阅读说明：本技术 一种基于无人船的雷达数据处理系统 (Radar data processing system based on unmanned ship ) 是由 王攀攀 王文强 燕居怀 王华超 崔雪梅 于 2019-10-25 设计创作，主要内容包括：本发明公开了一种基于无人船的雷达数据处理系统,包括数据采集模块以及数据处理模块；所述数据采集模块包括雷达传感器、遥感影像接收器、摄像模块、船体数据采集模块,所述数据采集模块将所采集的信息预处理后传输至数据处理模块；所述雷达传感器,所述雷达传感器发射电磁波对覆盖水域上的目标进行照射并接收其回波,获得目标跟踪数据并将接收到的电磁波处理为模拟信号。优点在于：本发明的数据处理模块通过模拟建模分析,计算出三维雷达数据,再通过激光雷达得到激光点云分类图、数字高程模型DEM、等高线、数字表面模型DSM、数字正射影像图DOM,最终计算出障碍物点得到障碍信息与前文所得障碍信息比对,保证最终得出的障碍信息准确无误。(The invention discloses a radar data processing system based on an unmanned ship, which comprises a data acquisition module and a data processing module; the data acquisition module comprises a radar sensor, a remote sensing image receiver, a camera module and a ship body data acquisition module, and the data acquisition module is used for preprocessing acquired information and then transmitting the information to the data processing module; the radar sensor emits electromagnetic waves to irradiate targets on a covered water area and receives echoes of the targets, target tracking data is obtained, and the received electromagnetic waves are processed into analog signals. Has the advantages that: the data processing module calculates three-dimensional radar data through analog modeling analysis, obtains a laser point cloud classification diagram, a digital elevation model DEM, a contour line, a digital surface model DSM and a digital orthographic projection image DOM through a laser radar, and finally calculates obstacle points to obtain obstacle information which is compared with the obstacle information obtained in the previous step, so that the accuracy of the finally obtained obstacle information is guaranteed.)

1. A radar data processing system based on an unmanned ship is characterized by comprising a data acquisition module and a data processing module;

the data acquisition module comprises a radar sensor, a remote sensing image receiver, a camera module and a ship body data acquisition module, and the data acquisition module is used for preprocessing acquired information and then transmitting the information to the data processing module;

the radar sensor emits electromagnetic waves to irradiate targets on a covered water area and receives echoes of the targets, target tracking data is obtained, and the received electromagnetic waves are processed into analog signals;

the remote sensing image receiver is used for receiving a remote sensing image downloaded by a satellite in real time and converting the remote sensing image into a digital signal;

the camera module at least comprises 10 panoramic cameras, wherein at least 50% of the panoramic cameras are positioned in the advancing direction of the ship body and used for acquiring video data around the ship body and converting the video data into digital signals;

the ship body data acquisition module is used for acquiring position data of a ship body, driving speed data of the ship body and acceleration data of the ship body and converting electric signals of the ship body into digital signals;

and the data processing module processes the data transmitted by the data acquisition module to obtain obstacle information.

2. The unmanned-vessel-based radar data processing system of claim 1, wherein the data processing module comprises a simulation modeling analysis:

s1, performing three-dimensional radar data conversion on the analog signals transmitted by the radar sensor through an NVIDIA Tegra K1 mobile processor;

s2, obtaining a laser point cloud data classification map, a Digital Elevation Model (DEM), a contour line, a Digital Surface Model (DSM) and a digital orthographic projection map (DOM) through laser radar data processing, and projecting three-dimensional data points to a grid map;

and setting all grids with relative heights larger than a certain threshold value as obstacle points to obtain obstacle information.

3. The unmanned-vessel-based radar data processing system of claim 2, wherein the distributed computing system is used for storing radar data, the radar data is processed at a high speed in a cloud computing manner by establishing a MapReduce model, a processing result is compared with obstacle information, consistent information is output, and inconsistent information is reintroduced into the step S1 for calculation.

