Effectiveness of computer vision system and back propagation neural network in poultry stunning prediction

被引：0

作者：

Ye, Changwen ^{[1
]}

Yousaf, Khurram ^{[1
]}

Zhao, Yang ^{[1
]}

Kang, Rui ^{[1
]}

Pang, Bin ^{[2
]}

Chen, Kunjie ^{[1
]}

机构：

[1] College of Engineering, Nanjing Agricultural University, Nanjing,210031, China

[2] College of Food Science and Engineering, Qingdao Agricultural University, Qingdao,266109, China

来源：

International Agricultural Engineering Journal | 2018年 / 27卷 / 01期

关键词：

Image segmentation - Binary images - Backpropagation - Torsional stress - Neural networks;

D O I：

暂无

中图分类号：

学科分类号：

摘要：

Electrical stunning is a vital part of poultry slaughter processing. In order to accurately recognize poultry electrical stunning level, a new hybrid classification method based on computer vision system (CVS) and back propagation neural network (BP-NN) was proposed. First, 300 same size sample images (320×240) were acquired by CVS. Secondly, image graying; Otsu’s thresholding, logic image operation and small regions removing methods were conducted on sample images to identify the region of interest (ROI) of final binary images. Three kinds of features, representing the total area (F1), the maximum area (F2) and the number of unconnected regions (F3) were extracted. Thirdly, a BP-NN was trained to simulate and predict the stunning level for poultry before killing. Linear fitting models were also developed to relate the stunning level and individual feature. Although the linear models of the total area, maximum area and the number of unconnected regions were significantly (P © 2018, Asian Association for Agricultural Engineering. All rights reserved.

引用

页码：289 / 297