CN-121984561-A - Satellite-to-ground data analysis method and device, electronic equipment and storage medium

CN121984561ACN 121984561 ACN121984561 ACN 121984561ACN-121984561-A

Abstract

The embodiment of the invention discloses a satellite-ground data analysis method, a device, electronic equipment and a storage medium, and relates to the technical field of satellite communication anti-interference. And integrating a lightweight ShuffleNet backbone network and a Coordinate Attention mechanism on the basis of YOLOv architecture to form an efficient SCA-YOLOv model so as to reduce the quantity of parameters and improve the positioning accuracy. And deploying the trained model on the edge node, analyzing the original video data in real time, and generating a detection result comprising the boundary frame coordinates, the category labels and the confidence coefficient. And then, associating a target track through matching cascade connection and DIoU matching algorithm, updating a target state by combining Kalman filtering, analyzing a target action to generate event information, transmitting the event information to an application center, and distributing the event information to a target terminal according to service requirements. The invention effectively solves the problems of poor real-time performance, low efficiency and incapability of flexibly coping with complex environments in the prior art.

Inventors

REN XIRAN
YE YIXUAN
CHEN PENGYU
SHU TIANYU
YU PENG
CHEN DANDAN
LIU WENJIE
FENG JIANYUAN
WANG SHIHUA
LI LINGYA

Assignees

亚太卫星宽带通信（深圳）有限公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (10)

1. A method of star-to-ground data analysis, the method comprising: acquiring original data of a target scene through Internet of things equipment, transmitting the original data to an edge computing node through a low-power consumption communication protocol, and integrating a ShuffleNet backbone network and a Coordinate Attention mechanism on the basis of YOLOv architecture to form an SCA-YOLOv model; The SCA-YOLOv model is deployed on the edge computing node, and the original data is analyzed through the SCA-YOLOv model to obtain a target detection result, wherein the detection result comprises a boundary frame coordinate, a category label and a confidence level; Matching cascade connection and DIoU matching are carried out on the detection result to obtain a target track and a matching result, and the states of the target track and the target are updated by combining Kalman filtering, so that actions of the target are analyzed and event information is generated by combining scene rules, wherein the event information comprises time, place, event type and target screenshot; And packaging the event information into a structured data packet through the edge computing node, transmitting the data packet to an application center through a non-ground network and a satellite Internet of things terminal, and distributing the data packet to a target terminal through each application center according to service requirements.
2. The method of claim 1, wherein the acquiring the original data of the target scene by the internet of things device, transmitting the original data to the edge computing node by the low power communication protocol, integrating ShuffleNet backbone network and Coordinate Attention mechanism on the basis of YOLOv architecture to form the SCA-YOLOv model comprises: The method comprises the steps of acquiring original data in real time through Internet of things equipment deployed in a target scene, and transmitting the original data to an edge computing node from a plurality of Internet of things equipment by adopting a low-power consumption communication protocol, wherein the original data comprises video streams and environmental parameters, the Internet of things equipment comprises a camera, an infrared sensor and a temperature and humidity sensor, and the low-power consumption communication protocol comprises LoRaWAN, zigbee; A lightweight ShuffleNet module is introduced into a YOLOv framework to serve as a backbone network, a Coordinate Attention module is embedded into a feature fusion path, the ShuffleNet module is used for reducing the quantity of parameters and the computational complexity, and the Coordinate Attention module is used for weighting through attention and enhancing the positioning accuracy.
3. The method for analyzing satellite-to-ground data according to claim 1, wherein the step of deploying the SCA-YOLOv model on the edge computing node, and analyzing the raw data by the SCA-YOLOv model to obtain a target detection result includes: Deploying the trained SCA-YOLOv model to the edge computing node to analyze the original video, and outputting a multi-scale feature map through each level of the ShuffleNet module; And capturing spatial context information of different sizes by combining with an SPPF module of the SCA-YOLOv model, and generating boundary frame coordinates, category labels and confidence degrees of the targets.
4. The method of analyzing satellite-to-ground data according to claim 1, wherein the performing matching cascade and DIoU matching on the detection result to obtain a target track and a matching result, and updating the states of the target track and the target in combination with kalman filtering includes: Matching cascade connection and DIoU matching are carried out on the detection results to obtain target tracks and matching results, the matching cascade connection is used for correlating the detection results of the same target in different frames, and the DIoU matching is used for calculating correlation indexes between the targets to improve the accuracy of target matching; And predicting and updating the target track by combining a Kalman filtering algorithm according to the historical track information of the target and the detection result, and updating the state information of the target, wherein the state information comprises position, direction, speed and acceleration.
5. The method of claim 4, wherein analyzing the motion of the object and generating event information in conjunction with a scene rule comprises: Analyzing the motion characteristics and the gesture changes of the target according to the state information of the target, judging the action type of the target, and generating event information by combining a preset scene rule, wherein the action type comprises static, moving, entering, leaving and traversing; And converting the action type of the target into an event type according to the event information in combination with a predefined scene rule, wherein the event type comprises regional invasion and personnel aggregation.
6. The star-to-ground data analysis method of claim 1, wherein said encapsulating the event information into structured data packets by the edge computation node comprises: And encapsulating the event information through the edge computing node to form a structured data packet, wherein the data packet comprises a time stamp, a geographic position, an event description and an associated target screenshot, and the structured data packet has a uniform data structure and format.
7. The method for analyzing satellite-to-ground data according to claim 1, wherein said distributing the data packet to the target terminal by each application center according to the service requirement comprises: processing and analyzing the event information according to service requirements and data characteristics by the application center to obtain a processing result, wherein the data characteristics comprise event types, time ranges and location areas; the processing result is distributed to related target terminals, wherein the target terminals comprise a service system and a user terminal, and the processing result comprises event classification, priority ordering and a corresponding response mechanism.
8. A satellite-to-ground data analysis apparatus, the apparatus comprising: The data acquisition and model construction module is used for acquiring original data of a target scene through Internet of things equipment, transmitting the original data to an edge computing node through a low-power consumption communication protocol, and integrating a ShuffleNet backbone network and a Coordinate Attention mechanism on the basis of YOLOv architecture to form an SCA-YOLOv model; The target detection module is used for deploying the SCA-YOLOv model on the edge computing node, and analyzing the original data through the SCA-YOLOv model to obtain a target detection result, wherein the detection result comprises a boundary frame coordinate, a category label and a confidence coefficient; The track matching and event generating module is used for carrying out matching cascade connection and DIoU matching on the detection result to obtain a target track and a matching result, updating the states of the target track and the target by combining with Kalman filtering, analyzing the action of the target and generating event information by combining with a scene rule, wherein the event information comprises time, place, event type and target screenshot; And the data packaging and satellite-to-ground transmission module is used for packaging the event information into a structured data packet through the edge computing node, transmitting the data packet to an application center through a non-ground network and a satellite Internet of things terminal, and distributing the data packet to a target terminal through each application center according to service requirements.
9. An electronic device comprising at least one processor and at least one memory, wherein, The memory has computer readable instructions stored thereon; the computer readable instructions are executed by one or more of the processors to cause an electronic device to implement the star-to-ground data analysis method of any one of claims 1 to 7.
10. A storage medium having stored thereon computer readable instructions, the computer readable instructions being executable by one or more processors to implement the method of satellite-to-ground data analysis of any one of claims 1 to 7.

