Candidate recognition of pulsar based on convolution neural network

Abstract

The invention discloses a pulsar candidate identification method based on a convolution neural network, which mainly solves the problems that the existing pulsar identification method relies on artificial time-consuming and labor-consuming, and the traditional machine learning method is slow in identification speed and low in accuracy. The implementation scheme is as follows: 1. Processing The survey data into images and each image is marked by two classifications. Establishing The training set of pulsar candidate images and non-pulsar candidate images, and Establishing the verification set and the test set 2, building a convolution neural network model; 3, training that convolution neural network model with a training set and evaluating the model with a verification set; 4. Recognizing Each image in the test set by the trained convolution neural network, and outputting the classification marks of each image, namely pulsar candidate image and non-pulsar candidate image,. The inventionhas the advantages of low complexity, less training time consumption, fast recognition speed and high recognition accuracy, and can be used for processing astronomical data.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method for identifying a pulsar candidate, which can be used for processing astronomical data. Background technique [0002] With the advancement of science and technology, radio telescopes have higher time and frequency resolution, and the amount of observed pulsar data shows an explosive growth trend. How to effectively and quickly identify pulsar candidates in massive pulsar observation data is a problem that astronomical institutions need to solve. Currently, there are two main methods for identifying pulsar candidates in pulsar observation data: [0003] The first method is to process the observation data into images, manually browse the images one by one, and identify the pulsar candidates. [0004] The second method is to process the observation data into images, then preprocess the images, put them into the trained machine learning model, and reco...

AI Technical Summary

Problems solved by technology

[0005] For the above two methods, the first method uses manual identification, which is time-consuming and labor-intensive, and the efficiency is very low

The second method uses machine learning to identify. Training the model is time-consuming. The data to be screened needs to be preprocessed in the early stage, and the accuracy of model identification is not high.

[0006] With the continuous improvement of the performance of pulsar sky survey observation equipment, more and more pulsar sky survey observation data, the existing methods can no longer meet the requirements of identifying pulsar candidates in massive observation data

Images