intervals that have consistent character inconsistencies, difference position indexes, and length changes, and meet the requirements for the number of consecutive time slices. Mark them as stable intervals and record the start and end time slices.

[0022] The stable intervals are summarized and numbered, and interval labeling is performed on the time slices to form stable interval labeling information containing interval number, start time slice, end time slice and coverage area.

[0023] Preferably, performing character-by-character position comparison on adjacent time slices includes: recording the total length of characters and marking length change information, comparing the character content at the same position one by one, marking positions with inconsistent character content and counting the number of inconsistent positions, and recording the specific position index corresponding to the inconsistent characters.

[0024] Preferably, obtaining the controlled mapping set includes the following steps:

[0025] Read the original acquisition records and combine them with the stable interval marker information to obtain the interval marker status. Extract the radio frequency reading information and visual analysis results corresponding to the time slices in the stable interval and register them in pairs to form a mapping relationship.

[0026] Organize the mapping relationships and sort them according to the starting time slice number of the stable interval. Arrange the mapping relationships within the same stable interval in ascending order of time slice number and connect them to form a mapping master set, while keeping the corresponding records of time slices in the unstable interval stored independently.

[0027] The delay suspension process is performed and the time slices in the unstable interval are grouped and marked. The grouped RF reading information and visual analysis results are then integrated with the mapping master set to form a controlled mapping set.

[0028] Preferably, grouping and marking time slices within the unstable interval includes: segmenting and organizing time slices that continuously belong to the unstable interval according to the time slice number, dividing time slices with consecutive time slice numbers into the same group, recording the corresponding start time slice number and end time slice number, and arranging the radio frequency reading information and visual analysis results corresponding to each group in chronological order and assigning a suspension mark.

[0029] Preferably, generating a unique attribution information fragment includes the following steps:

[0030] Extract the visual analysis results corresponding to the stable interval mapping relationship in the controlled mapping set, expand and record the character position, character content and total character length according to the time slice number, and perform character position comparison between adjacent time slices to form the change range;

[0031] The variation range is organized along the time slice number direction. Visual analysis results with consistent character inconsistency number, difference position index, and character length variation are divided into the same segment interval, and the corresponding start time slice number and end time slice number are recorded.

[0032] The visual analysis results within the segmented intervals are statistically analyzed one by one. The batch information with the longest consecutive occurrence is selected and bound to the start time slice number and end time slice number to generate a unique attributed information segment arranged in time progression.

[0033] Preferably, the process of comparing the position of characters between adjacent time slices to determine the range of change includes: recording the total length of characters, comparing the content of characters at the same position one by one, marking the positions where the character content is inconsistent and counting the number of inconsistent positions, and recording the specific position index of the inconsistent characters and the change in character length.

[0034] Preferably, the write processing is performed only on the converged visual resolution results, including the following steps:

[0035] Construct a write preparation queue corresponding to a unique ownership information segment, record RFID tag identification information, batch information content, start time slice number and end time slice number, and mark it as a non-triggered write state;

[0036] The delayed trigger control process is executed. The visual analysis results of the subsequent time slices of the unique ownership information fragment that has not been triggered to write state are continuously read and compared and recorded. The unique ownership information fragment that meets the continuous consistency condition is updated to the triggered write state and added to the write execution set.

[0037] After the write execution set is constructed, the inbound write is performed in chronological order, writing the RFID tag identification information, batch information and corresponding time range into the inventory record, completing the write processing only on the converged visual parsing results.

[0038] The intelligent data acquisition system of multimodal fusion RFID includes a multimodal time-series acquisition module, a change amplitude analysis module, a stable mapping construction module, a home segment generation module, and a delay writing control module;

[0039] The multimodal time-series acquisition module is used to read the same RFID tag multiple times within a continuous time slice within the identification area and simultaneously acquire packaging image data within the corresponding time slice; batch information is obtained by performing visual analysis on the packaging image data, and the multiple reading information corresponding to the same RFID tag and the visual analysis results are recorded in chronological order to form an original acquisition record with time-pointing characteristics.

[0040] The variation amplitude analysis module is used to calculate the variation amplitude between adjacent time slices based on the visual analysis results of the same RFID tag in the original collection record, and to segment and merge the variation amplitude along the time series, filter out continuous segments where the variation amplitude is in a convergent state, and generate stable interval marking information corresponding to time continuity.

[0041] The stable mapping construction module rearranges the radio frequency reading information in the corresponding time slice of the original acquisition record based on the stable interval marking information, retains only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result, and delays and suspends the mapping relationship in the unstable interval to obtain the controlled mapping set.

[0042] The attribution fragment generation module is used to compress and aggregate the content whose change amplitude is in a convergent state according to the temporal adjacency relationship for multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, and generate a unique attribution information fragment.

[0043] The delayed write control module is used to delay the triggering of the write operation based on the unique attribution information fragment, and only performs write processing on the converged visual parsing results.

