Solving bird image overlapping for automatic population counts of birds using image processing

Abstract

The counting bird population for assessment of the abundance of ecosystem is very important. The problem is overlapped bird in images. This research is purposed to solve bird image overlapping for counting population of birds using image processing. There are 4 techniques of image processing. Firstly, noisy dot and background are removed in bird image using multilevel thresholding. Secondly, the Region growing finds size of the bird by finding area value for segment of the image. Thirdly, the distance transform finds distance between images for segment of object. Finally, the erosion technique is used for object reduction which relationship of pixels in 4 directions. The experimental results have high precision as 0.96 and accuracy as 0.89 as recall 0.92 with different 6 images. The benefit of this research increases the performance of counting bird population which is one of the factors to discover an abundance of ecosystem.