Electric power material identification method and system based on side cloud collaboration

Abstract

The invention discloses an electric power material identification method and system based on side cloud cooperation. The method comprises the following steps: acquiring initial video data acquired by side cloud interaction intelligent equipment; analyzing the initial video data based on the selected visual language model, and extracting continuous video section data; controlling the cloud platform to obtain semantic information data; inputting the semantic information data into a selected target detection model for processing to obtain video frame data; detecting the video frame data based on a logic processing algorithm, and processing a detection result according to the image and video segmentation model to obtain target video frame data; and processing the target video data generated by reconstruction to obtain an identification result of the electric power materials. According to the electric power material identification method and system based on edge cloud collaboration provided by the invention, the edge side and the cloud layer in edge interaction are respectively improved, so that identification and management of electric power material information data are effectively improved, and the intelligent process of electric power warehouse management is promoted.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a method and system for identifying electric power materials based on edge-cloud collaboration. Background Art

[0002] With the construction of a green, modern, digital, and intelligent supply chain, refined management of power materials is gradually becoming a basic requirement for power warehouse management.

[0003] In the existing technology, RFID and other technologies are mainly used to improve the management ability of power materials. However, the information of power materials is essentially a multimodal information collection. Traditional technologies are difficult to meet the needs of multimodal data processing, resulting in poor identification and management effects of power materials.

[0004] It can be seen that how to improve the ability to identify and manage power materials in power warehouses has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides a method and system for identifying electric power materials based on edge-cloud collaboration, which effectively improves the recognition and management of electric power material information data and promotes the intelligent process of electric power warehouse management by respectively improving the edge side and cloud layer in edge interaction.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method and system for identifying electric materials based on edge-cloud collaboration, which is applied to a power warehouse where a number of edge-cloud interactive intelligent devices are deployed. The method for identifying electric materials based on edge-cloud collaboration includes:

[0007] Collecting initial video data acquired by the edge-cloud interactive intelligent device;

[0008] Analyzing the initial video data based on the selected visual language model, and extracting continuous video segment data containing power material displacement information from the analysis results;

[0009] Control the cloud platform to decompose the received continuous video segment data, and input the decomposition results into the selected image semantic extraction model to obtain corresponding semantic information data;

[0010] Input each of the semantic information data into the selected target detection model for processing to obtain corresponding video frame data;

[0011] The video frame data is detected based on a logic processing algorithm, and the detection result is processed according to a selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to mark the change in the number of power material identification boxes between adjacent video frame data identified by the mask layer;

[0012] The target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing to obtain the recognition result of the power materials in the power warehouse.

[0013] As one of the preferred solutions, the visual language model is the Qwen2-VL-2B model;

[0014] The analyzing the initial video data based on the selected visual language model comprises:

[0015] Processing the initial video data based on a preset video cutting algorithm to obtain a first processing result;

[0016] Controlling the Qwen2-VL-2B model to screen the power material displacement information of the first processing result to obtain a second screening result;

[0017] The second screening results are combined and packaged to obtain the continuous video section data.

[0018] As one of the preferred solutions, before inputting each of the semantic information data into the selected target detection model for processing, the power material identification method based on edge-cloud collaboration further includes:

[0019] De-redundancy, screening and aggregation processing are performed on each of the semantic information data.

[0020] As one of the preferred solutions, after the target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing, the power material identification method based on edge-cloud collaboration further includes:

[0021] The electric power materials in the electric power warehouse are analyzed according to the target video data, the acquired ELP distribution information data and the electric power material graphic list data, and the recognition result is synchronously updated based on the analysis result.

[0022] As one of the preferred solutions, after the target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing, the power material identification method based on edge-cloud collaboration further includes:

[0023] Based on the selected defect recognition model, defects of the power materials in the power warehouse are identified, and the defect recognition results are synchronously updated to the recognition results.