[0044] The technical effects and advantages provided by this application in the above technical solution are as follows:

[0045] This application constructs an intelligent data acquisition method for multimodal fused RFID. It retains time-directed information from multiple reads of the same RFID tag within a continuous time slice and extracts convergence segments from the fluctuation trajectory of visually analyzed content, allowing the visual analysis results to gradually stabilize over time. Based on this, a mapping relationship between RFID read information and visually analyzed content is established within the stable interval, constraining the data writing basis from the source. This ensures that the same tag always corresponds to a single batch of information, thereby improving the consistency of data entering the database under dynamic acquisition environments and reducing the impact of analysis fluctuations caused by posture changes on the final data results.

[0046] This application delays and suspends the mapping relationship within unstable intervals, and compresses and aggregates multiple batches of visually analyzed content segment by segment in a controlled mapping set. During the time progression, a unique attribution information fragment is generated, and the write operation is delayed and controlled. This ensures that the write behavior only responds to information that has been aggregated, thereby preventing intermediate fluctuations from entering the inventory record. This maintains a consistent state in the expression of the inbound data, ensures a stable correspondence during subsequent traceability, and reduces data anomalies caused by conflicts between multiple batches. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0048] Figure 1 This is the overall flowchart of this application.

[0049] Figure 2 A flowchart for generating stable interval marking information corresponding to time continuity for this application.

[0050] Figure 3 This is a flowchart for obtaining the controlled mapping set for this application.

[0051] Figure 4 This is a schematic diagram of the modules in this application. Detailed Implementation

[0052] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0053] This application provides, as follows: Figures 1 to 3 The intelligent data acquisition method for multimodal fusion RFID shown includes the following steps:

[0054] Step 1: Within the identification area, the same RFID tag is read multiple times within consecutive time slices, and packaging image data within the corresponding time slices is acquired simultaneously. Batch information is obtained by performing visual analysis on the packaging image data. The multiple read information and visual analysis results corresponding to the same RFID tag are recorded in chronological order to form an original collection record with time-pointing characteristics.

[0055] The specific steps to generate original acquisition records with time-related characteristics are as follows:

[0056] Within the identification area, target items entering the area are continuously observed, and time is discretely divided at fixed time intervals. The entire identification process is divided into multiple sequentially arranged time slices, each time slice corresponding to a unique time identifier. At the beginning of each time slice, an RFID reading action is triggered to read the same RFID tag, and the reading result is recorded within the same time slice, so that multiple reading information of the RFID tag can be obtained within multiple consecutive time slices.

[0057] Setting a fixed time interval refers to discretely sampling the identification process with a uniform and constant time period during continuous observation. The specific setting method is as follows:

[0058] Based on the moving speed of the target item within the recognition area, the radio frequency reading response time, and the required clarity of the packaging image acquisition, a time sampling period is predetermined to ensure that the time interval between two adjacent time slices remains consistent. This time interval ensures that multiple valid radio frequency readings are acquired throughout the entire process of the item passing through the recognition area, and also enables the packaging images between consecutive time slices to produce distinguishable visual differences during posture changes. At the same time, it avoids redundant information due to excessively short time intervals or loss of key changes due to excessively long time intervals.

[0059] By using this time interval as a unified unit of time division, the entire acquisition process forms a uniformly distributed time slice structure in the time dimension, thereby providing a stable and continuous data foundation for subsequent calculation of the change amplitude based on time series and extraction of stable intervals.

[0060] While performing radio frequency reading, the packaging image data of the current target item is acquired within the corresponding time slice. The acquired packaging image data covers the area on the surface of the item that contains batch information, and ensures that each time slice corresponds to one or more frames of image data.

[0061] In this process, the RFID tag reading information acquired in each time slice is bound to the packaging image data acquired in the same time slice through a unified time identifier, so that the RFID tag reading information and the packaging image data form a one-to-one correspondence in the time dimension, thereby obtaining a set of original data pairs arranged in the order of time slices, where each data pair contains the reading information of the same RFID tag and the packaging image data in the corresponding time slice.

[0062] After obtaining packaging image data within consecutive time slices, visual analysis processing is performed on the packaging image data of each time slice one by one. During the visual analysis process, the region containing batch information in the image is located, and character information is extracted from the region. The extracted character information is organized according to a preset character format to obtain the batch information corresponding to that time slice. After obtaining the batch information, the batch information is paired with the RFID tag reading information in the same time slice, and the corresponding batch information field is added to the original data pair set, so that the data structure corresponding to each time slice simultaneously contains RFID tag reading information, packaging image data, and batch information obtained by visual analysis.

[0063] The preset character format refers to a structured character description specification that is pre-established for the expression rules of batch information on the packaging of target items. This specification is used to limit the organization of batch information in terms of character composition, arrangement order and field structure. For example, it specifies that batch information is composed of a fixed length of letters and numbers, or is formed by concatenating date fields and serial number fields in a specific order, and defines the distribution range of each type of field in the character position.

[0064] This character format is usually obtained through statistical analysis of historical sample data in actual application scenarios. This involves collecting multiple batches of real packaging images, summarizing and organizing the character length, character type distribution, and arrangement patterns of the batch information, thereby forming a unified character structure template, which is then used as the basis for organizing the visual analysis results.

