• The function associates an equal number of pixels per constant grayscale interval • Takes full advantage of the available shades of gray • Enhances contrast of the image without modifying the structure Morphology - Top-Hat Transform Objects from an uneven background • Structuring element size should be bigger than the objects to be detected Deblurring • Remove focus blur • Remove motion blur Frequency Filtering • Remove noise • Remove repetitive patterns
UAV V/STOL with Augmented Co-Axial Power System EML4905 - Senior Design Project Spring 2009 Florida International University Department of Mechanical and Materials Engineering - 36 - The choice of the algorithm is base on the characteristic of the image that will be processed. 2. Geometric shape segmentation Different methods for shape segmentation • Edge detection Edge detection shows the edges in the image. In our case, it will show the shape edge and the character edge (under the best condition, usually come with noises). • Boundary detection based on threshold It can return the boundary of the parent object (the shape) and the child object (the character). This one is better for our project, which can separate the shape and the character directly. 3. Shape recognition • Geometric template matching • Transformation methods 4. Character segmentation Through the boundary of the shape, the points inside the shape are located. Since the contrast between the shape and the character is obviously, the character can be separated easily. 5. Character recognition The most common method is to utilize some training methods, neural network or support vector machine. It is important to note most of those character recognition applications are not required to deal with the rotation of the image since it is assumed that the frames are taken always with the same angle and overall positioning. In our case, the rotation of the character
UAV V/STOL with Augmented Co-Axial Power System EML4905 - Senior Design Project Spring 2009 Florida International University Department of Mechanical and Materials Engineering - 37 - is required and will be taking in consideration for any future work in developing the vision recognition system. In the case of UAV surveillance the pictures can be taken from different angles and different elevation, therefore, it is crucial to employ a vision recognition system that can analyze an image in rotational patterns that can be taken from different angles. The following images are an example of how the character recognition coding developed for this project is able to take a given picture and perform boundary recognition of the shapes inside the image. Figure 15: Vision Recognition Test - Original image to be process Figure 16: Vision Recognition Test – Image transferred to grayscale
UAV V/STOL with Augmented Co-Axial Power System EML4905 - Senior Design Project Spring 2009 Florida International University Department of Mechanical and Materials Engineering - 38 - Figure 17: Vision Recognition Test - Results for boundary recognition testing Figure 18: Vision Recognition Test - Negative of the image created to enhance contrast
You've reached the end of your free preview.
Want to read all 177 pages?
- Fall '08
- Mechanical Engineering, ........., Unmanned aerial vehicle