基于改进主成分分析和AdaBoost算法的运动虾苗识别方法

发布时间：2018-02-08 15:23

本文关键词： 图像识别识别算法运动虾苗改进PCA Ada Boost 行为量化　出处：《农业工程学报》2017年01期 　论文类型：期刊论文

【摘要】：针对虾行为量化过程中运动虾苗较难检测与识别的问题,该文以南美白对虾虾苗为例,提出了一种基于改进主成分分析(principal component analysis,PCA)+AdaBoost算法的运动虾苗自动识别方法。在室内自然光条件下,利用工业相机采集承装容器中虾苗的灰度图像。提取图像中大小为100×100像素的不同运动状态的虾苗图像,首先使用改进PCA算法进行主成分分析,并进行特征提取。根据特征参数的分布情况,对其进行归一化处理,利用归一化的特征构建多个弱分类器,利用Adaboost方法将弱分类器构建成强分类器。最后,利用强分类器对运动虾苗进行识别。试验结果表明,在150幅不同运动状态虾苗测试样本中,基于改进PCA+Adaboost方法的识别正确率98%,平均每个样本识别时间为0.027 898 s,满足行为量化中的自动识别要求。
[Abstract]:In order to solve the problem that it is difficult to detect and identify the movement shrimp seedlings in the process of behavior quantification, this paper takes the shrimp seedling of Penaeus vannamei as an example. An automatic recognition method for moving shrimp seedlings based on improved principal component analysis (PCA) principal component analysis (PCA) AdaBoost algorithm is proposed in this paper. The grayscale image of shrimp seedling in container was collected by industrial camera. The image of shrimp seedling with different motion states was extracted from the image with 100 脳 100 pixels. Firstly, the improved PCA algorithm was used for principal component analysis. According to the distribution of feature parameters, it is normalized to construct several weak classifiers using normalized features, and the weak classifier is constructed into strong classifier by Adaboost method. A strong classifier was used to identify the moving shrimp seedlings. The experimental results showed that in 150 samples of shrimp seedlings in different moving states, Based on the improved PCA Adaboost method, the recognition accuracy is 98 and the average recognition time per sample is 0.027 898 s, which meets the requirement of automatic recognition in behavioral quantization.
【作者单位】：中国水产科学研究院渔业机械仪器研究所;
【基金】：中国水产科学研究院中央级公益性科研院所基本科研业务费专项资金项目(2015A02XK06) 国家现代农业产业技术体系项目(CARS-47)
【分类号】：S917.4;TP391.41

【相似文献】