[0065] In actual processing, the extracted original character information will be reorganized and corrected according to the preset format, so that irregular or recognition-biased character sequences can be standardized and expressed, thereby obtaining batch information that meets the requirements of a unified format, ensuring the structural consistency of the parsing results in different time slices, and providing a stable input basis for subsequent calculation of change amplitude and determination of convergence interval.

[0066] When processing data from multiple consecutive time slices, the time slice order is kept unchanged, and the batch information within each time slice is arranged sequentially according to the time sequence. This allows the batch information corresponding to the same RFID tag in different time slices to form a record sequence arranged in a time progression. The batch information content obtained from each visual analysis is completely preserved in this record sequence, without deleting or replacing any record, to ensure that the entire process of batch information changing over time can be reflected.

[0067] After batch information extraction and pairing with RFID tag reading information, the data in all time slices are uniformly sorted according to time identifiers, so that the data pairs corresponding to each time slice are arranged sequentially in the overall record according to the time progression direction. After sorting, the batch information between adjacent time slices is compared one by one, and the differences in batch information parsed in different time slices are preserved in their original form.

[0068] In the process of preserving differences, the differences are not merged. Instead, the batch information within each time slice is stored as an independent record. This ensures that even if the batch information obtained from the parsing of the same RFID tag in multiple time slices has different content, these different contents can still correspond to their respective time slice positions in the record. In this way, the visual parsing results of the same RFID tag in consecutive time slices can be presented in a point-by-point unfolding form, thereby forming a continuous trajectory reflecting the change process of batch information in the overall record, while ensuring that the batch information corresponding to each time slice in this trajectory can be completely traced.

[0069] After completing the time sorting and point-by-point retention of differences, the multiple reading information of the same RFID tag in consecutive time slices is integrated with the batch information in the corresponding time slices to construct the original acquisition record structure indexed by time slices. In this original acquisition record structure, each time slice corresponds to a complete record. The record contains the RFID tag reading information in the time slice and the batch information obtained by parsing the packaging image data, and is arranged in the order of the time slices to form a complete record chain.

[0070] In this record chain, the arrangement of consecutive time slices reflects the change trajectory of batch information in the time dimension, so that the parsing results of the same RFID tag in different time slices can be presented in a continuous manner, and form an original collection record with time-oriented characteristics as a whole.

[0071] This original data collection record not only reflects the multiple readings of the same RFID tag within a continuous time slice, but also completely records the change process of the corresponding batch information, thus providing a continuous and complete data foundation for further processing of batch information based on time progression.

[0072] Step 2: For the visual analysis results corresponding to the same RFID tag in the original collection records, calculate the change amplitude between adjacent time slices, and divide the change amplitude into segments along the time series. Filter out continuous segments where the change amplitude is in a convergent state to generate stable interval marking information corresponding to time continuity.

[0073] The specific steps for generating stable interval marker information corresponding to time continuity are as follows:

[0074] Given that the original data collection records have been arranged in time slice order, the visual analysis results corresponding to the same RFID tag are extracted one by one. The batch information recorded in each time slice is expanded into a complete character sequence. Specifically, each batch of information is recorded in character order, with the position and content of each character recorded sequentially, so that the visual analysis results in each time slice form an ordered character set composed of multiple character positions. Based on this, starting from the first time slice, the current time slice and the next adjacent time slice are selected sequentially, and the batch information corresponding to the two is compared character by character position. During the comparison, the total character length is recorded first. When the character lengths of the two time slices are inconsistent, the length difference is recorded and marked as length change information. Secondly, when the character lengths are consistent, the character content at the same position is compared one by one. Positions with inconsistent character content are marked, and the number of inconsistent positions is counted. At the same time, the specific position index of these inconsistent characters in the entire character sequence is recorded.

[0075] After completing the item-by-item comparison of character content and length, the number of inconsistent characters, the index of the specific difference location, and the length change are integrated into the change amplitude result between the current two time slices. This result is recorded together with the corresponding time slice number, so that a complete change amplitude record is formed between each pair of adjacent time slices. In this way, a set of change amplitude records covering all adjacent time slices is gradually built up throughout the entire time process.

[0076] After completing the recording of the change amplitude between all adjacent time slices, the set of change amplitude records is continuously traversed in the order of the time slices. During the traversal, the first change amplitude record is used as the starting reference of the current segment, and its corresponding number of inconsistent characters, difference position index, and length change are used as the initial standard. Then, the change amplitude records corresponding to subsequent time slices are compared one by one based on this standard.

[0077] When a subsequent change amplitude record matches the current segment's starting reference in terms of the number of inconsistent characters, the difference position index is exactly the same, and the character length change is also consistent, the change amplitude record corresponding to that time slice is included in the current segment, and the same comparison operation continues to the next segment. When a change amplitude record is inconsistent with the current segment's starting reference in any of the above aspects, the merging process of the current segment is immediately terminated, and the starting and ending time slices of the current segment are recorded. Then, the inconsistent change amplitude record is used as the new segment's starting reference, and the above merging process is repeated.