4. The unmanned-vessel-based radar data processing system of claim 2, wherein the data acquisition module sequentially numbers each path of data according to respective inherent frame periods based on a same timestamp as a time reference when transmitting the data to the data processing module, and stores a corresponding relationship of the respective path of data frame numbers while storing the data.

5. The unmanned-vessel-based radar data processing system of claim 2, wherein the remote sensing image is processed by the steps of:

1) after receiving the remote sensing image, the remote sensing image receiver determines the resolution of the remote sensing image and intercepts the remote sensing image, and carries out data annotation on the intercepted remote sensing image;

2) preprocessing the intercepted remote sensing image by using a Canny edge detection algorithm, extracting the edge of the image, and overlapping the extracted image with an original image to highlight the channel characteristics for accelerating analysis;

3) constructing an image classification model, storing low-level image characteristic information in network parameters in training of classification tasks of a basic network, and transmitting the image characteristic information to a semantic segmentation model of the next level in a process of constructing a characteristic extraction model;

4) building a semantic segmentation model for segmenting road information in the remote sensing image; after training, the network parameters for extracting the road information are kept in the segmentation model.

6. The remote sensing image and deep learning-based road extraction method according to claim 5, characterized in that: the data marking of the intercepted remote sensing image in the step 1) is as follows: and observing and measuring the geographical range covered by the remote sensing image, and intercepting the original data of the image classification and semantic segmentation task by combining the actual situation of the channel to be extracted, wherein the original data is the RGB remote sensing image with the size of 256 × 256, namely 0.23 pixel per meter resolution scale.

7. The remote sensing image and deep learning-based road extraction method according to claim 5, characterized in that: the method for preprocessing the intercepted remote sensing image by using the Canny edge detection algorithm in the step 2) comprises the following steps:

removing image noise by Gaussian filtering to obtain a denoised image;

calculating gradients of the remote sensing image in the x direction and the y direction by using a Sobel operator through convolution operation, wherein the Sobel operator with the size of 3 has the following convolution kernels in the x direction and the y direction:

when the edge information is extracted, edges are reduced, only local maximum gradient is reserved, two thresholds are used through a Canny algorithm to distinguish edge pixels, a low threshold is used for filtering out a small gradient value caused by noise or color change, and a high threshold is used for distinguishing strong edge points and weak edge points.

8. The unmanned-vessel-based radar data processing system of claim 4, wherein when the MapReduce model is established, the radar type is transmitted to the Map function and Reduce function of the dynamic library in a main program by means of transmitting the radar type and ordering the Complere mode, the scanning period corresponding to the radar and the timestamp of the radar data and loading the dynamic library in the main program;

in the Map stage, a radar processing function is determined according to the type of a radar, the read radar data is analyzed, barrier information is calculated according to a track information linked list, then a Key/Value Key Value pair is established according to the modes of Partition and sort, the Key/Value Key Value pair with time as Key and all other contents as Value is established, the Key/Value Key Value pair is cut into different data blocks, and merging is carried out after sorting; the finally generated result comprises radar analysis data and an index file corresponding to the radar analysis data, wherein the index file identifies the initial position and the end position of each record, the timestamp and all barrier information; the calculation steps of the obstacle information are as follows:

inputting a track information linked list, data length and radar scanning period;

reading a piece of track information according to a fixed length;

checking whether the end of the whole track information chain is reached, if Yes, entering the next step, and if No, returning to the barrier information chain table until the end;

continuously checking whether the track data are in the same scanning period, and if Yes, pressing the track data into a temporary linked list; if the result is No, the next step is carried out;

taking out the flight path data of the same period from the temporary linked list to carry out obstacle information;

inserting the alarm result of the period into the tail part of the whole obstacle information;

and emptying the temporary linked list for recording the track data in the same period, and pressing the track data into the temporary linked list.