[0024] Another embodiment of the present invention provides an edge-cloud collaboration-based power material identification system, which is applied to a power warehouse where a number of edge-cloud interactive intelligent devices are deployed, including:

[0025] A collection module, used to collect the initial video data acquired by the edge-cloud interactive intelligent device;

[0026] An extraction module, configured to analyze the initial video data based on a selected visual language model, and extract continuous video segment data containing power material displacement information from the analysis results;

[0027] A semantic module is used to control the cloud platform to decompose the received continuous video segment data, and input the decomposition results into the selected image semantic extraction model to obtain corresponding semantic information data;

[0028] A video frame module, used for inputting each of the semantic information data into a selected target detection model for processing to obtain corresponding video frame data;

[0029] A detection module, configured to detect the video frame data based on a logic processing algorithm, and process the detection results according to a selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to detect the change in the number of power material identification boxes between adjacent video frame data identified by the mask layer mark;

[0030] The recognition module is used to input the target video data reconstructed from each of the target video frame data into the edge-cloud interactive intelligent device for processing to obtain the recognition result of the power materials in the power warehouse.

[0031] As one of the preferred solutions, the visual language model is the Qwen2-VL-2B model;

[0032] The extraction module comprises:

[0033] A cutting unit, configured to process the initial video data based on a preset video cutting algorithm to obtain a first processing result;

[0034] A screening unit, used for controlling the Qwen2-VL-2B model to screen the first processing result for power material displacement information to obtain a second screening result;

[0035] A packaging unit is used to combine and package the second screening results to obtain the continuous video section data.

[0036] As one of the preferred solutions, the edge-cloud collaborative power material identification system also includes:

[0037] The aggregation module is used to perform redundancy removal, screening and aggregation processing on each of the semantic information data.

[0038] As one of the preferred solutions, the edge-cloud collaborative power material identification system also includes:

[0039] The first synchronous updating module is used to analyze the power materials in the power warehouse according to the target video data, the acquired ELP distribution information data and the power material graphic list data, and synchronously update the recognition result based on the analysis result.

[0040] As one of the preferred solutions, the edge-cloud collaborative power material identification system also includes:

[0041] The second synchronous updating module is used to identify defects of the power materials in the power warehouse based on the selected defect identification model, and synchronously update the defect identification results to the identification results.

[0042] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0043] Based on edge-cloud collaborative interaction technology, the edge-cloud collaborative big model application architecture is adopted, which can greatly improve the intelligent identification ability of power materials and promote the intelligent management of power warehouses; the entire identification method adopts the edge-cloud collaborative big model application architecture, fully utilizing the computing resources of the edge and cloud, improving the processing efficiency and scalability of the system; introducing advanced big model technology, significantly improving the accuracy of material identification and analysis; through cloud model upgrades and automatic updates of edge models, the continuous improvement of system capabilities can be achieved, thereby improving the identification and management effects of power materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flowchart of a method for identifying electric power materials based on edge-cloud collaboration in one embodiment of the present invention;

[0045] Figure 2 It is a schematic diagram of a power material identification framework based on edge-cloud collaboration in one embodiment of the present invention;

[0046] Figure 3 It is a structural diagram of an electric power material identification system based on edge-cloud collaboration in one embodiment of the present invention;

[0047] Reference numerals:

[0048] Among them, 11, acquisition module; 12, extraction module; 13, semantic module; 14, video frame module; 15, detection module; 16, recognition module. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0051] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0052] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0053] It should be noted in advance that existing technologies have promoted technologies such as RFID to improve the management capabilities of power materials and optimize the consistency rate of account cards and materials. However, RFID is difficult to fundamentally solve the problem of account card correspondence, because it is very dependent on the correct setting of tags by staff. With the rapid development of artificial intelligence technology, machine vision technology has provided a new solution for improving the consistency rate of account cards and materials. Through "seeing is believing", it can effectively make up for the application limitations of RFID technology. However, considering that the information of account cards and materials is essentially a combination of multimodal information, traditional visual technology (such as target recognition based on the YOLO series of algorithms, etc.) is difficult to fully meet the actual application needs. In addition, there are currently more than 4,000 types of power materials, with many sizes and different models, and the traditional closed set detection model can no longer adapt.