[0078] In this way, the entire set of change amplitude records is divided into multiple consecutively arranged segment intervals. Each segment interval maintains the same number of inconsistent characters, difference position index, and length change, thus forming multiple interval structures with completely consistent internal change amplitudes as time progresses.

[0079] After segmenting the variation range, each segment is analyzed and processed individually. Specifically, the number of time slices covered by each segment is counted, and the consistency of the variation range records within the segment is used for screening. First, a minimum number of consecutive time slices is set; for example, segmented intervals with at least three consecutive time slices are considered candidate intervals. Second, the variation range records within the candidate intervals are checked again to ensure that all variation range records within the candidate interval are completely consistent in terms of the number of inconsistent characters, the index of the difference position, and the length variation, and that this consistency is not interrupted throughout the interval. Segmented intervals that meet both of these conditions are marked as stable intervals, and the start and end time slice numbers and the interval length are recorded. Segmented intervals that do not meet the requirement of a certain number of consecutive time slices or have interrupted internal consistency are not marked.

[0080] The minimum number of consecutive time slices is set to limit the duration of changes in amplitude while maintaining a consistent state over time. The specific rules for setting this minimum number are as follows:

[0081] The minimum number of consecutive time slices is determined by comprehensively considering the passage time of target items within the identification area, the time slice interval, and the fluctuation period of visual analysis results during attitude changes. This ensures that the number covers at least one complete stable observation phase while excluding accidental consistency caused by instantaneous identification errors or short-term attitude disturbances. Specifically, by statistically analyzing the change trajectory of visual analysis results corresponding to the same RFID tag during historical acquisition, the shortest continuous time slice range corresponding to the stable expression phase of batch information is determined. This range is used as the basis for determining the minimum number of consecutive time slices, while ensuring that this number is greater than the number of repeated consistency that may occur within a single fluctuation period, thereby avoiding misidentification of short-term repetitions as stable intervals.

[0082] By setting this rule, the segmented intervals obtained through screening have sufficient continuity in time, which can truly reflect the time range within which the visual analysis results enter a stable state, and provide a reliable basis for subsequent stable interval marking.

[0083] This filtering process ensures that all retained segmented intervals correspond to a consistent time range in the visual analysis results across multiple consecutive time slices, thereby extracting a set of intervals with continuity and consistency during the time progression.

[0084] After obtaining all stable intervals, these intervals are uniformly numbered, and the start and end time slices of each interval are marked according to the time slice order. Each time slice is marked to indicate whether it is within a certain stable interval, so that each time slice in the original acquisition record can clearly correspond to the interval state it belongs to. For time slices that are within stable intervals, the corresponding stable interval number and the start and end range information of the interval are marked in the record. For time slices that are not covered by any stable interval, they are marked as unstable intervals.

[0085] After completing the interval labeling of all time slices, the number, start time slice, end time slice, and coverage information of all stable intervals are summarized to form a complete set of stable interval labeling information. This set can not only clearly reflect the distribution of each stable interval in the time process, but also accurately indicate the interval belonging of each time slice in the entire time series, thus providing a clear data basis for subsequent sequential rearrangement and mapping of the original acquisition records based on stable intervals.

[0086] Step 3: Based on the stable interval marking information, the radio frequency reading information in the corresponding time slice of the original acquisition record is rearranged in order, and only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result is retained. The mapping relationship in the unstable interval is delayed and suspended to obtain the controlled mapping set.

[0087] The controlled mapping set is obtained through the following steps:

[0088] With the stable interval marking information already marked one by one to the corresponding time slot position of the original acquisition record, the original acquisition records formed by the same RFID tag in all time slots are processed one by one. The specific implementation process is as follows: each record is read sequentially from the starting time slot according to the time slot number, and the interval marking status corresponding to the time slot is obtained simultaneously during the reading process; during the reading process, each time slot is divided into two categories: time slots within stable intervals and time slots within unstable intervals, and mapping extraction processing is performed for time slots within stable intervals, that is, the radio frequency reading information and the corresponding visual analysis result are extracted from the time slot record, and the two items are registered in pairs with a one-to-one correspondence, while the time number of the time slot and the number of the stable interval to which it belongs are recorded.

[0089] After completing the mapping extraction of a single time slice, subsequent time slices are read. When multiple consecutive time slices belong to the same stable interval, the mapping relationships corresponding to these time slices are arranged in ascending order of time number, so that all time slices within the same stable interval form a continuous mapping segment. The starting time slice number is recorded at the beginning position of the mapping segment, and the ending time slice number is recorded at the end position. This results in a complete set of mapping relationships between radio frequency reading information and visual analysis results arranged in chronological order within each stable interval.

[0090] Meanwhile, for time slices within unstable intervals encountered during the sequential reading process, their mapping relationships are not included in the aforementioned mapping segments. Instead, the radio frequency reading information and visual analysis results of the time slice are extracted separately and registered together with their corresponding time numbers in an independent recording area, so that they remain in their original recording state and do not participate in the construction of the current mapping segment before subsequent processing.

