Unmanned aerial vehicles (UAV) systems could provide low-cost and high spatial resolution aerial images. These features and ease of operation make it a practical tool for applications in precision agriculture and horticulture. This paper highlights the application of UAV systems in tree counting, which is vital for tree inventory management and yield estimation. In this paper, two types of trees were discussed. One type is with non-uniform canopy area (e.g. container plants and citrus trees) and another is with uniform canopy area (e.g. Christmas). For the first type, aerial images obtained from 100 container-grown green and yellow plants were acquired with a stable, ground-based boom truck as a pretest. Two different index sets,  (2*G-R-B) and (V-H)/(V+H) were used to extract the green and yellow container plants from background, respectively. A counting algorithm based on average canopy area was developed to estimate the plant count. The effect of capturing heights and adjacent plant distance on the accuracy of the algorithm were discussed. Further, we had success on applying the algorithm on stitched images from the video files. The developed algorithm was able to count the trees with 2% accuracy. This also indicates that the algorithm could work well on low-resolution images. For the second type, we developed a counting algorithm based on a local maximum value at or near the center of coniferous tree. A 3-D intensity distribution of the images showed that local maximum of intensity of (r-g) matched well with the tree centers.  Also, Minimum distance filter (MDF) and threshold generated from histogram were used to remove the false identified tree locations.