Obstacles Detection Algorithm in Forest based on Multi-sensor Data Fusion

Abstract

In this paper an obstacle detection system was set up and a novel algorithm was proposed to detect obstacles in forest, aiming to improve the efficiency of automated operations and reduce the risk to cause the accident of forestry harvester in the complicated environment of forest areas. First a 2D laser scanner and an infrared thermal imager were used to collect the information of obstacles in forest areas. Second the features of each obstacle including temperature, color, aspect ratio, and rectangularity, were extracted from the laser points, visible images and infrared images. Then, we use the sample feature data to train a support vector machine (SVM) for object identification. The experiments were conducted on the various obstacles and the experimental results showed that the proposed algorithm was effective and provided a computation basis for automated operations of harvester