Search

CN-121985195-A - System and method for generating short video of travel beat

CN121985195ACN 121985195 ACN121985195 ACN 121985195ACN-121985195-A

Abstract

The invention discloses a system and a method for generating a short video of a travel shot, which relate to the technical field of short video generation and comprise the following steps of deploying high-speed cameras with inertia measuring units in a dynamic scene of a scenic spot, deploying environment light-sensing self-adaptive cameras in a static scene, and connecting all cameras with an edge computing gateway; the camera starts a corresponding shooting mode according to scene types, a dynamic scene tracks a motion track through a multi-machine-position space-time synchronization algorithm, a static scene adjusts shooting parameters according to rays and people flow, a cloud AI engine extracts behavioral characteristic emotion characteristics and basic characteristics of a user, the matched video templates are screened from a template library through a collaborative filtering algorithm in combination with scene characteristics to generate exclusive short videos, the system carries out association binding on the generated exclusive short videos and user characteristic data, establishes one-to-one or many-to-one short video attribution relation for the user to preview and download for use on line, and inconvenience caused by shooting by manual handheld shooting equipment is avoided.

Inventors

  • LI XIAOCHUN

Assignees

  • 浙江水科文化集团有限公司

Dates

Publication Date
20260505
Application Date
20251230

Claims (8)

  1. 1. The short video generation method for the travel beat is characterized by comprising the following steps of: S1, deploying high-speed cameras with inertia measurement units in a dynamic scene of a scenic spot, deploying environment light sense self-adaptive cameras in a static scene, and connecting all cameras with an edge computing gateway; s2, starting a corresponding shooting mode by the camera according to the scene type, tracking a motion track by a dynamic scene through a multi-machine-position space-time synchronization algorithm, and adjusting shooting parameters by a static scene according to rays and people stream; S3, extracting behavioral characteristics, emotion characteristics and basic characteristics of a user by the cloud AI engine, and screening an adaptive video template from a template library by combining scene characteristics through a collaborative filtering algorithm to generate a dedicated short video; S4, the system carries out association binding on the generated exclusive short videos and the user characteristic data, and establishes a one-to-one or many-to-one short video attribution relationship so as to be used for online previewing and downloading of the user.
  2. 2. The method for generating a short video for a trip according to claim 1, wherein in S2, the workflow of the dynamic scene shooting mode includes the steps of: After the high-speed camera is started, a multi-machine-bit communication link is established through an edge computing gateway, and a Network Time Protocol (NTP) and Precision Time Protocol (PTP) dual synchronization mechanism is adopted to finish multi-machine-bit time stamp calibration; Based on GPS positioning data and a scenic spot electronic map, a unified space coordinate system is established, the space position coordinates of each machine position are determined, and space reference calibration is completed; The high-speed camera acquires motion attitude data of the equipment in real time, the acquisition frequency is consistent with the frame rate of the camera, and the motion data and the video frame data are bound and transmitted to the edge computing gateway; detecting video frames shot by all machine positions, and realizing cross-view locking of multiple machine positions on the same user by extracting human body contours and key feature points of the user; The edge computing gateway performs fusion processing on the video frame data of multiple machine positions and the motion data of the high-speed camera, and removes data noise of the high-speed camera by adopting a Kalman filtering algorithm to obtain a motion smooth track of the equipment; And predicting the motion trail of the user by adopting a particle filtering algorithm based on the position data of continuous frames, so as to realize the real-time tracking of multiple machine positions on the motion trail of the user.
  3. 3. The method for generating the short video of the trip according to claim 2, wherein the cross-view locking of multiple machine positions to the same user is realized by extracting the human body outline and key feature points of the user, and the method comprises the following steps: Each machine position detects human body targets of the real-time video frames through YOLOv models, and outputs the boundary frame coordinates of the effective human body targets; Extracting and comparing human global contour features, human key feature points and clothing features of adjacent frames, and confirming that the current frame and the previous frame are the same user; And taking the target feature of the high-point machine as a reference, calculating the contour feature vector cosine similarity and the local feature vector weighted cosine similarity of other machine targets and the reference target, and binding the global user ID when the similarity meets a threshold value, so as to realize cross-view locking of multiple machine positions to the same user.
  4. 4. The method for generating a short video of a trip as set forth in claim 1, wherein S3 specifically includes the steps of: The cloud AI engine receives the encrypted video clips uploaded by the edge computing gateway and performs structural extraction on user characteristics; based on the scene characteristic labels, matching the same scene templates in the template library, and primarily screening a plurality of candidate templates; combining template selection data of the users with the same scene and the same characteristics, sequencing the remaining templates, and selecting an optimal template; and editing the video according to the optimal template, enabling the core behavior fragment of the user to correspond to the slow shot or special effect node of the template, enabling the common fragment to correspond to the conventional editing node, and automatically synthesizing the short video.
  5. 5. The method for generating a short video for a trip of claim 4, wherein the step of structurally extracting the user features comprises: Identifying key actions of a user based on a human body posture estimation model, counting duration time of each action, and marking core behavior fragments; and analyzing the key facial expressions of the user through the facial expression recognition model, and outputting pleasure, calm and other emotion labels.
  6. 6. The method for generating a short video for a trip according to claim 1, wherein in step S4, the user downloads: the user initiates an acquisition request through at least one channel of a two-dimensional code identifier, a scenic spot official applet and a tourist center self-service terminal which are arranged in a scenic spot; After the system receives the request, a verification interface is popped up, a user inputs an order number or an identity card number, the system is compared with the scenic spot ticketing system data, and the matching is passed; after verification is passed, the user enters a short video preview interface, and after the user selects the target resolution, the system immediately generates a short video with the corresponding resolution for the user to download to the local equipment or share to the social platform.
  7. 7. The method for generating a short video for a trip according to claim 6, wherein in step S4, a short video secondary optimization step is further involved: After the user obtains the short video, an optimization request can be initiated through an official applet to replace the score, adjust the transition style, add personalized characters, replace the filter and cut the lens segment; the cloud AI engine recalls template library resources or adjusts generation parameters according to user optimization requirements to generate optimized short videos; the user previews the optimized short video through the original identity verification channel, and supports the initiation of the optimization request again or the direct downloading and sharing.
  8. 8. A short video generation system adapted to the short video generation method of any one of claims 1 to 7, comprising: The scene camera deployment module deploys a high-speed camera with an inertial measurement unit in a dynamic scene of a scenic spot, deploys an environment light sense self-adaptive camera in a static scene, and all cameras are connected with an edge computing gateway; The scene shooting module is used for starting a corresponding shooting mode according to the scene type by the camera, tracking a motion track by a dynamic scene through a multi-position space-time synchronization algorithm, and adjusting shooting parameters according to rays and people stream by a static scene; The cloud AI engine extracts behavioral characteristics, emotion characteristics and basic characteristics of a user, and screens an adaptive video template from the template library by combining scene characteristics through a collaborative filtering algorithm to generate an exclusive short video; and the user preview downloading module is used for carrying out association binding on the generated exclusive short video and the user characteristic data by the system, and establishing a one-to-one or many-to-one short video attribution relationship for online previewing and downloading of the user.

