การระบุสายพันธุ์ของต้นไม้โดยการใช้ภาพจากอากาศยานไร้คนขับและการแบ่งส่วนด้วยความหมายโดยการเรียนรู้เชิงลึก

Abstract

Electricity from coal is an important source of energy for the country, accounting for 15% of its energy production. However, coal mining has environmental impacts as it requires the destruction of soil and rainforest areas. After the cessation of coal mining, a reforestation project was implemented in the area. Subsequently, issues arose regarding the measurement of the health of the replacement trees. Thus, a tool was needed to predict and identify the location of tree species used for reforestation, which would help with future monitoring of the health of the replacement forest. This thesis presents a method to predict and identify the location of tree species, specifically two types: the teak tree and the salao tree, in the reforestation area of the Mae Moh lignite mine rehabilitation zone in Mae Moh District, Lampang province. These experiments were conducted for all tree species and all month, different tree species and all month, each tree specise within the same month, tree species in a given month and for same tree species within the following month. The goal is to find the most suitable deep learning model for assessing the future forest's fertility after reforestation. From the results of experiments, it was found that deep learning models can identify the species and areas of teak and salao trees by measuring the weighted-f1 score, which was greater than 0.80 in all experiments.