[0091] After constructing the mapping segments for all time slices within stable intervals, the multiple mapping segments corresponding to different stable intervals are sorted in order as a whole. Specifically, they are sorted according to the starting time slice number of each stable interval, with the mapping segments with smaller starting time slice numbers placed first and the mapping segments with larger starting time slice numbers placed last, so that the mapping segments of multiple stable intervals form a sequential relationship in the time dimension. While sorting, the time order within each mapping segment is kept unchanged. That is, within the same stable interval, the mapping relationship is still arranged in the order of the original time slices. The time interval range between adjacent segments is recorded between each mapping segment so that the boundary positions between different stable intervals can be distinguished in subsequent processing.

[0092] After sorting, all stable interval mapping segments are sequentially connected to form a complete mapping master set. In this mapping master set, each record contains a time slice number, radio frequency reading information, and the corresponding visual resolution result. All these records originate from within the stable interval, thus ensuring that all mapping relationships in the mapping master set are established within the stable interval range. At the same time, the time slice data in the unstable interval recorded separately in the previous step are kept in their original order and are not inserted into the mapping master set, thereby creating a state in which stable interval mapping relationships are separated from unstable interval records in the overall structure.

[0093] After the main set of stable interval mappings is constructed, delayed suspension processing is performed on the records in unstable intervals. Specifically, the time slices in all unstable intervals are segmented and organized according to the time slice number. Time slices with consecutive time slice numbers that all belong to unstable intervals are grouped into the same group, and each group is assigned an independent suspension number. At the same time, the start time slice number and end time slice number of the group are recorded. After the grouping is completed, the radio frequency reading information and visual resolution results corresponding to the time slices of each group of unstable intervals are arranged in chronological order and stored together with their suspension numbers, so that this part of the mapping relationship forms an independent set of suspended records in the overall data structure.

[0094] In this process, mapping relationships within unstable intervals are not deleted or merged; instead, their complete record status is maintained, and they are identified by suspension numbers and time ranges, preventing them from participating in the output of stable interval mapping relationships at the current stage. After grouping and suspending all unstable interval records, the main set of stable interval mappings and the set of suspended unstable interval records are unified and organized to form a dataset containing two parts: the first part is the set of stable interval mapping relationships arranged in chronological order, and the second part is the set of unstable interval mapping relationships divided by suspension numbers. This method constructs a controlled mapping set, enabling the mapping relationship between RF reading information and visual analysis results to exhibit a hierarchical management state as time progresses, and providing a clear data organization form for subsequent processing.

[0095] Step 4: For multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, compress and aggregate the content whose change magnitude is in a convergent state according to the temporal adjacency relationship to generate a unique attribution information segment.

[0096] The specific steps for generating a unique attribution information fragment are as follows:

[0097] Under the premise that the controlled mapping set has been separated from the unstable interval suspension records according to the stable interval mapping relationship, the stable interval mapping relationship corresponding to the same RFID tag is extracted one by one, and then expanded sequentially from the smallest number according to the time slice number. A clear association is established between the visual analysis result in each record and its corresponding time slice number.

[0098] During the unfolding process, the visual analysis results within each time slice are recorded completely in character order, including the total character length, the specific position of each character, and the character content itself, so that each visual analysis result is presented in the form of a complete character sequence. On this basis, taking adjacent time slices as units, the comparison process is carried out one by one from the first record to the next. The visual analysis results corresponding to the current time slice and the next time slice are compared character by character position, and the consistency of character content is recorded one by one. At the same time, the change of character length and the position of the change are recorded. The number of inconsistent positions, the specific position index, and the character length change are collectively used as the change range content, and this change range content is bound to the current time slice number for recording.

[0099] Through the above-mentioned step-by-step expansion and comparison, all visual analysis results of the same RFID tag within the stable range are formed into a set of records arranged in time progression. Each record is associated with the change amplitude information between the previous time slice and the previous time slice, thereby constructing a continuous change amplitude record structure in the time dimension, providing a complete data foundation for subsequent segmented compression and aggregation.

[0100] After completing the correspondence between the visual analysis results and the change range records, the entire record set is continuously segmented starting from the time slice number. Specifically, the first visual analysis result is selected as the starting content of the current segment, and the change range corresponding to this starting content is used as the reference standard for the current segment. Then, the change range corresponding to each subsequent time slice is checked one by one from the starting time slice. When the change range corresponding to the subsequent time slice is consistent with the current segment reference standard in terms of the number of inconsistent characters, the index of the character difference position, and the change in character length, the visual analysis result corresponding to that time slice is included in the current segment, and the segment range continues to extend. When the change range corresponding to a certain time slice changes in any of the above three aspects, the extension of the current segment is immediately terminated, the starting time slice number and the ending time slice number of the current segment are recorded, and all visual analysis results in this segment are saved as an independent segment. Then, the segmentation process is repeated with the time slice where the change range changed as the new starting point.

[0101] By continuously executing this segmented processing, the visual analysis results throughout the entire time progression are divided into multiple independent segmented intervals. Each segmented interval corresponds to a continuous time slice range with consistent variation amplitude, thereby forming multiple visual analysis results with consistent variation characteristics in the time dimension.

