Search

CN-121998943-A - Air-ground collaborative inspection method based on heterogeneous platform

CN121998943ACN 121998943 ACN121998943 ACN 121998943ACN-121998943-A

Abstract

The invention discloses an air-ground collaborative inspection method based on a heterogeneous platform, belongs to the technical field of intelligent inspection, and is suitable for multi-scene inspection of cities, electric power, maritime waters and the like. Aiming at the problems of visual angle limitation, rechecking missing, inaccurate positioning, lack of self-evolution capability and the like of the traditional single-platform inspection, the method constructs a full-link mechanism of discovery, induction, rechecking, confirmation, learning and optimization through the functional complementation of air and ground equipment. The core steps include that the front end collects original data such as defect images, category labels and the like, an industry term knowledge base is called, a structured document is formed through a template engine, a patrol report with a traceability mark is output, and reliability is guaranteed by assistance of blockchain evidence storage. Meanwhile, a heterogeneous node coordination mechanism and a three-dimensional coordinate conversion model are established, cross-platform accurate coordination is achieved, closed loop rechecking and self-evolution optimization are combined, inspection efficiency and precision are considered, false alarms are reduced, and inspection intellectualization and reliability are remarkably improved.

Inventors

  • LI YONGJUN
  • Chen Shutu
  • XU MIAO
  • ZHANG DAN
  • WANG LIN

Assignees

  • 江苏欣网视讯软件技术有限公司
  • 南京欣网飞联无人机科技有限公司

Dates

Publication Date
20260508
Application Date
20260127

Claims (9)

  1. 1. The air-ground collaborative inspection method based on the heterogeneous platform is characterized by comprising the following steps of: the method comprises the steps that first, patrol original data are obtained through a front-end acquisition terminal, wherein the patrol original data at least comprise a defect target image, a defect type label, a confidence coefficient and corresponding space-time metadata; step two, constructing and maintaining an industry term knowledge base which comprises a defect category-industry term mapping table, a defect level evaluation rule, a causal relationship model and a standard report paragraph template and is stored in the form of an extensible markup language or graph database; step three, taking the defect type label as input, carrying out semantic retrieval and alignment in the industry term knowledge base through a natural language processing engine to generate a standardized description text which accords with industry specifications, and automatically determining defect levels according to defect level evaluation rules; Step four, dynamically filling the standardized description text, the defect level, the space-time metadata and the associated images into corresponding placeholders in a template based on a preset report template engine to form a structured intermediate document; And fifthly, outputting the structured intermediate documents in batches into Word and/or PDF inspection reports meeting the format requirements through a document conversion interface, and supporting automatic addition of digital signatures, two-dimensional codes and watermarks so as to finish traceable release of the reports.
  2. 2. The air-ground collaborative inspection method based on a heterogeneous platform according to claim 1, wherein the industry term knowledge base further comprises a synonym dictionary and a domain ontology to support normalization and context concept reasoning of synonymous defect descriptions.
  3. 3. The air-ground collaborative inspection method based on the heterogeneous platform according to claim 1 or 2, wherein the natural language processing engine adopts a pre-training language model based on a transducer and performs fine adjustment through a domain corpus so as to improve the accuracy of generating the professional terms.
  4. 4. A space-to-ground collaborative inspection method based on a heterogeneous platform according to any one of claims 1-3, wherein in step four, when multiple defect records exist in the same device, the template engine automatically merges similar defects and generates summary paragraphs, avoiding duplicate description.
  5. 5. The air-ground collaborative inspection method based on the heterogeneous platform according to any one of claims 1-4, wherein the report template engine supports user-defined templates and realizes template iteration and backtracking through a version control mechanism.
  6. 6. The air-ground collaborative inspection method based on a heterogeneous platform according to any one of claims 1-5, wherein after the fifth step, the generated inspection report hash value is further written into a blockchain memory certificate to realize report integrity and non-repudiation.
  7. 7. The air-ground collaborative inspection method based on the heterogeneous platform according to any one of claims 1-6 is characterized by further comprising a heterogeneous node collaborative mechanism construction step of establishing a coordinate sharing protocol, a clock synchronization mechanism and a standardized task transfer interface among heterogeneous nodes to realize efficient collaborative scheduling of multiple types of inspection nodes.
  8. 8. The air-ground collaborative inspection method based on the heterogeneous platform according to any one of claims 1 to 7, further comprising a three-dimensional coordinate conversion step of pixel-level detection results, wherein a mapping model of pixel coordinates and three-dimensional space coordinates is built based on real-time pose data of inspection nodes and sensor internal and external parameters, and accurate inverse solution from the pixel-level detection results to cross-platform sharing of three-dimensional target coordinates is achieved.
  9. 9. The air-ground collaborative inspection method based on the heterogeneous platform according to any one of claims 1-8, further comprising a system self-evolution optimization step of dynamically optimizing a front end detection threshold and an inspection path by the platform according to a historical inspection result, and reducing the repetition of similar false alarms.

