Search

CN-121526262-B - Cargo multimedia information intelligent extraction system and method

CN121526262BCN 121526262 BCN121526262 BCN 121526262BCN-121526262-B

Abstract

The invention belongs to the technical field of intelligent logistics, and discloses a system and a method for intelligently extracting goods multimedia information, wherein the system comprises the steps of collecting goods pictures, video clips and text labels uploaded by a goods owner, generating position coordinate binding by combining equipment positioning signals, and obtaining a multimedia position set; the method comprises the steps of performing content hierarchical scanning, separating visual elements and dynamic sequences to form a hierarchical structure to obtain a hierarchical content group, performing cross semantic bridging, constructing a correlation path among elements to obtain an integrated semantic chain, applying attribute extraction circulation to the integrated semantic chain, extracting cargo specification details and transportation constraint fragments to form an attribute set, generating a driver query sequence based on the structural attribute set and position coordinate binding, integrating path forward factors to perform adaptation sequencing to obtain a preferable driver list, and greatly improving the accuracy of freight supply and demand matching.

Inventors

  • LING YIBIN
  • WU YURONG
  • YANG JIAJIA

Assignees

  • 上海新颐科技软件有限公司

Dates

Publication Date
20260508
Application Date
20260115

Claims (7)

  1. 1. The intelligent cargo multimedia information extraction method is characterized by comprising the following steps of: Step S1, acquiring cargo pictures, video clips and text labels uploaded by a cargo owner, and generating position coordinate binding by combining equipment positioning signals to obtain a multimedia position set; step S2, performing content hierarchical scanning corresponding to the pictures and video clips in the media position set, and separating visual elements and dynamic sequences to form a hierarchical structure so as to obtain a hierarchical content group; Step S3, performing cross semantic bridging on the layered content group and the text labels to construct an inter-element association path to obtain an integrated semantic chain, wherein the method comprises the steps of extracting visual elements from the layered content group, performing semantic bridging on the visual elements and the text labels to generate a preliminary association path, and attaching bridging nodes to obtain a bridging path set; The method comprises the steps of selecting a path from a bridging path set, introducing text labels as reference anchors, executing semantic consistency check, and recording check nodes to obtain a check path group; setting bridging intensity values for the inspection path group, dynamically adjusting intensity based on inter-element distance and semantic overlapping, and applying an intensity propagation mechanism to obtain an intensity enhancement path set, wherein the intensity propagation mechanism transmits the credibility of a high-intensity bridging path to adjacent low-intensity paths, and the propagation rule is that if two paths share the same visual element or text entity, the intensity values of the low-intensity paths are according to the formula Enhancement is performed in which In order for the path enhancement value to be a value, In order to bridge the intensity value of the light, Intensity values for adjacent high intensity paths; carrying out branch scanning on paths in the strength enhancement path set, adding variation compensation and converting the variation compensation into a verification path set; Performing path compression circulation on the verification path set, merging adjacent nodes and attaching compression labels; Constructing an integrated semantic chain based on the verification path set, expanding the path to cover the multi-layer element association and embedding the multi-layer element association into an intra-chain fulcrum; S4, applying attribute extraction circulation to the integrated semantic chain, and extracting cargo specification details and transportation constraint fragments to construct an attribute set to obtain a structured attribute set; And S5, generating a driver query sequence based on the binding of the structured attribute set and the position coordinates, and merging the path forward factors to carry out adaptive sorting to obtain a preferred driver list.
  2. 2. The intelligent cargo multimedia information extraction method according to claim 1, wherein step S1 comprises: S11, receiving goods pictures and video clips from a goods owner device, synchronously capturing text label input, and attaching a transmission time sequence label to organize media elements to obtain an original multimedia package; S12, fusing equipment positioning signals to the original multimedia package, calibrating binding position coordinates through coordinate precision, constructing auxiliary anchor points by means of equipment sensor data to obtain a coordinate reinforced package, wherein the method comprises the steps of isolating cargo pictures, video clips and text labels from the original multimedia package, fusing the equipment positioning signals, recording the fusing time points, and obtaining a signal fusing group; performing coordinate precision calibration on the signal integration group, comparing the sensor data of the equipment and adjusting the binding strength, and obtaining an anchor point strengthening group through cross-verifying the corrected coordinate values by multiple signal sources; The method comprises the steps of binding and expanding elements in an anchor point strengthening group to adjacent media elements, and constructing a coordinate chain connection to obtain a chain binding group, wherein the mode of constructing the coordinate chain connection comprises the steps of expanding and binding the position coordinates of the media elements with higher binding strength in the anchor point strengthening group to the media elements adjacent in time sequence; applying position variation scanning to the chain binding group, adding variation compensation marks and converting into a coordinate enhancement packet; Step S13, dividing time axis marks corresponding to video clips in a coordinate enhancement package, binding associated position coordinates and embedding variation tracking points to obtain a time sequence position group, wherein the mode of embedding the variation tracking points comprises the steps that for each time axis mark point, a system associates the corresponding position coordinates according to the change condition of equipment positioning signals in the video recording process, if the position coordinates between adjacent time axis mark points change obviously, namely the distance exceeds a preset value, the variation tracking points are embedded at the time points; And S14, fusing the time sequence position group with the cargo picture and the text label, and expanding the binding range to cover inter-element variation to obtain a multimedia position set.
