Tree canopies extraction for tree outside forest areas using spectral, spatial, and vegetation indices attributes after applying the rule-based classification approach: a case study for Jodhpur, Rajasthan, India
This paper summarizes advances in rule based/object based extraction of tree crown in different stratum of TOFs area in Jodhpur Town using World view 2 data. The generation of image objects is offered through a convolution filter followed by an automatic image segmentation approach focused on spectral feature objects and range of Vegetation Indices (VIs). PCA analysis has also been performed to find variation among variables in the dataset and to represent the distribution of samples. The overall accuracy detection rate and the Kappa coefficient were 98.25% and 91.86, respectively. Accuracy assessment carried out for the quality and quantity of crowns extracted. Qualitative analysis detected maximum accurate detection for planted areas 0.971 followed by scattered 0.886 and linear 0.73 TOFs.Quantitave accuracy assessments for tree delineation indicated highest Accuracy Index (AI) % index for Plantation (97.73%) while Crown detection error of 14% noted for single crowns and 49% for clustered crowns. Ultimately, the reliability of this process is seen by contrasting these graded crowns with those obtained using traditional manual methods.