碩士 / 國立屏東科技大學 / 機械工程系 / 94 / The purpose of this study tries to discuss the possible techniques of machine vision system applied to fish counting based on the image processing technique. The content of the present study is focused on the exploration of image processing method. It was intended as an economical, fast and accurate tool for the general fish farmers. The steps of image processing methods for accurately counting fish are : (1).Inpute image graphic, (2).Binarying image, (3).Noisy dots removing, (4).Recursion counting, (5).Statistics revision.
In an experiment we use the present system to estimate known numbers (14, 28, 42, 56, 70, 84, 98, 111, 125, 136) of Paracheirodon innesi in a fish container that is 120mm in depth and 210mm in diameter. Each number was shot with camera at a depth of 15, 30, 45, 60, 75, 90 and 105mm.
The result shows that the average rate of accuracy with different samples is from 94.90% to 69.54%. The more the number is, the higher the overlapping rate is, and the lower the accuracy rate is.
The average rate of accuracy in different depths is from 64.47% to 79.69%. The deeper the water is, the lower the overlapping rate of the fish, and the higher the rate of accuracy is.
By calculating the Z Score of each fish's picture pixel in each number, the problem of overlapping fish counting could be revised. The average accuracy obviously increases 2~9% after revision, from 96.94% to 79.52% .
Keywords: Machine vision, fish counting, Image processing
| Identifer | oai:union.ndltd.org:TW/094NPUST489021 |
| Date | January 2006 |
| Creators | Huang, Ku Sung, 黃谷松 |
| Contributors | Lin, Yi Hong, Hsieh, Ching Chen, 林宜弘, 謝欽城 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 117 |
Page generated in 0.002 seconds