A Watershed-Segmentation-Based Improved Algorithm for Extracting Cultivated Land Boundaries

To accurately extract cultivated land boundaries based on high-resolution remote sensing imagery, an improved watershed segmentation algorithm was proposed herein a combination of pre- and post-improvement procedures. Image contrast enhancement used as the pre-improvement, while color distance Commission Internationale de l´Eclairage (CIE) space, including Lab Luv, regional similarity measure for region merging post-improvement. Furthermore, area relative error criterion (?A), pixel quantity (?P), consistency (Khat) were evaluating image accuracy. The in Red–Green–Blue (RGB) space selected to compare by extracting boundaries. validation experiments performed using subset Chinese Gaofen-2 (GF-2) with coverage 0.12 km2. results showed following: (1) contrast-enhanced exhibited obvious gain terms improving effect time efficiency algorithm. increased 10.31%, 60.00%, 40.28%, respectively, RGB, Lab, Luv spaces. (2) optimal scale parameters spaces C minimum areas 2000, 1900, D difference 1000, 40, 40. (3) boundary extraction 35.16% 29.58%, compared RGB space. accuracy ?A, ?P, Khat, that 76.92%, 62.01%, 16.83%, they 55.79%, 49.67%, 13.42% (4) Through visual comparison, efficiency, accuracy, comprehensive obviously better than color-based established evaluation indicators also proven be consistent evaluation. (5) method has satisfying transferability wider test 1 In addition, method, enhancement, perform CIE according simulated immersion results. It is useful attempt boundaries, which provides reference enhancing