Technical Field

The invention relates to the technical field of radar data processing, in particular to a radar data processing system based on an unmanned ship.

Background

At present, a lot of devices for obstacle avoidance by using radar data exist in the market, but the progress obtained by analyzing the radar data generated by the similar devices on the obstacle is often insufficient, at the moment, in order to ensure good obstacle avoidance, the obstacle avoidance range can be only expanded, namely, the obstacle is detoured in a large radius mode, and the operation obviously consumes more energy;

if the accurate avoidance of the obstacle is realized, only the avoidance parameters of the real obstacle can be calculated, the analysis effect of radar data in the prior art is poor, the obtained real obstacle information needs to be filtered and screened by an algorithm, otherwise, the validity of the data cannot be ensured; in order to quickly and effectively support the analysis and test work of radar images, the problem that real radar data can be generated, interference and failure can be eliminated is urgently needed at present, and it is worth explaining that along with continuous progress of society, the size of a ship body is greatly reduced due to the appearance of an unmanned ship, and meanwhile, the size of the ship body is more limited due to the fact that dye carried by the unmanned ship is more limited, so that the unmanned ship radar data processing system capable of accurately calculating obstacle avoidance parameters is particularly important.

Disclosure of Invention

The invention aims to solve the problems in the prior art, and provides a radar data processing system based on an unmanned ship.

In order to achieve the purpose, the invention adopts the following technical scheme: a radar data processing system based on an unmanned ship comprises a data acquisition module and a data processing module;

the data acquisition module comprises a radar sensor, a remote sensing image receiver, a camera module and a ship body data acquisition module, and the data acquisition module is used for preprocessing acquired information and then transmitting the information to the data processing module;

the radar sensor emits electromagnetic waves to irradiate targets on a covered water area and receives echoes of the targets, target tracking data is obtained, and the received electromagnetic waves are processed into analog signals;

the remote sensing image receiver is used for receiving a remote sensing image downloaded by a satellite in real time and converting the remote sensing image into a digital signal;

the camera module at least comprises 10 panoramic cameras, wherein at least 50% of the panoramic cameras are positioned in the advancing direction of the ship body and used for acquiring video data around the ship body and converting the video data into digital signals;

the ship body data acquisition module is used for acquiring position data of a ship body, driving speed data of the ship body and acceleration data of the ship body and converting electric signals of the ship body into digital signals;

and the data processing module processes the data transmitted by the data acquisition module to obtain obstacle information.

In the above unmanned ship-based radar data processing system, the data processing module includes a simulation modeling analysis:

s1, performing three-dimensional radar data conversion on the analog signals transmitted by the radar sensor through an NVIDIA Tegra K1 mobile processor;

s2, obtaining a laser point cloud data classification map, a Digital Elevation Model (DEM), a contour line, a Digital Surface Model (DSM) and a digital orthographic projection map (DOM) through laser radar data processing, and projecting three-dimensional data points to a grid map;

and setting all grids with relative heights larger than a certain threshold value as obstacle points to obtain obstacle information.

In the unmanned ship-based radar data processing system, the distributed computing system is used for storing radar data, the radar data is processed at a high speed in a cloud computing mode by establishing a MapReduce model, a processing result is compared with obstacle information, consistent information is output, and inconsistent information is imported into the step S1 again for calculation.

In the above radar data processing system based on the unmanned ship, when the data acquisition module transmits data to the data processing module, the data acquisition module sequentially numbers each path of data according to the respective inherent frame period by using the same timestamp as a time reference, and stores the corresponding relationship of each path of data frame number while storing the data.

