Development Of Grain Protein Near-Infrared Detector

The accurate and fast inspect to the quality of grain is an extremely essential segment regarding our grain production and market transaction, but at present for most grain enterprises of buying and selling in our country the grain detection still pauses in the stage of hand pinching and nipping. The subjective factor direct influences the detection results. Although our government emphasized the grain must be graded according to the Law of Volume and Weight, however, the law only has a rough estimation to the maturity and moisture content of the grain. The real quality information of the grain cannot be obtained accurately.Our grain departments unceasingly encourage the adoption and implementation of the advanced scientific food detection method and testing technology, and continuously improve the standard of detection of grain. The grain are divided into different grades so that the quality scale appeared based on the grain intrinsic quality which provide the technical basis for the insurance of qualitative grain purchasing and higher grades, higher price. In recent years, in the analysis domain, the near-infrared spectrum technology which developed quickly can be rated as Green Detecting Technology for its fast speed, lower cost, non-destructive samples without pretreatment, departing from chemical reagents. It provides a new method for non-destructive testing of material multi-ingredients content.The topic group, on the basis of the scene quality grading detecting system of the existing JDJ-1000, attempts to add the function of protein detection. The main researches are as follow: (1) Under the analysis of the domestic and foreign Near-infrared detector and technology and diffuse reflectance spectra analysis principle, combining with the features of JDJ-1000 quality grading detection, a set of grain protein Near-infrared detector is developed, which is composed of the near-infrared emitter, sensing signal detection modules, a singlechip data processing and analytical module, and software component. The source of light introduces three near-infrared light emitting diode (LED) of 1680nm, 1880nm and 2180nm wavelength respectively. The detector of light signal uses PD24-03 photodiode produced by Russian IBSG Corporation. This instrument has the features of simple structure, small volume, the ease of operation, and has no moving parts.(2) The structure and circuit design of grain protein near-infrared detector. Optimize the structure design of the path of rays of near-infrared spectrum detector in order to reject the mixed light or the disturbance of useless signals. Three Near-infrared LED illuminate by 45°oblique incidences direct at the central part of the sample surface. The detector which is installed above the central part of the sample receives diffuse reflectance spectrum through convex lens. The modular design of examining circuit ensures the stability of work, the accuracy of detection, and line anti-jamming ability. The circuit mainly includes five pieces of modul: a singlechip processor circuit, source control circuit, signal collecting and switching circuit, communicative circuit in series, and other auxiliary circuit.(3) Development of software work of the grain protein near-infrared detector. The software includes a low-position machine process and a high-position machine process (PC). The process of the low-position machine includes master process module, source control program module, signal collecting and switching module, displaying program module, and completes data acquisition, data processing, data communicative functions, and being compiled in assembly language. High-position machine program includes communication procedure module, data acquisition modules, data processing module, the completion of data storage, data analysis, results output demonstration, which are developed in the software platform of Visual Basic 6.0. After receiving the instruction from the PC, the low-position machine transfers the corresponding program module to carry on the work, and sends out the pulse signal to actuate the work of Near-infrared LED, simultaneously collects the frequency code data, and then transmits the data back to high-position machine (PC). After that the PC carries on the data storage, analysis, calculating and sends back the results to the low-position machine.(4) Grain protein Near-Infrared quantitative analysis model and sample database establishment. Taking the computer technology as a foundation, we can establish a quantitative analysis model of several algorithms and then choose the best one. Obtaining the protein content of thirty five samples by the Kaishi Fixation Method of Nitrogen, twenty seven samples are choosed as the adjustment collection, and the other eight samples as the testing collection. After scanning the spectrum of the twenty seven samples in the detector, we establish the quantitative analysis model the way of Multiple Linear Regressions (MLR) and BP nurnal network respectively, and carry on the forecast to all samples. The result is that the correlation coefficient R2 between the value forecasted by multiple linear regressions of the testing collection and the chemistry value is 0.6491, and the correlation coefficient R2 between the value forecasted by BP nurnal network of the testing collection and the chemistry value is 0.9372. The forecast effect of the BP nurnal network is better than the multiple linear regressions.(5) The experimental study of grain protein near-infrared detection. The topic group has carried out the detecting experiment to each essential part and the overall performance of the detector. The results show that the voltage supplying for LED is stable, driving circuit is reliable, the noise level is low. Each time before collecting the spectrum the detector gathers and subtracts its own useless signals so that reduces the influence because of its own drifting and temperature. The detecting repeatability, the reappeared samples and the accuracy of equipment confirmed the operating performance. The detecting repeatability shows us a satisfied result. Each sample forecast results varied greatly, but taking the average value of many times is possible to improve the accuracy.In sum, this paper introduces us the design of a set of grain protein Near-infrared detector which can meet the needs of fast detection of grain protein, and expand the testing function of the scene quality grading system of the grain type JDJ-1000. The ring-like array distribution form greatly simplified the light structure and improved the light collecting effect of sensor. The design of the software and hardware of the circuit modules enhanced the working efficiency of the detector. The limited error during repeated detections and the accuracy of detection through several times'sampling made the design meet the anticipated requirements.