[0054] In view of this, an embodiment of the present invention proposes a large and small model collaborative working framework for power material management and control, which introduces large models, especially multimodal large models, into the field of power material management and control, and provides a reliable and flexible solution for improving the consistency rate of accounts, cards and materials of power materials. Specifically, an embodiment of the present invention provides a power material identification method based on edge-cloud collaboration. For details, please refer to Figure 1 , Figure 1 The figure shows a flow chart of a method for identifying electric power materials based on edge-cloud collaboration in one embodiment of the present invention, which specifically includes steps S1 to S6:

[0055] S1: Collecting initial video data acquired by the edge-cloud interactive intelligent device;

[0056] S2: analyzing the initial video data based on the selected visual language model, and extracting continuous video segment data containing power material displacement information from the analysis results;

[0057] S3: Control the cloud platform to decompose the received continuous video segment data, and input the decomposition result into the selected image semantic extraction model to obtain corresponding semantic information data;

[0058] S4: Inputting each of the semantic information data into the selected object detection model for processing to obtain corresponding video frame data;

[0059] S5: Detecting the video frame data based on a logic processing algorithm, and processing the detection results according to the selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to mark the change in the number of power material identification boxes between adjacent video frame data identified by the mask layer;

[0060] S6: Input the target video data reconstructed from each of the target video frame data into the edge-cloud interactive intelligent device for processing to obtain the recognition result of the power materials in the power warehouse.

[0061] For easier understanding, see Figure 2 , Figure 2 A schematic diagram of an edge-cloud collaboration-based power material identification framework in one of the embodiments of the present invention is shown. Improvements are made to the edge side and the cloud side respectively, and a novel large and small model edge-cloud collaborative working architecture is used to achieve effective utilization of multi-modal information of multiple power materials, thereby supporting effective management and control of materials.

[0062] like Figure 2 As shown, preferably, the framework side will be equipped with a quantized Qwen2-VL-2B model. Qwen2-VL-2B is a multimodal large model for local deployment on mobile devices such as mobile phones. It has the advantage of being lightweight. Since the Qwen2-VL-2B model is only responsible for some simple tasks in this framework and is assisted by a retrieval enhancement module, the quantized version can be directly deployed to reduce computing power overhead on the edge side. In order to facilitate the update and capability enhancement of the edge side model and adapt to the expansion of new materials in the future, the embodiment of the present invention uses LongChain or Llamaindex to introduce graphic information of electric power materials, including electric power material appearance image-text pairs (text includes electric power material names, electric power material classifications, etc.), State Grid Electric Power Logistics Service Platform (ELP) distribution information, etc., so that it can better adapt to the identification needs of various types of electric power materials. In addition, in order to meet the detection needs of material defects, a YOLO defect recognition model can be deployed on the side.

[0063] The computing resources available on the cloud side of the framework are generally more abundant, so large visual models such as RAM++, GroundingDINO, and SAM2 are set up here to collaboratively complete tasks such as visual inventory tracking of materials. It should be noted that in order to achieve the best open set detection performance to match the identification requirements of diverse and diverse power materials, the embodiment of the present invention uses SAM2-Adapter for SAM2. The large plural visual models on the cloud side mainly complete the functions according to the call instructions of the Qwen2-VL-2B model on the edge side. Its core model is SAM2, which mainly realizes the visual tracking of power material targets.

[0064] From the overall perspective of edge-cloud collaboration, the Qwen2-VL-2B multimodal large model, which is the core of the entire framework, is deployed on the edge, which is different from the general deployment method. This is done to give full play to its capabilities and efficiency advantages, maximize the intelligent capabilities of each warehouse, and maintain good scalability to support the organic integration and connection of multiple intelligent applications.

[0065] Specifically, the process of the power material identification method based on edge-cloud collaboration in the embodiment of the present invention is as follows:

[0066] 1. The edge-side intelligent device in the power warehouse (i.e., edge-cloud interactive intelligent device) cuts the real-time monitoring video into several 5-second video segments, and sends them to the visual language model for analysis. Of course, the specific cutting method can adopt the existing cutting algorithm, or the cutting algorithm can be built into the visual language model as a functional module, which is not specifically limited in the embodiments of the present invention. The present invention is preferably a Qwen2-VL-2B model. Qwen2-VL-2B will screen out continuous video segments containing material displacement behaviors, combine and package them and upload them to the cloud. At the same time, Qwen2-VL-2B will extract relevant material multimodal data pairs, recent ELP distribution information, etc. through retrieval enhancement based on LongChain or Llamaindex.

