The Leaf Of The Northeast Common Woody Plants’ Information Digitization And The Study On The Species Identification

In this study, by choosing a total of50of the Northeast China common woody plant, scan about1000plant leaf image, the use of computer image processing and analysis software, accurate measurement of leaf original digital features a total of31, and set the formula28relative digital feature index. The coefficient of variation to determine stability of data, using SPSS19.0canonical discriminant analysis on main identification and evaluation, application of index sequential addition, stepwise elimination method, can be used as the characteristic indexes of selected combination of the most effective experimental species identification. The main results are as follows:(1) Stability evaluation using the coefficient of variation was28relative digital indicators. Reach:near circular degree, the ratio of length to width, the wide position, the ratio of rectangle area, the ratio of rectangle perimeter, long arc distance ratio, shorter arc length ratio, on the pitch area ratio, the ratio of leaf capsule area, the ratio of leaf capsule perimeter coated perimeter than small variation coefficients of these10indicators, better stability than others, can be used as alternative indicators.(2) The only using single feature index to recognize50kinds of plants is not feasible. The classification results show that single feature index, all single index of plant individual comprehensive judgement right rate did not exceed50%, the comprehensive judgement right rate of the size of the order:the order of the ratio of length to width> on the pitch area ratio> the wide position> near circular degree> the ratio of rectangle area> the ratio of rectangle perimeter> shorter arc length ratio> long arc length ratio> the ratio of leaf capsule perimeter> the ratio of leaf capsule area, wherein. Among them, the synthetic discriminant ratio of length to width accuracy is highest,49.9%.(3) Verified that the combined two feature indexes of significant correlation can also improve the rate of correct discrimination indicator plant. The correlation between changes in indicators, classification results of discriminant analysis index combination of no direct relation, characteristic indexes are can be any combination of discriminant analysis.(4) The10indexes are all integrated on50kinds of plant are canonical discriminant analysis. The classification results show, comprehensive judgement right rate reached99.9%, the rate of correct classification of49kinds of plants are100%, only to distinguish No.5plant a correct rate of94.7%, but still can not guarantee to achieve complete identification of individual all leaves. The structure of the matrix, the classification results reflect the impact on plant identification indexes the size order:the ratio of length to width> near circular degree> on the pitch area ratio> the wide position> the ratio of leaf capsule area> the ratio of rectangle perimeter> the ratio of rectangle area> the ratio of leaf capsule perimeter> shorter arc length ratio> long arc length ratio.(5) According to the structure characteristics of10index matrix combination classification results,10indicators of discriminant analysis the size order, by sequentially adding index method, the formation of index combination of canonical discriminant analysis. The index "the ratio of leaf capsule perimeter","a long arc length ratio" adding no gain on the classification results, the remaining8indicators combination of canonical discriminant analysis, classification results and10indicators combined results exactly the same specification is invalid, the two indicators that invalid.(6) After5step gradual elimination method, finally obtains two groups the most effective combination of index can reach full recognition, respectively,"the ratio of length to width, on the pitch area ratio, the wide position, the ratio of leaf capsule area, the ratio of rectangle area, longer arc length ratio" and "near circular degree, on the pitch area ratio, the wide position, the ratio of leaf capsule area, the ratio of rectangle area, longer arc length ratio". The two characteristic indexes combined with6indexes, comprehensive evaluation on50kinds of plants, the correct rate is100%, can identify all individuals.