Moving Target Feature Extraction Based On Wavelet Transform

The tracking of moving target has great market foreground andapplication value. In order to realize the steady tracks of target, First of allfeature extraction will be carried on to the image.In this paper, the purpose of research is how to extract the corner featurefrom one image. It can be used in many fields such as airplane andautomobile .By researching all kinds of ways of many papers and offeringthis solution.1,Feature extraction ways(1),The feature extraction in the spatial domain. The edge is the most basic feature .It exists in target and background ,Target and target ,area and area .The edge is the most significant feature thatthe image segmentation was dependent on ,Also is the foundation of thegrain feature and shape feature. The edge is insusceptible to the alternation ofthe grey,It can be regarded as the matching feature point .The grain are also one kind of basic targets feature ,It points out therange regulation and part pattern that has arisen again and again in the image.The grey matrix of one image give out the orientation of grey alternation,Adjacent to intermission and alternation scope,It is the base to analyse theranging the regulation and the part pattern.The shape is also one type of significant feature,The extraction of area'sshape feature have three kinds of means,First type ,The shape feature inarea is extracted by means of the different transforms to area's internal orexternal form. Second type ,The area's shape feature being extracted in theimage spatial domain. Third type ,Utilize the image's data structure todepict the area's shape features.(2),The wavelet transform extraction feature.When the wavelet transform to be used for extracting the feature, it canextract the target feature in the spatial domain and the frequency domain ,Moreover analyses the signal's high frequency section with the brief timewindow and analyses the signal's low frequency section with the long timewindow,When the wavelet depict the signal, In the practice application ,Most of features may be recognized by means of the matching on the wideryardstick ,The merely a little feature need to utilize fine yardstick messagecan discriminate .Hence , The wavelet foundation's selection is veryagile .On the basis of the different signal,Adopt the relevant mother waveletin order to extract the interesting feature.2,Corner detection ways One kind of corner detection way is based on the edge link code. Thatway possess 3 moves , First is to pre-divide cutting image,Second is toextract link code. Third is to detect the corner. The corner is named by thepoint of intersection. Another type of corner detection ways directly extracted the cornersfrom the grey scale image. The corner detection algorithm of grey scaleimage has two types. The corner detection ways based on formwork givensimilarity between formwork and the son window of every one image,thereby decide whether the corners relevant against the formwork exist in theimage. Another type of corner detection ways based on gradient, we cancalculate the edge's curvature to decide the existence of corners. Thenumbers were calculated through corners not merely have something to dowith the fringe intensity , but also have something to do with the fringeorientation variance ratio. Still there are some corner detection algorithms for instance:The cornerdetection ways based on self-adapting right line, Rosenfeld's algorithm,Plessey's algorithm,Susan's algorithm and CSS algorithm,Still there is onekind of CSS modification algorithm ,it can be called New-CSS.Still thereis one kind of fresh corner detection algorithm —The area algorithm, andthat algorithm may obtain the corner's place ,Bearing ,Angle size messagein the image.