[0102] After completing the segmentation of all intervals, the visual analysis results within each segment interval are compressed and aggregated segment by segment. During the processing, all visual analysis results within each segment interval are statistically analyzed one by one. The specific implementation process is as follows: count the number of time slices in which each type of batch information appears in the segment interval, and record the continuous occurrence range of the batch information in the interval. In the statistical process, the batch information with the most consecutive occurrences of time slices and the longest coverage time range in the segment interval is selected as the content to which the segment interval belongs. At the same time, the content to which the segment interval belongs is bound and recorded with the start time slice number and end time slice number of the segment interval.

[0103] After compressing and aggregating a single segment interval, the same process is performed on all segment intervals in sequence, so that each segment interval generates a corresponding attribution record. These attribution records are arranged in the order of time slice numbers, thus forming a unique set of attribution information fragments arranged in time progression. In this set, each information fragment contains uniform batch information content and a corresponding time range, and each information fragment comes from a continuous time slice interval with a consistent range of change.

[0104] Through the above processing, the visual resolution results of multiple batches of the same RFID tag in the controlled mapping set are integrated into several time segments with unique attribution relationships, thereby providing a clear and continuous attribution expression form for subsequent processing.

[0105] The convergence of the change amplitude specifically means that, during the continuous advancement of time slices, the differences in the visual resolution results between adjacent time slices at the character content level no longer expand or change. The number of inconsistent characters, the index of the difference position, and the changes in character length remain consistent across multiple consecutive time slices. In other words, the change amplitude between consecutive time slices is stable and unchanging. In this state, although there may be differences in the visual resolution results, these differences are fixed in the time dimension and no longer fluctuate with changes in the object's posture. This indicates that the resolution results within the current time interval have entered a stable expression stage from the dynamic fluctuation stage, which can serve as the basis for subsequent segmentation, merging, and attribution determination.

[0106] Step 5: Delay the triggering control of the inbound writing operation based on the unique attribution information fragment, and only perform the writing process on the converged visual analysis result, thereby eliminating the conflict of multiple batches of information corresponding to the same RFID tag.

[0107] Only the converged visual analysis results are written; the specific steps are as follows:

[0108] After a set of unique attribution information segments arranged in chronological order has been formed, each unique attribution information segment corresponding to the same RFID tag is processed one by one and sorted according to the starting time slot number of each information segment. Information segments with smaller starting time slot numbers are arranged first, and information segments with larger starting time slot numbers are arranged last, thus forming a write preparation queue arranged in the direction of time progression.

[0109] In the write preparation queue, a complete record is created for each uniquely owned information segment. The record includes the RFID tag identification information corresponding to the information segment, the batch information after unified aggregation, the start time slice number, and the end time slice number. At the same time, an initial status identifier is attached to each record, marking it as a non-triggered write state. After the queue is built, all uniquely owned information segments are kept in a non-triggered write state, and visual analysis result updates from subsequent time slices are continuously received during the time progression. These updates are compared and recorded one by one with the end time slice number of each uniquely owned information segment in the current queue, so that each uniquely owned information segment undergoes a complete time extension process before entering the write process, thus making the data entry write operation based on a complete time range rather than a single time slice.

[0110] During the process of establishing and continuously updating the write preparation queue, delayed trigger control processing is performed on each uniquely assigned information segment starting from the beginning of the queue. Specifically, the currently processed information segment is selected, and the end time slice number of the information segment is used as the observation starting point. The newly added visual analysis results in subsequent time slices are read one by one. These newly added visual analysis results are compared with the batch information content recorded in the current information segment, and the comparison results are recorded. When the newly added visual analysis results are consistent with the batch information content of the current information segment in several consecutive time slices after the end time slice, and no new change amplitude is recorded within the consecutive time slices, the state of the current information segment is updated from the non-triggered write state to the triggered write state, and the information segment is added to the write execution set.

[0111] If a new visual analysis result appears in a time slice after the end of the current time slice, and the batch information content corresponding to the visual analysis result differs from the content recorded in the current information segment, then the state of the current information segment remains unchanged, and the observation range is extended to the next time slice. The same line-by-line comparison and recording of the visual analysis results in subsequent time slices continues until no new difference records appear within a continuous time slice.

[0112] Through the above-mentioned observation and status update process, each unique ownership information fragment undergoes a complete stable extension phase after its corresponding time range, thereby achieving delayed trigger control of the data entry write operation.

[0113] After the write execution set has been formed, each uniquely owned information segment in the set is processed for inbound writing in chronological order. During the process, the RFID tag identification information recorded in the information segment and the corresponding batch information are written into the inventory record on a unit basis. The start time segment number and end time segment number of the information segment are attached as time range identifiers during the writing process, so that each inbound record can clearly reflect its corresponding time interval.

[0114] After completing the writing of the current information fragment, the status of the current uniquely belonging information fragment is updated to the writing completed status, and the next information fragment in the writing execution set is processed until all information fragments in the set have completed the writing process.

