Number of trees and canopy application of remote sensing by drone

Estimation of the number of trees and their canopy coverage in forest nursery fields is of great significance for the management of forest nursery. The calculation of the number of tree plants and their canopy coverage by conventional techniques is laborious and time consuming. It is of great practical significance to explore a new technology that is fast, efficient, low-cost and accurate.

In recent years, the remote sensing technology of UAV has developed rapidly. The application research results have been achieved in land remediation, land use planning, fine detection of mine geological disasters, investigation and evaluation of mountain flood disasters, calculation of building density and floor area ratio. However, in these applications, the drones used are mostly expensive aerial survey drones, and the data processing is complicated.
1 Experimental area and data

Figure 1 corrected drone image

The experimental area is two ginkgo forests in the Yudu District of Chengdu. The experimental area is flat. The data used is the image obtained by the DJI PHANTOM2 VISION+ drone. The image is a fisheye image, and the fisheye lens is corrected and geometrically corrected for the fisheye image. The corrected image is shown in Figure 1.

As can be seen from Figure 1, the image was taken in the fall and the ginkgo leaves were golden yellow. There are 3 ginkgo nursery fields in the image, and 2 of them are selected as test plots.

2 methods and steps

2.1 Constructing the difference yellow leaf index

Through the analysis of image features and luminance values, the yellow leaf information of the tree canopy can be increased by the combination of the difference between the red band (B3) and the blue band (B1). For this reason, we explored the construction of the difference yellow leaf index (HYZS). , its expression is as follows:

HYZS=B3-B1 (1)

Using the above expression, the difference yellow leaf index is calculated.

2.2 Canopy Coverage Information Extraction

The Ginkgo biloba crown extraction model was established using the difference yellow leaf index to establish the yellow leaf state, as follows:

If HYZS>40 then

The pixel is a ginkgo tree crown pixel, and its value is 1:

Else (2)

The pixel is a ginkgo tree crown pixel, and its value is 0: Endif

The above model was used to extract the ginkgo crown information, and the extraction results were comprehensively processed to remove small areas below 5 pixels. The result is shown in Figure 2. The red color is the ginkgo tree crown.

Figure 2 Ginkgo tree crown (red)

The extracted canopy result raster data is vectorized, and the polygon map covered by the canopy is obtained. The canopy map with the area above 1 m2 is selected, and the canopy is covered with the GIS database. The result is shown in Fig. 3.


Figure 3 Ginkgo tree crown covered polygon (red)

2.3 Analysis of canopy coverage of woodland plots

Digitize the number of nursery plots by digitizing the screen. There are 2 Ginkgo nursery plots in the experimental area to establish their GIS database. The spatial analysis of the plot layer and the canopy overlay GIS data layer is performed, and statistical analysis is performed on the basis of this, and the statistical result data is obtained, and the canopy coverage of each block is calculated, as shown in Table 1. The canopy coverage is the total area covered by all canopies in the plot, accounting for the total area of ​​the plot.

2.4 Estimation of the number of trees in the plot

In the canopy GIS database, add the number of plants field. Plot 1: Select a polygon whose canopy cover area is less than or equal to the average area (3.43m2). Each polygon is a canopy of a number. Therefore, the number of plants is assigned a value of 1. For a polygon whose area is larger than the average area, divide its area by 3.43 m2 and round it up, and use this number as the number of trees in the polygon. By inspection, the number of trees with 2 polygons is incorrectly calculated. One is miscalculated as 2 trees and the other is 3 trees miscalculated as 4 trees. After modification, the number of strains obtained was 79. Plot 2: According to the above method, the analysis and calculation were carried out also on the average area (3.43 m2), and the number of plants of the plot 2 was 83.

2.5 Accuracy analysis

In the case of canopy extraction, the extracted canopy is a ginkgo tree crown in the yellow leaf state, and therefore, the false extraction rate is zero. A small amount of ginkgo leaves have not been extracted due to their lack of yellowness, mainly in the edge of the canopy. By sampling, it is estimated that up to 5% is not extracted, so the leakage rate is 5%. Therefore, the overall accuracy of its extraction is high. The canopy coverage is a relative number, and its accuracy depends on the extraction precision of the canopy cover and the interpretation accuracy of the land extent. Because the block range is interpreted as visual interpretation, its accuracy j is higher. As far as the estimation of the number of plants is concerned, for the forest land plot 1, there is no leakage estimate. Two more strains were counted in the automatic extraction calculation. The number of estimates is 2.53% of the total number of 79. For plot 2, 83 strains were automatically extracted and the actual number of strains was 89. The number of missing strains was 6 and accounted for 6.74% of the total total of 89. Therefore, the accuracy of the automatic extraction estimation of the number of plants is high.

MMC:http://www.mmcuav.com/

Drone:http://www.mmcuav.com/drones/

Payloads:http://www.mmcuav.com/payloads/

Accessories:http://www.mmcuav.com/accessories/

GCS:http://www.mmcuav.com/gcs/