[0067] 2. After the video segment is uploaded to the cloud, it will first be decomposed into video frames, where the first frame of each video segment will be sent to the selected image semantic extraction model for processing. The embodiment of the present invention is preferably RAM++ to complete the extraction of semantics.

[0068] 3. Since the semantics extracted by RAM++ contains redundant information, it is necessary to deduplicate, filter and aggregate each of the semantic information data separately. Qwen2-VL-2B completes the filtering and aggregation to adapt to and match the subsequent target detection. The embodiment of the present invention preferably detects labels through Grounding DINO.

[0069] Fourth, the filtered and aggregated semantics will be sent to Grounding DINO to guide it to complete the target detection for each video frame. The logical processing algorithm is used to monitor the change in the number of recognition boxes between every two frames. When the number of recognition boxes in the nth frame is more than that in the (n-1)th frame, the sequence number of the nth frame is recorded.

[0070] 5. Taking each frame with an increased number of recognition frames as the starting point, the selected image and video segmentation model is used for processing. The preferred embodiment of the present invention is SAM2. SAM2 completes the generation of the mask based on the recognition frame information and propagates it to subsequent frames. In this way, all materials passing through the camera's field of view will be segmented and marked. At this point, the video frames can be reassembled into a video, which is the reconstruction link.

[0071] 6. Return the segmented and labeled video to the edge-side smart device. The video will contain a mask, which can be analyzed to obtain the identification results of the power materials in the power warehouse.

[0072] 7. Enable the Qwen2-VL-2B model to combine ELP distribution information, material graphic data pairs, segmented and annotated video and defect recognition model recognition results, and the positional relationship between the camera and each key point in the warehouse to perform reasoning, derive the type, quantity, behavior and other information of the materials, and update the recognition results synchronously with the derivation results, so as to obtain more detailed power material information and improve the macro-coordination and control capabilities of power warehouses. In addition, defects in power materials can also be identified to provide accurate data support for subsequent information aggregation and collection. Preferably, based on the selected defect recognition model, defects in power materials in the power warehouse are identified, and the defect recognition results are synchronously updated to the recognition results.

[0073] It should be noted that since most of the time, the camera's field of view does not involve the movement of materials, the timeliness of the video processing by each model within the framework can be guaranteed. Similarly, multiple warehouses can share the cloud-based visual processing model group, and the collaboration between warehouses is also achieved through the Qwen2-VL-2B model.

[0074] In addition, the fine-tuning and updating of large models has certain requirements for computing power. If training equipment with high computing power is set up in each warehouse, it will not only be costly but also cause waste of resources. The present invention is to set up intelligent devices capable of edge-cloud interaction in the power warehouse. In addition to being able to better complete the identification and inventory of materials, the proposed framework also has the characteristic of convenient upgrading. The framework can support the sinking of new knowledge on the cloud side through the organic collaborative operation of the edge and the cloud. Therefore, the operation and maintenance upgrade of the framework is simplified to the upgrade work on the cloud. By fine-tuning the RAM++ model on the cloud side and correspondingly modifying the knowledge base and prompt engineering of the Qwen2-VL-2B model on the edge side, the framework can obtain the ability to accurately identify new types of power materials.

[0075] To summarize, the present invention mainly designs an edge-cloud collaborative framework that links the Qianwen 2 (Qwen2-VL-2B) visual language model, the Segment-Anything-Model-2 (SAM2) image and video segmentation model, the Recognize-Anything-Model-Plus (RAM++) image semantic extraction model, the Grounding DINO target detection model, the YOLO model and other models, and proposes a corresponding working operation mode, which can support the effective understanding and application of multimodal material information, and realize the identification and inventory control of power materials.