Description

Satellite-to-ground data analysis method and device, electronic equipment and storage medium Technical Field The present invention relates to the field of satellite communications anti-interference technologies, and in particular, to a satellite-to-ground data analysis method, a satellite-to-ground data analysis device, an electronic device, and a storage medium. Background With the acceleration of global informatization, satellite internet of things is playing an increasingly important role as a key technology for connecting remote areas with city centers. However, the conventional communication network (e.g. 4G/5G) only covers about 20% of land area, resulting in large-area ground communication blind areas in remote areas, wide oceans, deserts, mountains and other environments. In these areas, the internet of things equipment bears important roles such as emergency management, infrastructure monitoring and environmental protection, but lacks effective ground network support, so that the data is difficult to timely transmit back to the urban command center. Although the satellite internet of things can make up the deficiency of the ground network by utilizing the wide coverage characteristics of the satellite internet of things, the connection of the global scope is realized, the connection is limited by the satellite-to-ground bandwidth, and massive data (particularly video data collected by a camera) collected by the internet of things equipment cannot be transmitted back through the satellite efficiently, so that the cost is high and the real-time requirement is difficult to meet. The traditional method relies on manual screening of monitoring data, has low efficiency and is easy to miss, and particularly, the method is worry-free in the scenes of earthquake relief, ecological monitoring and the like which need to process a large amount of information in real time. Therefore, there is an urgent need for a satellite-to-ground data analysis method capable of automatically analyzing and compressing data and only returning key information, adapting to satellite-to-ground transmission bandwidth, and ensuring efficient and real-time data transmission and intelligent monitoring in a non-ground network environment. Disclosure of Invention The embodiment of the invention provides a satellite-to-ground data analysis method, which aims to solve the problems that the prior art relies on manual screening of mass monitoring data, has low efficiency, is easy to miss detection, is difficult to meet the real-time requirement, and cannot flexibly cope with complex environments. The technical scheme is as follows: According to one aspect of the invention, the satellite-ground data analysis method comprises the steps of collecting original data of a target scene through Internet of things equipment, transmitting the original data to an edge computing node through a low-power consumption communication protocol, integrating ShuffleNet backbone network and a Coordinate Attention mechanism on the basis of YOLOv architecture to form an SCA-YOLOv model, deploying the SCA-YOLOv8 model on the edge computing node, analyzing the original data through the SCA-YOLOv8 model to obtain a detection result of the target, wherein the detection result comprises boundary frame coordinates, category labels and confidence, performing matching cascade and DIoU matching on the detection result to obtain a target track and a matching result, updating the states of the target track and the target through Kalman filtering, analyzing actions of the target and generating event information through combining scene rules, wherein the event information comprises time, place, event type and target screenshot, packaging the event information into a structured data packet through the edge computing node, transmitting the data packet to an application center through a non-ground network and a satellite Internet of things terminal, and distributing the data packet to the application center according to the service requirements. In one embodiment, original data of a target scene are acquired through Internet of things equipment, the original data are transmitted to an edge computing node through a low-power communication protocol, a ShuffleNet backbone network and a Coordinate Attention mechanism are integrated on the basis of an YOLOv framework to form an SCA-YOLOv model, the original data are acquired in real time through Internet of things equipment deployed in the target scene, the original data are transmitted to the edge computing node from a plurality of Internet of things equipment through the low-power communication protocol, the original data comprise video streams and environment parameters, the Internet of things equipment comprises a camera, an infrared sensor and a temperature and humidity sensor, the low-power communication protocol comprises LoRaWAN, zigbee, a lightweight ShuffleNet module is introduced into the YOLOv framework to serve as a bac