In the above radar data processing system based on unmanned ship, the processing steps of the remote sensing image are as follows:

1) after receiving the remote sensing image, the remote sensing image receiver determines the resolution of the remote sensing image and intercepts the remote sensing image, and carries out data annotation on the intercepted remote sensing image;

2) preprocessing the intercepted remote sensing image by using a Canny edge detection algorithm, extracting the edge of the image, and overlapping the extracted image with an original image to highlight the channel characteristics for accelerating analysis;

3) constructing an image classification model, storing low-level image characteristic information in network parameters in training of classification tasks of a basic network, and transmitting the image characteristic information to a semantic segmentation model of the next level in a process of constructing a characteristic extraction model;

4) building a semantic segmentation model for segmenting road information in the remote sensing image; after training, the network parameters for extracting the road information are kept in the segmentation model.

In the above radar data processing system based on unmanned ship, the data annotation of the intercepted remote sensing image in step 1) is: and observing and measuring the geographical range covered by the remote sensing image, and intercepting the original data of the image classification and semantic segmentation task by combining the actual situation of the channel to be extracted, wherein the original data is the RGB remote sensing image with the size of 256 × 256, namely 0.23 pixel per meter resolution scale.

In the above unmanned ship-based radar data processing system, the preprocessing of the intercepted remote sensing image by using the Canny edge detection algorithm in step 2) is as follows:

removing image noise by Gaussian filtering to obtain a denoised image;

calculating gradients of the remote sensing image in the x direction and the y direction by using a Sobel operator through convolution operation, wherein the Sobel operator with the size of 3 has the following convolution kernels in the x direction and the y direction:

when the edge information is extracted, edges are reduced, only local maximum gradient is reserved, two thresholds are used through a Canny algorithm to distinguish edge pixels, a low threshold is used for filtering out a small gradient value caused by noise or color change, and a high threshold is used for distinguishing strong edge points and weak edge points.

In the unmanned ship-based radar data processing system, when a MapReduce model is established, the radar type is transmitted to a Map function and a Reduce function of a dynamic library in a main program in a manner of transmitting the radar type and sequencing the Complere mode, a scanning period corresponding to the radar and a timestamp of radar data and loading the dynamic library in the main program;

in the Map stage, a radar processing function is determined according to the type of a radar, the read radar data is analyzed, barrier information is calculated according to a track information linked list, then a Key/Value Key Value pair is established according to the modes of Partition and sort, the Key/Value Key Value pair with time as Key and all other contents as Value is established, the Key/Value Key Value pair is cut into different data blocks, and merging is carried out after sorting; the finally generated result comprises radar analysis data and an index file corresponding to the radar analysis data, wherein the index file identifies the initial position and the end position of each record, the timestamp and all barrier information; the calculation steps of the obstacle information are as follows:

inputting a track information linked list, data length and radar scanning period;

reading a piece of track information according to a fixed length;

checking whether the end of the whole track information chain is reached, if Yes, entering the next step, and if No, returning to the barrier information chain table until the end;

continuously checking whether the track data are in the same scanning period, and if Yes, pressing the track data into a temporary linked list; if the result is No, the next step is carried out;

taking out the flight path data of the same period from the temporary linked list to carry out obstacle information;

inserting the alarm result of the period into the tail part of the whole obstacle information;

and emptying the temporary linked list for recording the track data in the same period, and pressing the track data into the temporary linked list.

Compared with the prior art, the invention has the advantages that:

1. the ship body data acquisition module processes data transmitted by the remote sensing image receiver and the camera module to obtain obstacle information.

2. According to the method, a data processing module firstly calculates three-dimensional radar data through analog modeling analysis, then obtains a laser point cloud data classification map, a digital elevation model DEM, a contour line, a digital surface model DSM and a digital orthographic image map DOM through a laser radar, and finally calculates obstacle points to obtain obstacle information which is compared with the obstacle information obtained in the previous step.

3. According to the method, the radar data are stored by using a distributed computing system, the radar data are processed at a high speed in a cloud computing mode by establishing a MapReduce model, and a processing result is compared with obstacle information, so that the finally obtained obstacle information is accurate and correct.

Detailed Description

The following examples are for illustrative purposes only and are not intended to limit the scope of the present invention.