[0076] For details, see Figure 3 , Figure 3A schematic diagram of the structure of an electric material identification system based on edge-cloud collaboration provided by another embodiment of the present invention is shown, which is applied to an electric warehouse where a number of edge-cloud interactive intelligent devices are deployed, and specifically includes:

[0077] The acquisition module 11 is used to acquire the initial video data acquired by the edge-cloud interactive intelligent device;

[0078] An extraction module 12, configured to analyze the initial video data based on a selected visual language model, and extract continuous video segment data containing power material displacement information from the analysis results;

[0079] The semantic module 13 is used to control the cloud platform to decompose the received continuous video segment data, and input the decomposition results into the selected image semantic extraction model to obtain corresponding semantic information data;

[0080] The video frame module 14 is used to input each of the semantic information data into the selected target detection model for processing to obtain corresponding video frame data;

[0081] A detection module 15, configured to detect the video frame data based on a logic processing algorithm, and process the detection result according to a selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to change the number of power material identification frames between adjacent video frame data identified by the mask layer mark;

[0082] The identification module 16 is used to input the target video data reconstructed from each of the target video frame data into the edge-cloud interactive intelligent device for processing to obtain the identification result of the power materials in the power warehouse.

[0083] Further, in the above embodiment, the visual language model is the Qwen2-VL-2B model;

[0084] The extraction module 12 comprises:

[0085] A cutting unit, configured to process the initial video data based on a preset video cutting algorithm to obtain a first processing result;

[0086] A screening unit, used for controlling the Qwen2-VL-2B model to screen the first processing result for power material displacement information to obtain a second screening result;

[0087] A packaging unit is used to combine and package the second screening results to obtain the continuous video section data.

[0088] Furthermore, in the above embodiment, the edge-cloud collaborative power material identification system further includes:

[0089] The aggregation module is used to perform redundancy removal, screening and aggregation processing on each of the semantic information data.

[0090] Furthermore, in the above embodiment, the edge-cloud collaborative power material identification system further includes:

[0091] The first synchronous updating module is used to analyze the power materials in the power warehouse according to the target video data, the acquired ELP distribution information data and the power material graphic list data, and synchronously update the recognition result based on the analysis result.

[0092] Furthermore, in the above embodiment, the edge-cloud collaborative power material identification system further includes:

[0093] The second synchronous updating module is used to identify defects of the power materials in the power warehouse based on the selected defect identification model, and synchronously update the defect identification results to the identification results.

[0094] The power material identification method or system based on edge-cloud collaboration provided by an embodiment of the present invention has the beneficial effects of at least one of the following:

[0095] (1) Based on edge-cloud collaborative interaction technology, a large model application architecture of edge-cloud collaboration is adopted, which can greatly improve the ability to identify power materials and promote the intelligent management of power warehouses. The entire identification method adopts a large model application architecture of edge-cloud collaboration, making full use of the computing resources of the edge and cloud, improving the processing efficiency and scalability of the system. Advanced large model technology is introduced to significantly improve the accuracy of material identification and analysis. Through cloud model upgrades and automatic updates of edge models, the system capabilities can be continuously improved, thereby improving the identification and management effects of power materials.

[0096] (2) By deploying the quantized Qwen2-VL-2B multimodal large model on the edge and visual large models such as RAM++, Grounding DINO, and SAM2 on the cloud, an edge-cloud collaborative intelligent material management framework is implemented. This method uses the Qwen2-VL-2B model for video analysis and screening, combines the cloud visual model to complete material identification, segmentation, and tracking, and infers the type, quantity, and behavior of materials through multi-source information fusion.

[0097] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for identifying electric power materials based on edge-cloud collaboration, characterized in that: Applied to a power warehouse where a number of edge-cloud interactive intelligent devices are deployed, the power material identification method based on edge-cloud collaboration includes: Collecting initial video data acquired by the edge-cloud interactive intelligent device; Analyzing the initial video data based on the selected visual language model, and extracting continuous video segment data containing power material displacement information from the analysis results; Control the cloud platform to decompose the received continuous video segment data, and input the decomposition results into the selected image semantic extraction model to obtain corresponding semantic information data; Input each of the semantic information data into the selected target detection model for processing to obtain corresponding video frame data; The video frame data is detected based on a logic processing algorithm, and the detection result is processed according to a selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to mark the change in the number of power material identification boxes between adjacent video frame data identified by the mask layer; The target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing to obtain the recognition result of the power materials in the power warehouse.

