A Spiking Corner Detection Method Based on Gray Image

Abstract

The invention discloses a method for detecting a Spiking corner point based on a grayscale image, which belongs to the technical field of image processing and focuses on solving the problems of low corner point detection accuracy, complicated operation and slow detection speed in the prior art. The present invention mainly utilizes the characteristics of the Spiking neural network to invent a detection template to detect the corner information of the image, and determine the corner of the gray image by the characteristic that the voltage value of the detection template reaches the minimum when the voltage value is at the corner. points, the present invention can better detect the corner points of the image, and extract the time information and spatial information of the corner points of the grayscale image. The application scenarios of the present invention are mainly in image corner detection, character object tracking and recognition, motion estimation and other computer scenarios where corner information is applied. Involves Spiking neural network, machine learning.

Description

technical field

[0001] A method for detecting Spiking corners based on grayscale images. The invention is applied to image corner feature extraction, optical flow calculation, target recognition, tracking, motion estimation, three-dimensional reconstruction and other computer vision occasions involving corner applications. Involves Spiking neural network, machine learning. Background technique

[0002] The corner feature of the image is an important feature in image processing, and its feature has strong robustness, and has the characteristics of rotation invariance and no change with illumination. This feature plays a very important role in moving target tracking, streamer calculation, motion evaluation and analysis, and construction of 3D scenes. Regarding the definition of corner points, the relatively unified statement so far is: the edge point with the largest local curvature has many corner points for a general image, so the method of related research has strong unive...

Examples

Embodiment Construction

[0029] The Spiking corner detection method based on grayscale images is as follows:

[0030] (1) Perform grayscale processing on the color image, if it is a grayscale image, no processing is required. ,In experiment Figure 5 (a) is a color picture, and grayscale processing is performed on the color picture.

[0031] (2) Encode the grayscale image to obtain the image pulse ignition time.

[0032] a=T-|x i -T s |

[0033]

[0034] This encoding scheme takes a relative time Ts as the standard, and encodes any real number xi with encoding to its ignition time relative to Ts. Where ti is the time value after encoding, T is the constant time encoding measure, and Ts is set as the current pixel value pi. After the pulse time encoding, the time encoding is the ignition time ti=1, and the non-ignition time ti=0.

[0035] (3) Initialize the weight detection template: the specific template is as follows figure 2 As shown, firstly, the upper semicircle of the circular weight ...

technical field

[0001] A method for detecting Spiking corners based on grayscale images. The invention is applied to image corner feature extraction, optical flow calculation, target recognition, tracking, motion estimation, three-dimensional reconstruction and other computer vision occasions involving corner applications. Involves Spiking neural network, machine learning. Background technique

[0002] The corner feature of the image is an important feature in image processing, and its feature has strong robustness, and has the characteristics of rotation invariance and no change with illumination. This feature plays a very important role in moving target tracking, streamer calculation, motion evaluation and analysis, and construction of 3D scenes. Regarding the definition of corner points, the relatively unified statement so far is: the edge point with the largest local curvature has many corner points for a general image, so the method of related research has strong unive...

Examples

Embodiment Construction

[0029] The Spiking corner detection method based on grayscale images is as follows:

[0030] (1) Perform grayscale processing on the color image, if it is a grayscale image, no processing is required. ,In experiment Figure 5 (a) is a color picture, and grayscale processing is performed on the color picture.

[0031] (2) Encode the grayscale image to obtain the image pulse ignition time.

[0032] a=T-|x i -T s |

[0033]

[0034] This encoding scheme takes a relative time Ts as the standard, and encodes any real number xi with encoding to its ignition time relative to Ts. Where ti is the time value after encoding, T is the constant time encoding measure, and Ts is set as the current pixel value pi. After the pulse time encoding, the time encoding is the ignition time ti=1, and the non-ignition time ti=0.

[0035] (3) Initialize the weight detection template: the specific template is as follows figure 2 As shown, firstly, the upper semicircle of the circular weight ...

Images