基于定标模型云共享的奶牛粪水微型NIR现场速测系统

针对还田利用中粪水成分现场获取困难的问题,该研究设计开发了基于定标模型云共享的奶牛粪水微型近红外（Near-Infrared,NIR）现场速测系统,并对系统进行了验证。该系统主要包括微型近红外光谱传感器、云服务器、Android客户端。微型近红外传感器采集被测样品光谱数据,通过蓝牙协议将数据传输给Android客户端,再通过移动网络将光谱数据传送到云服务器;云服务器利用部署在云端的定标模型对接收到光谱数据进行计算、分析得到定量预测结果,并将预测结果回送Android客户端,可以实现总氮（Total Nitrogen,TN）、总磷（Total Phosphorus,TP）、总钾（Total Potassium,TK）、铵态氮（Ammonium Nitrogen,NH4+-N）、硝态氮（Nitrate Nitrogen,NO3--N）、酰胺态氮（Amide Nitrogen,CONH2-N）,有效磷（Available Phosphorus,AP）、有效钾（Available Potassium,AK）、有机质（Organic Matter,OM）、pH值共10种粪水成分的快速检测。结果表明：对粪水TN、TP、TK、NH4+-N、NO3--N、CONH2-N、AP、AK、OM、pH值的预测相对误差分别约为9.251%、4.261%、8.238%、8.906%、17.825%、15.123%、9.829%、5.507%、10.558%、2.969%。微型近红外光谱技术结合定标模型云共享能够实现粪水中多种成分的现场速测,完成了定量模型的资源共享,且因微型近红外光谱传感器的便携性、Android客户端的操作简便性、定标模型云共享的低成本和无需用户具备专业知识要求等优点,具有广阔的市场前景和重要的现实应用价值。

Micro NIR on-site and rapid detection system for cow manure slurry based on cloud sharing of calibration model

Abstract: Manure slurry is a mixture of urine, feces, flushing water, and disinfectant in the livestock and poultry breeding industry. About two billion tons of manure slurry can be produced in China every year. Among them, the composition varies greatly, due to the complex sources. There are many influencing factors on the composition of manure slurry, such as the seasons, regions, breeding scale, fecal cleaning, and manure slurry treatment. Currently, the fixed composition value cannot accurately be calculated the amount of manure slurry, when returning to the field. Therefore, it is urgent to develop an accurate, and rapid detection system suitable for the compositions in the manure slurry on site. Fortunately, Near-Infrared (NIR) spectroscopy can offer a great potential to detect manure composition at present. Nevertheless, most reports were usually focused on the near-infrared spectrometer at the large-scale laboratory level. This kind of instrument can be confined to the application of near-infrared spectroscopy analysis in fields, due to the large bulk volume, low portability, and high price. The near-infrared spectrometer can be further developed towards the miniaturization for the technical and cost feasibility during on-site detection, particularly with the development of Micro-Electromechanical Systems (MEMS) and Micro-Opto-Electro-Mechanical Systems (MOEMS) in recent years. In addition to the need for a stable and reliable hardware system, the calibration model is another important application premise of near-infrared technology. But, there is a great challenge to establish the calibration model using cloud sharing technology. It is very necessary to access the shared resources via the various data computing services anytime, anywhere, and on-demand through the Internet. In this study, a micro NIR onsite and rapid system was proposed to detect the composition of manure slurry during field return using a calibration model under cloud sharing. A complete function was also achieved for the data acquisition, upload, prediction and storage at the same time, according to the design scheme of "micro NIR sensor + calibration model cloud sharing + Android client + mobile network". The micro NIR sensor was first used to collect the spectral data of the measured sample, then to transmit the data into the Android client through the Bluetooth protocol, and finally to the cloud server through the mobile network. A calibration model was deployed in the cloud server to calculate the received spectral data. A quantitative prediction was obtained to further send back to the Android client in real time. A rapid detection was realized for the ten parameters of manure slurry compositions, such as the Total Nitrogen (TN), Total Phosphorus (TP), Total Potassium (TK), ammonium nitrogen (NH4+-N), nitrate nitrogen (NO3--N), amide nitrogen (CONH2-N), Available Phosphorus (AP), Available Potassium (AK), Organic Matter (OM), and pH value. The prediction relative errors of AK, TN, AP, TP, pH, NH4+-N, and TK in the manure slurry were less than 10%, and the rest of OM, CONH2-N, and NO3--N were between 10%-18%. Consequently, the micro NIR technology combined with the calibration model in cloud sharing can be expected to realize the onsite, rapid and accurate detection of various compositions in the manure slurry, particularly for the full sharing of the quantitative model. Moreover, the broad market prospect can be gained for the practical application, due to the portable micro NIR sensor, the simple operation of the Android client, and the lower cost of the system. There is also no need for the professional requirements of the users in the cloud sharing of the calibration model. The finding can provide data support for the high precision return of manure slurry to the field.