[0115] Throughout the writing process, only the unique ownership information fragment formed after segmented compression and aggregation is written, and the visual analysis results of multiple batches that differ in the original collection records are not directly written. This ensures that each RFID tag corresponds to only one batch information fragment in the inventory record, and that the batch information comes from the unified aggregation results of continuous records within the time range. This eliminates the conflict of multiple batch information corresponding to the same RFID tag during the overall warehousing process, and keeps the inventory data consistent in both the time and content dimensions.

[0116] This application provides, as follows: Figure 4 The multimodal fusion RFID intelligent data acquisition system shown includes a multimodal time-series acquisition module, a change amplitude analysis module, a stable mapping construction module, a home segment generation module, and a delay write control module.

[0117] The multimodal time-series acquisition module is used to read the same RFID tag multiple times within a continuous time slice within the identification area and simultaneously acquire packaging image data within the corresponding time slice. Batch information is obtained by performing visual analysis on the packaging image data. The multiple reading information corresponding to the same RFID tag and the visual analysis results are recorded in chronological order to form an original acquisition record with time-pointing characteristics.

[0118] The variation amplitude analysis module is used to calculate the variation amplitude between adjacent time slices based on the visual analysis results of the same RFID tag in the original collection record, and to segment and merge the variation amplitude along the time series, filter out continuous segments where the variation amplitude is in a convergent state, and generate stable interval marking information corresponding to time continuity.

[0119] The stable mapping construction module rearranges the radio frequency reading information in the corresponding time slice of the original acquisition record based on the stable interval marking information, retains only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result, and delays and suspends the mapping relationship in the unstable interval to obtain the controlled mapping set.

[0120] The attribution fragment generation module is used to compress and aggregate the content with convergent change magnitude according to the temporal adjacency of multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, and generate a unique attribution information fragment.

[0121] The delayed write control module is used to delay the triggering of the write operation based on the unique attribution information fragment, and only performs write processing on the converged visual parsing results.

[0122] The intelligent data acquisition method for multimodal fusion RFID provided in this application is implemented through the aforementioned intelligent data acquisition system for multimodal fusion RFID. For details of the specific methods and processes of the intelligent data acquisition system for multimodal fusion RFID, please refer to the embodiments of the intelligent data acquisition method for multimodal fusion RFID described above, which will not be repeated here.

[0123] The foregoing has only described certain exemplary embodiments of this application by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of this application. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of this application.

Claims

1. A smart data acquisition method using multimodal fusion RFID, characterized in that, Includes the following steps: Within the identification area, the same RFID tag is read multiple times within consecutive time slices, and packaging image data within the corresponding time slices is acquired simultaneously; batch information is obtained by performing visual analysis on the packaging image data, and the multiple read information corresponding to the same RFID tag and the visual analysis results are recorded in chronological order to form an original acquisition record with time-pointing characteristics. For the visual analysis results corresponding to the same RFID tag in the original collection records, the change amplitude between adjacent time slices is calculated, and the change amplitude is segmented and merged along the time series. Continuous segments with the change amplitude in a convergent state are selected to generate stable interval marking information corresponding to time continuity. Based on the stable interval marking information, the radio frequency reading information in the corresponding time slice of the original acquisition record is rearranged in order, and only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result is retained. The mapping relationship in the unstable interval is delayed and suspended to obtain the controlled mapping set. For multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, the content with the convergence of change magnitude is compressed and aggregated segment by segment according to the temporal adjacency relationship to generate a unique attribution information segment. The write operation is delayed based on the unique attribution information fragment, and the write process is only performed on the converged visual parsing results.

2. The intelligent data acquisition method for multimodal fusion RFID according to claim 1, characterized in that, The process of creating original acquisition records with time-related characteristics includes the following steps: The same RFID tag is continuously observed and divided into multiple time slices at fixed time intervals, and multiple reads are performed to obtain RFID tag reading information. Packaging image data is collected synchronously for each time slice, and RFID tag reading information is bound to packaging image data through a unified time identifier to form a set of raw data pairs arranged by time slice; Visual analysis processing is performed on the packaging image data of each time slice in the raw data set to extract batch information and match it with the RFID tag reading information of the corresponding time slice to obtain a record sequence containing batch information; The recorded sequences are sorted by time identifiers and the batch information differences between adjacent time slices are preserved. These are then integrated to form the original acquisition records indexed by time slices, thereby constructing original acquisition records with time-pointing characteristics.

3. The intelligent data acquisition method for multimodal fusion RFID according to claim 1, characterized in that, Generating stable interval marker information corresponding to time continuity includes the following steps: Extract the visual analysis results corresponding to the same RFID tag in the original collection records, expand the batch information according to the character position and content, and perform character position comparison on adjacent time slices to record the number of inconsistent characters, the index of the difference position and the length change, forming a change range record; Organize the change range records and traverse them in time slice order. Use the first change range record as the starting reference and group change range records with the same number of inconsistent characters, difference position index and length change into the same segment interval. Process segmented intervals and count the number of time slices. Filter segmented intervals that have consistent character inconsistencies, difference position indexes, and length changes, and meet the requirements for the number of consecutive time slices. Mark them as stable intervals and record the start and end time slices. The stable intervals are summarized and numbered, and interval labeling is performed on the time slices to form stable interval labeling information containing interval number, start time slice, end time slice and coverage area.

