实时采集中异常值的自动甄别与纠错方法研究

随着科学技术进步，规模猪的生长过程研究手段也日益现代化。传统上对猪的生长研究大多采用人工收集数据，不仅麻烦费事，而且极易产生猪的应激反应，对猪的生长产生影响。随着各种传感器和现代通讯技术在养猪事业中的应用，数据的收集变得更加科学、方便。然而，由于采集数据对象的特殊性，使得在采集数据的过程中，客观存在挤、拱、撞等现象，从而造成采集的数据存在偏差，对后期分析研究猪的生长性状产生一定影响，因此，必需加以修正。鉴于此提出了将经典算法与神经网络方法相结合来自动甄别与纠正采集的数据，通过Matlab仿真及在安徽菩提果公司研发的“9SC-05猪用选种选料自动测定设备系统”的应用实践，表明该方法具有纠错准确率高、速度快、适应性好等优点。

Study of method of automatically abnormalvalue discrimination and error correction in real-time data acquisition system

With the progress of science and technology, the research method of scale pig’s growth process is also increasingly modern. Traditionally, manual methods of data collection are mostly used to study the pig’s growth, which are not only troublesome, but also easily cause pig’s stress reaction, thus affect the normal growth of pig. With the application of various sensors and modern communication technology in pig industry, the data collection methods are becoming more and more scientific and convenient. However, due to the particularity of collect objects, there objectively exist all kinds of phenomenons, such as crowding, arching and colliding, in the process of collecting data, which may make the data misaligned and subsequent analysis and research in growth characteristics of pig influenced, therefore, the data collected must be corrected. So, this paper puts forward the classical algorithm combined with neural network to automatically identify and correct the data collected. Through the simulation of Matlab and Anhui bodhi fruits company’s application practice in choosing pig breeds and feed in automatic measuring system, it is proved that this method is of error correction accuracy, high speed and high adaptability.