  3. 3. The intelligent cargo multimedia information extraction method according to claim 2, wherein step S2 includes: S21, selecting a cargo picture from a multimedia position set, performing content layered scanning to separate static visual elements, and splitting contour textures and background components to obtain a picture layered layer; S22, applying dynamic sequence segmentation to the video segments in the corresponding multimedia position set, capturing a motion track, layering dynamic elements, and connecting sequences through inter-frame transition marks to obtain video layering; Step S23, element alignment is carried out on the picture layering layer and the video layering layer, cross-layer connection is constructed, alignment anchor points are set, and an alignment layering layer is obtained; S24, binding and associating the alignment layering layer with the position coordinates, and expanding and connecting the alignment layering layer to the time sequence position group element to obtain an expanded layering layer; and S25, performing element clustering circulation corresponding to the expansion layering layer, combining the visual elements and attaching clustering labels to obtain a layering content group.
  4. 4. The intelligent cargo multimedia information extraction method according to claim 3, wherein step S4 comprises: s41, starting attribute extraction circulation from an integrated semantic chain, and scanning cargo specification details to split size materials and a plurality of components to obtain a specification fragment group; s42, extracting the extended transportation constraint fragments of the specification fragment group, and fusing ageing loading and unloading and path marks to construct a constraint sub-chain to obtain a constraint extended group; s43, fusing the specification fragment group and the constraint expansion group, establishing a link between attributes and setting a fusion node to obtain a fusion attribute group; step S44, performing attribute priority ranking corresponding to the fusion attribute group, adjusting fragment weights and attaching ranking anchors to obtain a ranking attribute group; Step S45, binding the ordering attribute group to the tail end of the integrated semantic chain, and expanding the ordering attribute group to position coordinate binding to obtain an expanded attribute group; Step S46, applying attribute verification cycle to the extended attribute group, checking the link consistency and attaching a verification tag to obtain a structured attribute set.
  5. 5. The intelligent cargo multimedia information extraction method according to claim 4, wherein step S5 comprises: S51, extracting cargo specification details and transportation constraint fragments from the structured attribute set, binding the cargo specification details and the transportation constraint fragments with position coordinates to generate an initial driver query sequence, and attaching attribute tags to obtain a sequence draft; step S52, merging path forward factors into the sequence draft, simulating a driver track, embedding forward matching points, expanding track branches, and obtaining factor enhancement sequences; And step S53, performing adaptive sorting based on the factor enhancement sequence, calculating the path overlapping degree, adjusting the ranking, and adding a sorting mark to obtain a sorting sequence group.
  6. 6. The intelligent cargo multimedia information extraction method according to claim 5, wherein step S52 comprises: Step S521, selecting a query sequence from a sequence draft, introducing a path-following factor as a track simulation input, and constructing a simulation sub-path to obtain a simulation track group; Step S522, embedding forward matching points into the simulation track group, binding a dynamic expansion factor range based on position coordinates, and applying matching point diffusion to obtain an expansion track group; Step S523, performing branch integration on tracks in the extended track group, setting integration nodes and recording branch variation to obtain an integrated track group; Step S524, executing factor strengthening cycle corresponding to the integrated track group, adjusting the forward factor weight and attaching strengthening labels to obtain a strengthening track group; step S525, converting the enhanced track group into a factor enhanced sequence, and binding the factor enhanced sequence to the structured attribute set element.
  7. 7. An intelligent cargo multimedia information extraction system for implementing the intelligent cargo multimedia information extraction method according to any one of claims 1 to 6, comprising: The positioning module is used for collecting cargo pictures, video clips and text labels uploaded by a cargo owner and generating position coordinate binding by combining equipment positioning signals to obtain a multimedia position set; the content analysis module is used for executing content layering scanning corresponding to the pictures and the video clips in the media position set, separating visual elements and dynamic sequences to form a hierarchical structure, and obtaining a layered content group; The semantic bridging module is used for carrying out cross semantic bridging on the layered content group and the text labels to construct an inter-element association path and obtain an integrated semantic chain, and comprises the steps of extracting visual elements from the layered content group, carrying out semantic bridging on the visual elements and the text labels to generate a preliminary association path, and adding bridging nodes to obtain a bridging path set, wherein the semantic bridging adopts a cross-modal similarity calculation method to calculate cosine similarity between visual element semantic vectors and text entity semantic vectors, when the similarity exceeds a preset threshold value, the system judges that semantic association exists, generates a preliminary association path, adds bridging nodes at midpoint positions of the paths, and records visual element numbers, text entity contents and similarity values; The method comprises the steps of selecting a path from a bridging path set, introducing text labels as reference anchors, executing semantic consistency check, and recording check nodes to obtain a check path group; Setting bridging intensity values for the inspection path group, dynamically adjusting intensity based on inter-element distance and semantic overlapping, and applying an intensity propagation mechanism to obtain an intensity enhanced path set, wherein the intensity propagation mechanism is used for transmitting the credibility of a high-intensity bridging path to adjacent low-intensity paths, and the propagation rule is that if two paths share the same visual element or text entity, the intensity values of the low-intensity paths are according to the formula Enhancement is performed in which In order for the path enhancement value to be a value, In order to bridge the intensity value of the light, Intensity values for adjacent high intensity paths; carrying out branch scanning on paths in the strength enhancement path set, adding variation compensation and converting the variation compensation into a verification path set; Performing path compression circulation on the verification path set, merging adjacent nodes and attaching compression labels; Constructing an integrated semantic chain based on the verification path set, expanding the path to cover the multi-layer element association and embedding the multi-layer element association into an intra-chain fulcrum; the attribute extraction module is used for applying attribute extraction circulation to the integrated semantic chain, extracting cargo specification details and transportation constraint fragments to construct an attribute set, and obtaining a structured attribute set; And the adaptation recommendation module is used for generating a driver query sequence based on the binding of the structured attribute set and the position coordinates, and merging the path forward factors to perform adaptation sequencing to obtain a preferred driver list.

