• 研究简报 •

### 基于FCM和分水岭算法的无人机影像中林分因子提取

1. 东北林业大学信息与计算机工程学院 哈尔滨 150040
• 收稿日期:2018-03-26 修回日期:2018-12-14 出版日期:2019-05-25 发布日期:2019-05-20
• 基金资助:
中央高校科研业务费（2572018BH02）；林业公益性行业科研专项（201504307-03）。

### Extraction of Stand Factors in UAV Image Based on FCM and Watershed Algorithm

Li Dan, Zhang Junjie, Zhao Mengxi

1. 1. College of Information and Computer Engineering, Northeast Forestry University Harbin 150040
• Received:2018-03-26 Revised:2018-12-14 Online:2019-05-25 Published:2019-05-20

Abstract: [Objective] The purpose of the investigation and monitoring of forest resources is to identify and implement the quantity and quality of the national forest resources, macro grasp of the development and change of forest resources, and provide data support for the sustainable development of national forest resources, it is the foundation of the management of national forest resources.[Method] This paper takes the Pinus sylvestris var. mongolica plantation as the research object in the city forestry demonstration base of Northeast Forestry University, using the multi rotor unmanned aerial vehicle(UAV)DOM as the data source, applying FCM clustering algorithm, watershed segmentation algorithm and a series of digital image processing technologies such as morphological operation, threshold segmentation, image smoothing, gray image and binary, extracting the stand factors of Pinus sylvestris var. mongolica plantation. FCM clustering algorithm and threshold segmentation method is used to extract treetop markings, then the watershed segmentation algorithm is used to iterate the treetop image, and the single tree crown segmentation image is obtained. According to the result of single tree crown segmentation, the characteristics of single tree are extracted and then the value of each stand factor is calculated.[Result] In the module of forestland extraction, the greenness segmentation successfully separates the forestland from the non-forestland, according to the color characteristics of the image. It determines the range of the single tree crown segmentation. In the module of single tree crown segmentation, both threshold segmentation and FCM clustering algorithm can be used to extract the treetop markers from the forestland image effectively. It has achieved good segmentation effect that applying watershed segmentation algorithm based on marker to single tree crown segmentation, most of the single tree crowns are separated from each other, but some areas still have problem of less segmentation or over segmentation. The stand factors include canopy density, number density, average crown width, average DBH, average tree height and volume. The measurement accuracy of the canopy density is 96.67%, the measurement accuracy of the woodland area is 81.23%, the measurement accuracies of the stumpage number and average crown width are related to the treetop extraction method and the two parameters(the size of structural elements of morphological corrosion and the window size of median filter)in the watershed segmentation. Parameter combination experiments on two method of treetop extraction are carried out respectively, the result show that the measurement accuracies of the stand factors of the two treetop extraction method using the proper combination of parameters are all above 80%, the average measurement accuracies are all above 90%, the maximum average measurement accuracy of the threshold segmentation method is 94.49%, the maximum average measurement accuracy of the FCM clustering algorithm is 93.17%.[Conclusion] The method of forest resource investigation by using the orthoimage of artificial forest taken by UAV is presented in this paper, which embodies the information construction of forestry. The application of advanced computer science and technology and unmanned aerial vehicle technology to the traditional field of forestry has effectively improved the efficiency and accuracy of forest resource investigation. The method proposed in this paper is suitable for the extraction of the stand factors of high canopy density forest, and the measurement accuracy meets the actual demand.