Description

Air-ground collaborative inspection method based on heterogeneous platform Technical Field The invention relates to the technical field of intelligent patrol, in particular to an air-ground collaborative patrol method based on a heterogeneous platform. Background With the development of unmanned system technology, unmanned inspection has been widely used in various scenes. The conventional unmanned inspection system generally relies on a single platform to independently execute tasks, but in practical application, the system has a plurality of defects: 1. The single visual angle limitation is that the confidence coefficient of part of suspected targets is low under the conditions of long distance, complex terrain and shielding, false detection or omission is easy to generate, and the real state of the inspection target is difficult to comprehensively and accurately identify. 2. The cross-platform rechecking is impossible, namely the data shot by a single platform lacks high-resolution compensation or verification by a heterogeneous sensor, and a closed-loop evidence chain is difficult to form, so that the reliability of the inspection result is insufficient. 3. The positioning accuracy is insufficient, the single-node positioning depends on the onboard camera, the coordinate of the same target is often larger in error, and the requirements of accurate inspection and subsequent processing cannot be met. 4. The system lacks learning ability, the routing inspection strategy is difficult to dynamically optimize, and the problems of repeated false alarms, low-efficiency path planning and the like exist, so that the routing inspection efficiency is low and the intelligent level is not high. The current air-ground inspection system also faces the problem of difficult coordination of heterogeneous nodes, air and ground equipment independently run, a standardized task triggering and information transmission mechanism is lacked, and preliminary screening-rechecking linkage is difficult to realize. Meanwhile, the existing unmanned plane and the ground robot lack unified data formats, space-time references and task interfaces, so that cross-platform collaborative detection cannot be realized. The traditional detection only outputs the target pixel position, the heterogeneous node rechecking requirement cannot be met, and the system has no feedback mechanism to optimize the front end detection threshold value and the inspection path, so that similar false alarms occur repeatedly, and the application effect of the inspection system is further limited. Disclosure of Invention The invention provides a space-ground collaborative inspection method based on a heterogeneous platform, which is used for realizing efficient discovery, accurate confirmation and trusted reporting of suspected targets by establishing a task relay mechanism for triggering high-precision local inspection by low-cost wide-area sensing by utilizing the function complementation characteristics of air and ground equipment, and continuously improving inspection efficiency through system-level feedback optimization. In order to solve the technical problems, the technical scheme provided by the invention is that the air-ground collaborative inspection method based on the heterogeneous platform comprises the following steps: The method comprises the steps that first, inspection original data are obtained through a front-end acquisition terminal, the inspection original data at least comprise defect target images, defect type labels, confidence degrees and corresponding space-time metadata, the front-end acquisition terminal can comprise heterogeneous inspection equipment such as unmanned aerial vehicles and ground robots, the space-time metadata comprise key information such as inspection time and equipment position coordinates, and basic data support is provided for subsequent report generation and target tracing. Step two, an industry term knowledge base is constructed and maintained, wherein the knowledge base comprises a defect type-industry term mapping table, a defect level evaluation rule, a causal relation model and a standard report paragraph template and is stored in the form of an extensible markup language or a graph database, and the industry term knowledge base further comprises a synonym dictionary and a domain ontology so as to support normalization and upper and lower concept reasoning of synonymous defect description, ensure that defects of different expressions can be uniformly identified and processed, and improve industry suitability. And thirdly, taking the defect type label as input, carrying out semantic retrieval and alignment in the industry term knowledge base through a natural language processing engine to generate a standardized description text conforming to industry specifications, and automatically determining the defect level according to a defect level evaluation rule, wherein the natural language processing engine adopts a p