Research On Improving The Performance Of Target Recognition Based On Joint-Transform Correlator

Target identification is important in many civilian and defense applications.Optical correlator provides a simple technique in both fast verification and identification of data.Traditional optical recognition methods include: matched filter,joint transform correlator,etc.The traditional optical recognition of the system suffer from many problems and constraints of practical considerations in their application and how to improve the recognition performance of the system is a notable problem.Joint-transform correlator is easy to be implemented and used for real-time detection of targets,but the correlation peak is not high and not sharp in general,and the recognition ability is low.In this paper,the mathematical transformation is applied to the optical system,with the using of the fast-transform function of the computer.As an alternative,we aim at improving the performance of target recognition based on optical joint-transform correlator.Combining the optical wavelet joint-transform correlator and Logpolar-Mellin Transform,we carry out the real-time detection of targets detection and the distortion-invariant recognition,the main work is as follows:1.In this paper,the optical image recognition technology is applied in biological targets detection,and the optical wavelet transform is used to recognize multiple red blood cells,which leads to improve the system's ability of recognition.In the joint-transform correlator,wavelet transform can be implemented by wavelet filtering at the spectrum plane,hence the optical wavelet joint-transform correlator can be built.Compared with the classical joint-transform correlator,Matlab simulation results show that,due to the introduction of the wavelet transform,the correlation peak becomes sharper,the peak-to-noise ratio is significantly enhanced and the half-width of the peak is greatly reduced.Moreover,in the multi-target recognition,different targets are easily distracted and identified.The detection ability of this system is improved.2.The optical image recognition is applied in distortion of scale and angles changing at the same time of biological targets detection.Mellin transformation is used to identify targets at different scales;Polar coordinate-transformation is used to solve the identified problem of the rotation-image target recognition.The method that is preprocessing of biological targets of the distortion by using Mellin transformation on the log-polar transformation is used in the computer.It combines with joint-transform correlator building recognition system of thescale and rotating the image.Matlab simulation results,compared with the traditional correlator to the distorted image unrecognized,show that the introduction of Mellin transformation on the log-polar transformation,which excludes the interference caused by the displacement,scale and rotation at the same time,achieves the relevant identification and detection of the distortion targets.3.An image of the distortion recognition system based on optical wavelet joint-transform correlator is presented,combining the optical wavelet joint-transform correlator and Logpolar-Mellin Transform.The scale and rotation of the optical image can be recognized by the system which has a high discernment.The reference image and the tested Target image are pretreat by using Logpolar-Mellin Transform.The images are used as the input image of the optical wavelet joint-transform correlator.Afterward,appropriate parameters of the wavelet filter function are selected by the computer on the spatial frequency spectrum plane.Finally the system gets well cross-correlation peak on the output plane.Matlab simulation results show that the system can eliminate the interference caused by the scaling and rotation of the target image at the same time,receiving excellent PNR and ACR.