Reduction Algorithm Based On Neighborhood Rough Sets And Application In Target Detection Of Scene Image

It is an era of big data, new data is producing every moment of our day,how to find the potential and valuable knowledge from these data efficiently is ahot research. Rough set is objective in the mining of knowledge which is widelyused in data mining. Due to the limitation of the classical rough set theory canonly deal with discrete data, the classical rough set has been extended to theneighborhood rough set by professor Hu.In order to deal with the dynamic data efficiently, this article explores theneighborhood rough set from the the perspective of information, the maininnovation points are as follows:(1) In the neighborhood rough set, by analyzing the decision attributevalues of samples in neighborhood, the inconsistent neighborhood matrix isdefined. The inconsistent neighborhood matrix can be used to narrow theresearch range when adding more attributes with respect to existingcondition attributes. It is faster to calculate the significance of attributes bymeans of condition entropy.A relationship of conditional entropy andpositive region has been found under the neighborhood systems. Therefore, aattribute reduction algorithm based on this.(2) Considering most realistic data is changing, how to deal with dynamic data has become the urgent problem. The original reduction set may be invalidand need to be updated, when the new simple is added to the neighborhooddecision system. The existed incremental methods has analyzed the changingrules of the neighborhood decision from algebra view. But this paper studies thechanging rules of the conditional entropy and the reduction set after adding newsimple from the perspective of information, furthermore, the research showedthat only inconsistent neighborhood of the new simple can cause these changes.An incremental attribute reduction under the information view is presented withthe framework of neighborhood system, which can update the original reductionset dynamically by reducing the new simple and the inconsistent neighborhood.The presented algorithm can avoid repeated reduction effectively.(3) The proposed attribute reduction algorithm is applied to the featureselection of target image detection, according to the characteristics of Hog highdimensional features, the principle of Hog feature extraction is analyzed, and theHog feature is selected by the means of cascade reduction.