Specific Target Extraction Based On The Feature Of The Object In High Resolution Remote Sensing Image

In view of the deficiencies of the current target recognition technology on specific targetextraction, the existing software with numerous features but low utilization, advance theinformatization surveying and mapping, and meet the demand of geographic conditions monitoring changesfound, this paper presents a mixed techniques of feature extraction for specific target extraction,through the study of the feature forming, the feature extraction with the objects of segmentedimage, and syncretize improved the rule expression algorithm of decision tree on targetextraction, in which several module package, develop a massive data processing object orientedchange monitoring system on the basis of the existing.Mainly related to technology:(1) Feature set construction. Through the programming of multiple feature algorithm,including spectral feature, texture feature and shape feature, and test the algorithm accuracy,through the quantitative comparison of the feature values to verified the correctness of therealization algorithm of the design, experiments show that the feature set can be very good tomeet the requirements for feature extraction.(2) Feature extraction and organization technology. Advance the standard to measure theseparability among features based on the separability criterion, through analyzing the distance toachieve the purposes of the most effective features extraction and realize the judgment ofmulti-dimensional feature space. The filtered features through the tree structure for visualizationto solve the key problem of its organization structure.(3) Feature subset selection technique. Implementation a method of information extractionof the decision tree of automatic optimization expression, puts forward a model to describe theconcise rules establishment way, solve a major problem of the image feature informationextraction, combine the feature extraction techniques with decision tree algorithm to describe theknowledge rules effective, so as to improve the extraction precision better.