Reducing the effect of forest canopy to measure distances between trees using UAV image

Abstract

In the aerial images taken from the forests, the crowns of the trees are recorded. Using DN, texture and crown pattern, the type of trees can be predicted with a certain accuracy and used in forest studies. DN is the amount of light intensity that assigns a number to each component of digital images depending on the number of image bits. The presence of foliage causes the space under the canopy to be affected by shade and semi-shade, and these areas have close to zero DNs. The canopy of trees has less effect on aerial images in autumn and winter, but the need to study the forest throughout the year, in spring and summer will be a problem. On the other hand, some LIDAR and RADAR sensors have the ability to pass through the canopy of trees, but these sensors are not always available and their imaging cost is higher than UAVs that capture in the visible range. The need to use image processing methods is to be able to prevent the loss of information about the crown of trees in visible aerial images, and to extract information about the areas affected by the shadow and semi-shadow of the crown for interpretation. Due to the 8-bit image, in this research, 8 level of transfer have been done. At each step of the transition, the DN of each pixel is compared with the average DN of the image. DNs that are larger than the average are assigned a value of 1 and other DNs are assigned a value of 0. The output of this step is an image with an 8-digit binary number assigned to each pixel. By converting the binary number to decimal, the DNs of the image are reconstructed and the image is improved. Due to the increased structural similarity of the fused image (0.8762) compared to the original image (0.2783), we expect the image to improve. In the second step, the transferred pixels are multiplied by the new mean and then subtracted from the value of the old DNs. This creates an image where the tree trunks are dark and other parts of the image are light. In the next step, we open and close the image. This operation consists of two stages: Erosion and Dilation. Erosion is performed on a binary image and the remaining parts of the crown are shortened or thinned. This morphological operation works evenly on the whole image and makes the trunks of the trees thinner, so the Dilation operation is applied to remove the spots created in the image, creating large small gaps between the crown The trees are shown in the picture, which shows the trunks of the trees. Finally, by forming a graph between the trees, the distances between them are calculated in terms of pixels of the image.