Description

Cargo multimedia information intelligent extraction system and method Technical Field The invention relates to the technical field of intelligent logistics, in particular to an intelligent goods multimedia information extraction system and method. Background In modern logistics service, accurate collection and efficient transfer of cargo information have key effects on realizing accurate matching of supply and demand and optimizing transportation resource allocation, and the current mainstream logistics platform generally supports the cargo owner to upload related data of the cargo in a multimedia form such as shooting photos, recording videos, editing text remarks and the like, and the multimedia data can intuitively record information such as appearance form, stacking state, packaging form, site environment and the like of the cargo, so that an original basis is provided for the subsequent scheduling and transportation scheme formulation. The conventional logistics information processing technology has remarkable limitation in the aspect of automatically converting the multimedia content uploaded by a cargo owner into structured freight data which can be used by intelligent matching, the problem is particularly remarkable in an instant freight matching scene which needs quick response, particularly, when the cargo owner uploads pictures, videos and text descriptions containing freight information simultaneously, the data from different modes have remarkable differences in expression form, information granularity and semantic hierarchy, a system is difficult to automatically identify and associate content fragments which describe the same freight attribute in each mode data, the key information such as freight types, specification parameters, quantity characteristics, loading and unloading requirements, transportation timeliness constraints and the like which are scattered in different media carriers cannot be effectively integrated and structurally converted, the conventional system can only independently store and simply display various multimedia files, rely on manual review and then perform information aggregation and secondary input, the processing mode not only causes a large amount of repeated labor and response delay, but also causes the omission or misjudgment of key freight characteristics due to subjective differences of manual understanding, and the fact that the key freight characteristics are not suitable for a driver is due to the fact that the key freight characteristics are automatically extracted from the multimedia content, the key information is difficult to automatically extract the key information fragments which are distributed in the multimedia content, the condition that the key information is based on the characteristics of the physical freight characteristics can not be matched with the actual position of the information carrier, and the actual position of the cargo can not be matched with the actual carrier, and the actual position of the intelligent logistics information can not be easily achieved, and the capability of matching can not be matched with the actual performance of the freight platform. In view of the above, the present invention provides a system and a method for intelligently extracting multimedia information of goods to solve the above problems. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides the following technical scheme that the intelligent extracting method for the goods multimedia information comprises the following steps: Step S1, acquiring cargo pictures, video clips and text labels uploaded by a cargo owner, and generating position coordinate binding by combining equipment positioning signals to obtain a multimedia position set; step S2, performing content hierarchical scanning corresponding to the pictures and video clips in the media position set, and separating visual elements and dynamic sequences to form a hierarchical structure so as to obtain a hierarchical content group; S3, performing cross semantic bridging on the layered content groups and the text labels, and constructing inter-element association paths to obtain an integrated semantic chain; S4, applying attribute extraction circulation to the integrated semantic chain, and extracting cargo specification details and transportation constraint fragments to construct an attribute set to obtain a structured attribute set; And S5, generating a driver query sequence based on the binding of the structured attribute set and the position coordinates, and merging the path forward factors to carry out adaptive sorting to obtain a preferred driver list. An intelligent extracting system for goods multimedia information, comprising: The positioning module is used for collecting cargo pictures, video clips and text labels uploaded by a cargo owner and generating position coordinate binding by combining equipment positioning signals to obtain a multimedia po