A novel algorithm based on fractal texture analysis to classifyplant species in forest eco-system

Plants are one of the most fundamental and important species on earth. Forest eco-system has been an important source of various economically and socially significant plant species which are very difficult to be properly detected and identified. An automated image based species identification technique will be highly helpful, since it is a tedious job to collect and identify the samples manually in dense forests. In this study, we implement aaccuracy under normal conditions. However, keeping in mind the various natural and non-natural hindrances and noises, this study also dealt with the analysis of influence of those factors. The factors like fogginess of the surrounding, brightness of the picture and resolution of the camera used to take the picture did not have much influence on the % accuracy of the results. These factors affected the accuracy only at extreme ranges, whereas, the other tested factors like salt and pepper noise and digital zooming would influence the accuracy in a specific level even under a narrow range. On the whole, this method proved to be highly selectable for species identification of plants in forest eco-system than any other image processing method in practice fractal texture analysis using multi-level decomposition for identification of leaf species. Fresh leaves of 9 different dominative species of South Indian forests have been considered for study. The various images of the leaf samples of these plants were analyzed using a novel improved algorithm based on fractal texture analysis. The results showed that all the plants were identified with 98%.