CN-121644839-B - Scenic spot unmanned aerial vehicle live broadcast control system and method based on AI interaction

CN121644839BCN 121644839 BCN121644839 BCN 121644839BCN-121644839-B

Abstract

The application relates to the technical field of artificial intelligence and unmanned aerial vehicle control, discloses a scenic spot unmanned aerial vehicle live broadcast control system and method based on AI interaction, and aims to solve the problems that existing scenic spot live broadcast depends on manual operation, has poor interactivity, cannot normally run and has weak environmental adaptability. The method comprises the steps of starting the unmanned aerial vehicle through an automatic parking apron, completing self-checking, and receiving a flight request or an interaction instruction through a user platform. According to the scheme, the 7×24-hour time-oriented live broadcast is realized, the user interaction rate is improved by more than 80%, the natural language-driven course adjustment and personalized virtual anchor explanation are supported, the autonomous obstacle avoidance capability in a complex environment is achieved, and the privacy compliance is ensured through AI face blurring.

Inventors

JIN XIN
YANG MEI
Xiao Ganhong

Assignees

南昌大学

Dates

Publication Date: 20260505
Application Date: 20260204

Claims (9)

1. Scenic spot unmanned aerial vehicle live broadcast control system based on AI interaction, characterized by comprising the following parts: The unmanned aerial vehicle flight unit is used for executing aerial shooting and mobile live broadcasting tasks and returning high-definition video streams in real time through a 5G network; an automatic parking apron network is distributed at key nodes of scenic spots, provides automatic take-off and landing, charging and environment self-checking services for the unmanned aerial vehicle, and supports continuous operation; The AI interaction engine integrates natural language processing and computer vision and voice synthesis modules and is used for analyzing user interaction instructions, generating intelligent explanation contents and driving a virtual anchor to respond in real time; the user interaction platform provides a mobile terminal and a Web interface, and supports online users to send flight route requests, ask questions and interact, and enjoy live E-commerce activities; The central scheduling server is in charge of receiving user requests, planning flight paths, coordinating unmanned aerial vehicle resource allocation, scheduling AI services and managing live stream distribution; The data fusion and decision module is used for integrating scenic spot geographic information, tourist distribution thermodynamic diagrams, meteorological data and user behavior data to generate a dynamic flight strategy and content recommendation scheme; the live broadcast pushing flow and content management module is responsible for carrying out real-time processing on the video flow, superposing the AI interaction engine to generate content and pushing the content to the third-party live broadcast platform; the safety monitoring and emergency response module monitors the state of the unmanned aerial vehicle, airspace compliance and privacy protection policy execution conditions in real time and triggers an automatic return or scram mechanism when abnormal; the system calculates priority weights according to request time stamps, user member grades and award amount when a plurality of users submit flight requests simultaneously, responds to high-weight requests preferentially, integrates a plurality of requests with similar paths into one-time combined flight task through a task merging mechanism, generates a flight path by adopting a multi-objective optimization function comprising path smoothness, landscape coverage rate, obstacle avoidance safety margin and energy consumption weight, and starts a task merging algorithm when the central scheduling server receives the flight requests of the plurality of users within the same time period, thereby setting N pending requests exist at a moment, each request comprises a starting point Pi, an ending point Qi, a submitting time Ti and a priority weight Wi, and the system firstly calculates the similarity of paths between any two requests The definition is: Wherein, the To request Is set of preliminary planned path points, To request If the initial planning path point set of (a) 0.6 And Second, judging that the two tasks can be combined, generating a combined flight path after the two tasks are combined, covering all starting points and all end points, adopting a traveling business problem solver to optimize the access sequence, aiming at minimizing the total flight distance, executing the combined tasks by the same unmanned aerial vehicle, sequentially completing the view actions appointed by each user in the flight process, and responding to each user inquiry through an AI interaction engine; the AI interaction engine comprises a natural language processing module, wherein the natural language processing module is used for analyzing a natural language request containing time or space semantics, and calculating a target azimuth angle by combining geographic position information and an astronomical algorithm so as to drive the unmanned aerial vehicle to generate a specific sightseeing route; After the path planning engine receives the task instruction, a multi-target optimization algorithm is started to generate a comprehensive flight strategy, and an objective function of path planning is defined as follows: Wherein, the The path smoothness is measured by the integral reciprocal of the course angle change rate; Calculating the overlapping length ratio of the central line 50m buffer zone of the flight path and the marked scenic spots for the coverage rate of the scenic spots; Estimating the estimated energy consumption based on the flight distance, the wind resistance coefficient and the battery discharge curve; In order to avoid the obstacle safety margin, the times of the path crossing the high risk area are counted, and the weight coefficient is set as follows =0.3、 =0.4、 =0.2、 =0.1, Ensuring that the route has both visual appeal and operation safety, and outputting the planning result in the form of a Waypoint sequence, wherein the planning result comprises longitude and latitude height, flight speed, hover time and camera cradle head attitude angle of each Waypoint.
