雉鸡亚种种群的图像识别研究

雉鸡的亚种众多,在其表型与遗传上具有地理变异。雉鸡的一些亚种适应于不同的地理条件(例如:寒带、热带、干旱区、高原)形成了独特的群体形态特征,尤其体现在雄性个体的羽色图案上。羽色图案的复杂性给雉鸡亚种的人工鉴定带来了挑战,为此,本研究探索应用图像识别技术对雉鸡亚种进行识别。对3种识别算法:K近邻分类算法,概率神经网路和符号分类器进行了比较。通过序列前向选择法、序列后向选择法、序列浮动前向选择法和序列浮动后向选择法等算法对差异纹理特征进行提取和比较。本研究对25个雉鸡亚种的高质量博物馆标本的图片进行分析验证。结果表明,目前的图像识别技术只能识别羽色图案差异较大的5个类群,其中概率神经网路具有最高的雉鸡亚种识别鉴定效率。

Image Recognition in the Common Pheasant(Phasianus colchicus)Subspecies

There are many subspecies of common pheasant,and there are geographical variations in their phenotypes and genetics.Some subspecies of common pheasant have formed unique morphological characteristics to adapt to different conditions(e.g.,cold,heat,aridity,plateau),which is reflected in the feather color patterns of male birds.The complexity of feather color patterns poses challenges to the manual identification of common pheasant subspecies.For this reason,we explored the application of image recognition technology to identify common pheasant subspecies.We compared three recognition algorithms,viz.K nearest neighbor classification algorithm,probabilistic neural network,and symbol classifier.We extracted various textural features and compared them using algorithms such as sequence forward selection,sequence backward selection,sequence floating forward selection,and sequence floating backward selection.We analyzed and verified the pictures of high-quality museum specimens of 25 common pheasant subspecies.The tested image recognition technology identified five groups characterized by large differences in feather color patterns.The probabilistic neural network yielded the highest identification efficiency.