Researching Corner Detection Based On Non-causal Fractional Gradient Operator

In recent years,due to the continuous deepening of research in the field of computer information technology,the computing level of computers has also been greatly improved,so that fractional calculus can be realized.For the calculus operation is no longer limited to the integer order,the operation order is generalized to the fraction even to the complex number,thus completely realizes the continuous order calculus.The application of fractional calculus involves areas such as medical image processing,seismic analysis,and climate research.For example,control theory,electromagnetic field theory,signal processing,and other directions require the use of fractional calculus.Harris corner detection,which is commonly used in image feature extraction,has the disadvantage that it is extremely sensitive to noise and also has a large number of pseudo corners.Therefore,the relationship between causal and anti-causal system and forward and backward fractional calculus is studied.This paper proposes a corner detection method based on non-causal fractional gradient operator to improve the traditional Harris algorithm and reduce the disadvantageous effects of these deficiencies in feature processing.This paper investigates the relationship between causal and anti-causal systems,forward and backward fractional calculus,and obtains the system's amplitude and phase characteristics.Based on the above discussion,we design the non-causal fractional gradient operator.This paper proposes a corner detection method based on non-causal fractional gradient operator.This method can retain part of image edge details when extracting corner points,and has the advantage of effectively suppressing noise and enhancing medium and high frequency.That is,when the corner points are extracted,some image edge detail information is retained,so that the corner point detection of the image achieves a better effect.By comparing and analyzing the method of this paper with the traditional integer-order corner detection method,it is proved that the proposed method achieves a trade-off between anti-noise ability and detection accuracy.Even an image with too much noise,the accuracy rate can be improved by at least 3%,the missed rate can bereduced to about 5%,and the operation speed can be basically consistent with Harris.In addition,in order to better verify the proposed method can be used for feature matching,then this method is applied to the practical application of image mosaic.