12 December 2011 Science Briefs

SNAMP Publication #7: Object-based image analysis of downed logs in a disturbed forest landscape using lidar


Blanchard, S., M. Jakubowski, and M. Kelly

Research Highlights:

  • Lidar (light detection and ranging) was used to map downed logs within sparse and open forest canopy. Our study examined a small area within the Last Chance project in the Tahoe National Forest, the northern SNAMP study area.

  • The study focused on using object-based image analysis (OBIA) for mapping downed logs using lidar data. OBIA is a new method in remote sensing that analyzes objects, instead of pixels in an image.

  • Specific lidar data metrics that are useful in mapping downed logs were determined.

  • By selecting these metrics we were able to map downed logs with relatively high accuracy (73 percent were identified correctly). The classified downed log objects yield useful information for characterizing the logs such as pattern, orientation, and dimensions.

  • However, substantial analyst time was required for visual interpretation and delineation of objects. Automating the identification of homogeneous objects around logs proved difficult. OBIA image classification methods combined with lidar data were not entirely satisfactory in terms of efficiency or image classification automation.

  • Our results show that lidar data can be used to map downed logs but the OBIA approach is time consuming and may be inefficient (versus manual delineation).

  • Lidar data coupled with OBIA mapping approaches can be used to supplement traditional field-based downed log mapping methods, but there can be substantial analyst time required to achieve high accuracies.

Experiment:

The objective of this study was to examine the efficacy of airborne lidar-derived data products and their use in rule-based OBIA analysis and image classification for identifying and quantifying downed logs. We tested the methods in a forested landscape with an open canopy on steep terrain.

OBIA is an image processing and analysis framework in which image classification and analysis focuses on discrete groups of similar pixels, or “objects,” rather than individual pixels.

The study area within the Tahoe National Forest near the Last Chance Fuels Treatment Project in Placer County, CA was burned by a forest fire in 2001. Downed logs are clearly visible in the acquired lidar data and on the ground from field surveys. Data metrics such as elevation, intensity, and point density were extracted from the lidar data to distinguish downed logs, ground, and canopy cover using the OBIA software eCognition.

Results:

The results show that downed logs in an open canopy forested landscape can be successfully delineated from lidar-derived metrics using the OBIA framework. Accuracy was relativity high, although some areas were misidentified. Over-identification occurred in areas with clusters of logs as well as in areas of vegetation and canopy cover. Under-identification occurred in areas where the collection pattern of the lidar sensor was most pronounced in the calculated lidar images. OBIA methods combined with lidar data were not entirely satisfactory in terms of efficiency or automation. Substantial analyst time was required for visual interpretation and manual object delineation in areas where homogenous log objects were difficult to extract automatically.

Conclusions:

  1. Downed logs were successfully delineated using lidar-derived metrics in the OBIA framework. The Triangulated Irregular Network (TIN) surface of lidar points, the elevation minimum, and the total number of points layers were the most useful metrics for the image classification.

  2. Accuracy was relatively high: 73 percent of the identified downed logs were delineated correctly.

  3. The delineated downed log objects yield useful information for characterizing the logs such as pattern, orientation, and dimensions.

  4. Lidar data coupled with an OBIA framework for identifying downed logs on the forest floor was effective but inefficient in terms of:

    • Substantial analyst time was required for visual interpretation and to derive the OBIA rule sets.

    • An automatic determination of an OBIA delineation rule set was not achieved.

    • The data structure of the lidar data makes homogenous log objects difficult to extract in an OBIA framework.

  5. Airborne lidar data coupled with OBIA techniques can be used to compliment field-based methods for identifying and characterizing understory downed logs in forests.

Full Reference: Blanchard, S. D., Jakubowski, M. K., and Kelly, M. 2011. Object-Based Image Analysis of Downed Logs in Disturbed Forested Landscapes Using Lidar. Remote Sensing. 3(11): 2420-2439.

The full paper is available here.

To learn more about lidar data, check out our lidar FAQs sheet. For more information on lidar data in SNAMP, see the spatial team website: http://snamp.cnr.berkeley.edu/teams/spatial, and our spatial team newsletters that focus on lidar: Vol. 2, No. 3, and Vol. 5, No. 1.

No Comments Yet

Post a New Comment