2. The power material identification method based on edge-cloud collaboration as claimed in claim 1 is characterized in that: The visual language model is the Qwen2-VL-2B model; The analyzing the initial video data based on the selected visual language model comprises: Processing the initial video data based on a preset video cutting algorithm to obtain a first processing result; Controlling the Qwen2-VL-2B model to screen the power material displacement information of the first processing result to obtain a second screening result; The second screening results are combined and packaged to obtain the continuous video section data.

3. The power material identification method based on edge-cloud collaboration as claimed in claim 1 is characterized in that: Before inputting each of the semantic information data into the selected target detection model for processing, the power material identification method based on edge-cloud collaboration also includes: De-redundancy, screening and aggregation processing are performed on each of the semantic information data.

4. The power material identification method based on edge-cloud collaboration as claimed in claim 1 is characterized in that: After the target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing, the power material identification method based on edge-cloud collaboration further includes: The electric power materials in the electric power warehouse are analyzed according to the target video data, the acquired ELP distribution information data and the electric power material graphic list data, and the recognition result is synchronously updated based on the analysis result.

5. The power material identification method based on edge-cloud collaboration as claimed in claim 1 is characterized in that: After the target video data reconstructed from each of the target video frame data is input into the edge-cloud interactive intelligent device for processing, the power material identification method based on edge-cloud collaboration further includes: Based on the selected defect recognition model, defects of the power materials in the power warehouse are identified, and the defect recognition results are synchronously updated to the recognition results.

6. An edge-cloud collaborative power material identification system, characterized in that: Applied to power warehouses where a number of edge-cloud interactive intelligent devices are deployed, including: A collection module, used to collect the initial video data acquired by the edge-cloud interactive intelligent device; An extraction module, configured to analyze the initial video data based on a selected visual language model, and extract continuous video segment data containing power material displacement information from the analysis results; A semantic module is used to control the cloud platform to decompose the received continuous video segment data, and input the decomposition results into the selected image semantic extraction model to obtain corresponding semantic information data; A video frame module, used for inputting each of the semantic information data into a selected target detection model for processing to obtain corresponding video frame data; A detection module, configured to detect the video frame data based on a logic processing algorithm, and process the detection results according to a selected image and video segmentation model to obtain each target video frame data with a mask layer, wherein the detection process is configured to detect the change in the number of power material identification boxes between adjacent video frame data identified by the mask layer mark; The recognition module is used to input the target video data reconstructed from each of the target video frame data into the edge-cloud interactive intelligent device for processing to obtain the recognition result of the power materials in the power warehouse.

7. The power material identification system based on edge-cloud collaboration as claimed in claim 6 is characterized in that: The visual language model is the Qwen2-VL-2B model; The extraction module comprises: A cutting unit, configured to process the initial video data based on a preset video cutting algorithm to obtain a first processing result; A screening unit, used for controlling the Qwen2-VL-2B model to screen the first processing result for power material displacement information to obtain a second screening result; A packaging unit is used to combine and package the second screening results to obtain the continuous video section data.

8. The power material identification system based on edge-cloud collaboration as claimed in claim 6 is characterized in that: The edge-cloud collaborative power material identification system also includes: The aggregation module is used to perform redundancy removal, screening and aggregation processing on each of the semantic information data.

9. The power material identification system based on edge-cloud collaboration as claimed in claim 6 is characterized in that: The edge-cloud collaborative power material identification system also includes: The first synchronous updating module is used to analyze the power materials in the power warehouse according to the target video data, the acquired ELP distribution information data and the power material graphic list data, and synchronously update the recognition result based on the analysis result.

10. The power material identification system based on edge-cloud collaboration as claimed in claim 6, characterized in that: The edge-cloud collaborative power material identification system also includes: The second synchronous updating module is used to identify defects of the power materials in the power warehouse based on the selected defect identification model, and synchronously update the defect identification results to the identification results.

Images