Deep space rock image segmentation and identification method based on spiking neural network

Abstract

The invention discloses a deep space rock image segmentation and identification method based on a spiking neural network. The method comprises the following steps of constructing an image segmentation model based on a convolutional neural network (CNN), conducting semantic segmentation on the preprocessed rock image in the deep space environment, and segmenting the image into the sky, the ground, the large-size rock and the small-size rock, converting the CNN-based segmentation model into a spiking neural network (SNN)-based segmentation model, performing rock image segmentation by using an SNN-based segmentation model, and identifying the large-size rock by using the SNN-based image identification model converted in the mode. According to the method, the advantages of the traditional CNN in the aspects of feature extraction and precision in image segmentation and recognition are reserved, and meanwhile, the advantages of the SNN in the aspects of low power consumption, low time delay and the like can be exerted. The segmentation of the deep space rock image and the identification work of the large-size rock are completed through the network which better meets the real-time requirement, and a foundation is laid for the subsequent navigation and obstacle avoidance of the deep space detector.

Description

technical field [0001] The invention belongs to the technical field of vision-based image segmentation and image recognition after a deep-space probe lands, and in particular relates to a deep-space rock image segmentation and recognition method based on a pulse neural network. Background technique [0002] Since the 21st century, with the development of computer science and space technology, deep space exploration technology has become the only means for human beings to protect the earth, enter the universe, and find a new living home, which has attracted great attention from all countries in the world. Deep space exploration refers to the exploration of the moon and celestial bodies or space beyond the moon, and is an important part of human space activities. At present, soft landing on the surface of stars, conducting on-site patrol sampling, and sending the obtained samples back to the earth for research have also become the focus of deep space exploration in various cou...

AI Technical Summary

Problems solved by technology

In the field of image segmentation, convolutional neural networks are currently widely used for image segmentation, and relatively mature network structures and training algorithms have been formed, such as FCN, U-net, FPN, etc., but in the space environment, how to improve deep space detectors? Adapting to the small sample problem and resource shortage problem has become a big challenge

Images