2. The system of claim 1, wherein the unmanned aerial vehicle flight unit is equipped with multispectral imaging equipment, and comprises a 4K visible light camera, a thermal infrared imager and a laser radar, wherein three types of sensor data are synchronously transmitted to a central dispatching server after being aligned through time stamps, the thermal infrared imager is used for capturing temperature distribution characteristics in a night-tour mode to enhance visual expressive force, and the laser radar is used for constructing a local three-dimensional point cloud map to support accurate obstacle avoidance and path optimization under complex terrains.
3. The system of claim 2, wherein the unmanned aerial vehicle flight unit embeds an edge calculation module, runs a lightweight YOLOv target detection model for identifying scenic spot landmark buildings, vegetation categories and tourist gathering areas in real time during flight, and embeds the identification results as semantic tags into video stream metadata for AI interaction engines to call to generate narrative content with geographic and cultural contexts.
4. A system according to claim 3, wherein the automated tarmac network is of modular design, each tarmac being provided with an ambient sensing sensor group comprising an anemometer, a rain sensor and an illumination intensity meter for monitoring local weather conditions in real time, automatically sending a fly-off warning to a central dispatch server when wind speeds exceeding a threshold or rainfall intensities exceeding a preset value are detected, and initiating a hatch closure and equipment waterproofing procedure.
5. The system of claim 4, wherein the natural language processing module in the AI interaction engine employs a BERT architecture-based domain adaptation model that uses scenic spot tour guide words, historical literature and common problem corpora for fine tuning on a pre-training basis to enable accurate understanding of local cultural terms, dialect expressions and historical classics, supporting accurate parsing of complex requests including specific location and time semantics.
6. The system of claim 5, wherein the AI interaction engine configures a multi-persona virtual anchor library comprising a history character image, a local cultural speaker and a cartoon IP image, the user selects an anchor style through the interaction platform, the AI system generates commentary conforming to the language style of the selected character according to the character setting and knowledge base, and outputs an audio stream having emotion intonation through a speech synthesis technique.
7. The system of claim 6, wherein the user interaction platform sets a flight route clicking function, after a user selects a starting point and an ending point on a two-dimensional or three-dimensional map of a scenic spot, the system converts a request of the selected starting point and the selected ending point into a geographic coordinate sequence and submits the geographic coordinate sequence to the central scheduling server, and the server generates a flight route which meets safety constraint and has optimal visual appreciation by adopting an improved a-type algorithm in combination with real-time airspace occupation, residual electric quantity of an unmanned aerial vehicle and meteorological data.
8. The system of claim 1, wherein the data fusion and decision module is connected to a scenic spot ticketing system API to obtain in real time the number of persons entering the scenic spot and the residence time data of each area, and combines the density information of the guests identified by AI in the video stream to generate a minute-level updated thermal profile of the guests, the thermal profile being an important input parameter for path planning, and guiding the unmanned aerial vehicle to cover the high-popularity area preferentially to promote live attraction.
9. A control method applied to the scenic spot unmanned aerial vehicle live broadcast control system based on AI interaction as claimed in any one of claims 1 to 8, comprising the steps of: Starting an unmanned aerial vehicle flight unit in a standby state through an automatic apron network, and entering a flight state after self-checking and environment evaluation are completed; Receiving a flight route request or an interactive questioning instruction of an online user through a user interaction platform, wherein the flight route request comprises target geographic position coordinates or a view requirement of natural language description; transmitting the received user request to a central scheduling server, and calling a data fusion and decision module by the server to generate a comprehensive flight strategy, wherein the comprehensive flight strategy comprises an optimal flight path, a predicted flight duration, an energy consumption prediction and an AI explanation theme; the central scheduling server sends a flight control instruction to a specified unmanned aerial vehicle, and the unmanned aerial vehicle executes an aerial photographing task along a planned path and returns a high-definition video stream in real time; the AI interaction engine synchronously analyzes the user questioning content, generates semantic answers by combining the scenic spot knowledge graph, and outputs an audio stream and synchronously plays the unmanned aerial vehicle shooting picture through a voice synthesis technology; the live broadcast pushing flow and content management module processes the original video flow in real time, and adds image-text explanation, interactive question-answer feedback and e-commerce commodity link information generated by AI and pushes the original video flow to the main stream live broadcast platform; the safety monitoring and emergency response module continuously monitors the flight state, the communication link quality and the privacy protection execution condition of the unmanned aerial vehicle, and immediately executes a preset emergency program when abnormality is found; after the task is finished, the unmanned aerial vehicle automatically returns to the nearest idle parking apron to finish landing, charging and data uploading, and the system updates the equipment state to standby.

