Kalman Filter Method For Image Processing Technology

Kalman filter is an adaptive filter widely used in state reconstruction and parameter estimation of complex systems.In addition to its good application in state estimation of complex systems,it also has good applications in the noise suppression of high-noise polluted images,and the restoration and reconstruction of distorted images.To this end,this project carried out research on image preprocessing methods based on Kalman filtering,which effectively improved the accuracy of target recognition by enhancing the quality of the image.The work of this project is as follows(1)A Kalman filtering method based on restoration of the sequential tracking target image is proposed.When the noise of the contaminated image is too large or the image is deformed,because the filtering of a single damaged image cannot meet the task's demand for image quality,Kalman filtering would be performed in complementation and redundancy on the information of multiple damaged images obtained sequentially to solve the problem,and prepares for the non-linear transformation of the damaged image.(2)A Kalman filter method based on fuzzy neural network for water quality evaluation is proposed.In the application engineering field of water quality monitoring,the distributed neural network proposed is constructed by convolutional neural network and fuzzy neural network.The digital image of sampled water is input to obtain its various indicators,and then used Fuzzy neural network to analysis indicators for the monitoring and scoring of water quality.The distributed parameters of the fuzzy neural network are optimized by Kalman filter to obtain better pattern recognition accuracy.(3)A Kalman filter method based on linear regression for face recognition is proposed.The face recognition method based on linear regression is described.The linear regression parameters are input into the Kalman filter in stages,and the classification model is optimally estimated to reconstruct the weight of the face to be recognized.Without the conventional image feature extraction process,a considerable accuracy of face recognition can still be obtained.This method can be used to suppress noise in image enhancement,which can effectively improve the accuracy of face recognition.