Description

System and method for generating short video of travel beat Technical Field The invention relates to the technical field of short video generation, in particular to a system and a method for generating a short video of a trip. Background Scenic spot tourism has become a leisure mode of people, and each big scenic spot in the scenic spot has become the punching place when people travel, and along with the popularization of short video social contact, record play in the tourism process is one of the core demands of tourists. At present, tourists mainly acquire video content in a mode of self-shooting, assisting shooting by peers or hiring a tourist staff and the like, but the manual shooting mode has a plurality of inconveniences in high-speed movement or dangerous scenes such as a slideway, a roller coaster, drifting and the like, one is difficult for the tourists to independently shoot, potential safety hazards are easily caused by handheld equipment, the other is easy to cause the problem of picture shaking when the common tourists shoot, the video quality is influenced, and the third is that the shot video also needs to be manually clipped, is time-consuming and is contrary to the requirement of easy play. Disclosure of Invention The invention provides a system and a method for generating a short video of a travel shoot, which are used for solving the defects of manual shooting in the prior art and improving the convenience of shooting cards for travel in scenic spots. The invention provides the following technical scheme: In a first aspect, the present invention provides a method for generating a short video of a trip, including the steps of: S1, deploying high-speed cameras with inertia measurement units in a dynamic scene of a scenic spot, deploying environment light sense self-adaptive cameras in a static scene, and connecting all cameras with an edge computing gateway; s2, starting a corresponding shooting mode by the camera according to the scene type, tracking a motion track by a dynamic scene through a multi-machine-position space-time synchronization algorithm, and adjusting shooting parameters by a static scene according to rays and people stream; S3, extracting behavioral characteristics, emotion characteristics and basic characteristics of a user by the cloud AI engine, and screening an adaptive video template from a template library by combining scene characteristics through a collaborative filtering algorithm to generate a dedicated short video; S4, the system carries out association binding on the generated exclusive short videos and the user characteristic data, and establishes a one-to-one or many-to-one short video attribution relationship so as to be used for online previewing and downloading of the user. As a further improvement of the present invention, in S2, the workflow of the dynamic scene shooting mode includes the steps of: After the high-speed camera is started, a multi-machine-bit communication link is established through an edge computing gateway, and a Network Time Protocol (NTP) and Precision Time Protocol (PTP) dual synchronization mechanism is adopted to finish multi-machine-bit time stamp calibration; Based on GPS positioning data and a scenic spot electronic map, a unified space coordinate system is established, the space position coordinates of each machine position are determined, and space reference calibration is completed; The high-speed camera acquires motion attitude data of the equipment in real time, the acquisition frequency is consistent with the frame rate of the camera, and the motion data and the video frame data are bound and transmitted to the edge computing gateway; detecting video frames shot by all machine positions, and realizing cross-view locking of multiple machine positions on the same user by extracting human body contours and key feature points of the user; The edge computing gateway performs fusion processing on the video frame data of multiple machine positions and the motion data of the high-speed camera, and removes data noise of the high-speed camera by adopting a Kalman filtering algorithm to obtain a motion smooth track of the equipment; And predicting the motion trail of the user by adopting a particle filtering algorithm based on the position data of continuous frames, so as to realize the real-time tracking of multiple machine positions on the motion trail of the user. As a further improvement scheme of the invention, the cross-view locking of multiple machine positions to the same user is realized by extracting the human body outline and key characteristic points of the user, and the method comprises the following steps: Each machine position detects human body targets of the real-time video frames through YOLOv models, and outputs the boundary frame coordinates of the effective human body targets; Extracting and comparing human global contour features, human key feature points and clothing features of adjacent frames, and confirming