4. The intelligent data acquisition method for multimodal fusion RFID according to claim 3, characterized in that, Perform character-by-character position comparison on adjacent time slices, specifically including: recording the total length of characters and marking length change information, comparing the content of characters at the same position one by one, marking positions with inconsistent character content and counting the number of inconsistent positions, and recording the specific position index corresponding to the inconsistent characters.

5. The intelligent data acquisition method for multimodal fusion RFID according to claim 3, characterized in that, Obtaining the controlled mapping set involves the following steps: Read the original acquisition records and combine them with the stable interval marker information to obtain the interval marker status. Extract the radio frequency reading information and visual analysis results corresponding to the time slices in the stable interval and register them in pairs to form a mapping relationship. Organize the mapping relationships and sort them according to the starting time slice number of the stable interval. Arrange the mapping relationships within the same stable interval in ascending order of time slice number and connect them to form the main mapping set. Delay suspension is performed on time slices within the unstable interval, and the time slices within the unstable interval are grouped and marked. The grouped RF reading information, visual analysis results and mapping master set are then integrated to obtain the controlled mapping set.

6. The intelligent data acquisition method for multimodal fusion RFID according to claim 5, characterized in that, The time slices within the unstable interval are grouped and marked. Specifically, the time slices that belong to the unstable interval are segmented and organized according to the time slice number. Time slices with consecutive time slice numbers are grouped into the same group, and the corresponding start time slice number and end time slice number are recorded. At the same time, the radio frequency reading information and visual analysis results corresponding to each group are arranged in chronological order and assigned a suspension mark.

7. The intelligent data acquisition method for multimodal fusion RFID according to claim 5, characterized in that, Generating a unique attribution information fragment involves the following steps: Extract the visual analysis results corresponding to the stable interval mapping relationship in the controlled mapping set, expand and record the character position, character content and total character length according to the time slice number, and perform character position comparison between adjacent time slices to form the change range; The variation range is organized along the time slice number direction. Visual analysis results with consistent character inconsistency number, difference position index, and character length variation are divided into the same segment interval, and the corresponding start time slice number and end time slice number are recorded. The visual analysis results within the segmented intervals are statistically analyzed one by one. The batch information with the longest consecutive occurrence is selected and bound to the start time slice number and end time slice number to generate a unique information segment arranged in time progression.

8. The intelligent data acquisition method for multimodal fusion RFID according to claim 7, characterized in that, The process involves comparing the position of characters between adjacent time slices to determine the magnitude of change. This includes recording the total length of characters, comparing the content of characters at the same position one by one, marking positions with inconsistent character content and counting the number of inconsistent positions, and recording the specific position index of the inconsistent characters and the change in character length.

9. The intelligent data acquisition method for multimodal fusion RFID according to claim 7, characterized in that, The write process is performed only on the converged visual resolution results, including the following steps: Construct a write preparation queue corresponding to a unique ownership information segment, record RFID tag identification information, batch information content, start time slice number and end time slice number, and mark it as a non-triggered write state; The delayed trigger control process is executed. The visual analysis results of the subsequent time slices are continuously read and compared for the unique ownership information fragments that have not been triggered to write state. The unique ownership information fragments that meet the continuous consistency condition are updated to trigger write state and added to the write execution set. The write execution set is executed in chronological order to write into the inventory record, writing the RFID tag identification information, batch information and corresponding time range into the inventory record, and completing the write processing only on the converged visual analysis results.

10. A multimodal fusion RFID intelligent data acquisition system, used to implement the multimodal fusion RFID intelligent data acquisition method according to any one of claims 1-9, characterized in that, It includes a multimodal time series acquisition module, a variation amplitude analysis module, a stable mapping construction module, a class segment generation module, and a delay write control module; The multimodal time-series acquisition module is used to read the same RFID tag multiple times within a continuous time slice within the identification area and simultaneously acquire packaging image data within the corresponding time slice; batch information is obtained by performing visual analysis on the packaging image data, and the multiple reading information corresponding to the same RFID tag and the visual analysis results are recorded in chronological order to form an original acquisition record with time-pointing characteristics. The variation amplitude analysis module is used to calculate the variation amplitude between adjacent time slices based on the visual analysis results of the same RFID tag in the original collection record, and to segment and merge the variation amplitude along the time series, filter out continuous segments where the variation amplitude is in a convergent state, and generate stable interval marking information corresponding to time continuity. The stable mapping construction module rearranges the radio frequency reading information in the corresponding time slice of the original acquisition record based on the stable interval marking information, retains only the mapping relationship between the radio frequency reading information in the stable interval and the visual analysis result, and delays and suspends the mapping relationship in the unstable interval to obtain the controlled mapping set. The attribution fragment generation module is used to compress and aggregate the content whose change amplitude is in a convergent state according to the temporal adjacency relationship for multiple batches of visual analysis results corresponding to the same RFID tag in the controlled mapping set, and generate a unique attribution information fragment. The delayed write control module is used to delay the triggering of the write operation based on the unique attribution information fragment, and only performs write processing on the converged visual parsing results.

Images