Description

Scenic spot unmanned aerial vehicle live broadcast control system and method based on AI interaction Technical Field The invention belongs to the technical field of artificial intelligence and unmanned aerial vehicle control, and particularly relates to a scenic spot unmanned aerial vehicle live broadcast control system and method based on AI interaction. Background Along with the deep fusion of artificial intelligence and unmanned system technology, the application of intelligent unmanned equipment in the field of cultural tourism gradually evolves from single function demonstration to normalized and platform service modes, and an unmanned aerial vehicle serving as an intelligent carrier with air movement capability has shown remarkable advantages in multiple scenes such as aerial photography, inspection, emergency rescue and the like, and the capabilities of flexible deployment, wide area coverage and real-time transmission provide a brand new technical path for scenic spot content production and interactive experience. Under the background, the scenic spot live broadcast system based on the unmanned aerial vehicle gradually becomes an important component part of intelligent travel construction, and natural wind, light and human landscapes are presented through a high-altitude visual angle, so that the visual field limitation of the traditional fixed camera is broken through, and the immersion and participation degree of users on the line are improved. The unmanned aerial vehicle live broadcast system for scenic spot scenes particularly emphasizes the automatic operation capability and man-machine interaction level of the system, and has the core aims to realize autonomous take-off and landing, route planning, video acquisition and real-time push flow under the unattended condition, and enhance the bidirectional interaction between spectators and equipment through an intelligent means, so that online users can not only ' see ', but also participate in ' and ' influence ' the generation process of live broadcast contents, thereby constructing the real-sense dynamic, interactive and sustainable operation digital text travel content ecology. The prior art still faces multiple technical bottlenecks when realizing live broadcasting functions of unmanned aerial vehicles in scenic spots, firstly, most systems rely on manual remote control operation, flight tasks are difficult to be automatically executed in all weather and all time periods, so that live broadcasting frequency is low, labor cost is high, operation is not sustainable, secondly, interaction modes are limited to unidirectional video stream output, an AI-driven real-time semantic understanding and feedback mechanism is lacking, air line adjustment or visual angle switching cannot be performed in response to audience voice or text instructions, interactivity is seriously insufficient, and once again, flight control and AI decision systems are mutually separated, a unified intelligent closed loop is not formed, AI is only used for later content labeling or voice broadcasting and cannot be deeply involved in flight process regulation, finally, the system is poor in adaptability to complex scenic spot environments, when facing to real constraints such as crowded, electromagnetic interference and weather changes, dynamic obstacle avoidance and task re-planning capacity based on multi-mode perception is lacked, safety and stability are difficult to guarantee, and the problems are particularly required to be solved in a novel unmanned aerial vehicle live broadcasting system with deep fusion of intelligent machine, man-machine interaction and advanced process control under a context traveling scene requiring high concurrency and strong interaction and long-period operation. Disclosure of Invention The invention aims to make up the defects of the prior art, provides a scenic spot unmanned aerial vehicle live broadcast control system and method based on AI interaction, and can effectively solve the problems in the background art. The existing scenic spot live broadcast system has the problems of fixed live broadcast viewing angle, lack of interaction capability, high operation cost, discontinuous content output and the like, and is difficult to meet the requirements of users on immersive, dynamic and participatable real-time scenic experience. The invention realizes the fundamental transition of live broadcast in scenic spots from passive watching to active participation, from temporary activity to normalization service and from single video to intelligent interactive content platform by constructing a system architecture integrating unmanned aerial vehicle autonomous flight control, AI intelligent content generation, multi-mode man-machine interaction and cloud collaborative management. The system comprises an unmanned aerial vehicle flight unit, an automatic parking apron network, an AI interaction engine, a computer vision and voice synthesis module, a cent