Study On Rice Quality Inspection System

Rice is the primary food for the majority of the people in our country; the focus of rice production in our country is increasing the yield of rice, the research for enhancing the quality of rice and breeding the rice varieties ware rare in our country. Rice production was increasing in recent years, but the quality improved little. With the rising of living standard and improving of meal structure continuously, the quality of rice had put forward higher requirements. On the current rice market, high-quality rice price was higher, sold briskly, but some poor quality rice, although lower selling price, the market was still left out in the cold. In order to improve rice quality, it was necessary to improve the quality selection of rice varieties, but also to strengthen the quality inspection of the rice. The time of the research about the detection of rice quality in our country was late, the detection of the rice quality mainly relies on manual identification methods and sensory evaluation, which were impactied greater with subjective influence and operated cumbersome. In addition, its accuracy was low and the repeatability was bad. At present, for the status of the quality detection of domestic rice, this paper studied separately the rapid detection methods about the fatty acid content, moisture content, rice-shaped and appearance quality of rice, and developed a comprehensive evaluation system for detecting rice quality, inorder to make the inspection of rice quality more rapid, accurate, and objective.1. Studied the rapid detection method of fatty acid content of rice based on the colorimetric assay of the reverse micelles system, then develope a reverse micelle colorimetric reagents. The reverse micelles colorimetric reagents were made up of the reagents,.which include the reverse micelle system of AOT and isooctane, Tris reagent, phenol red reagent, concentrated hydrochloric acid. The reverse micelle colorimetric reagent has a characteristic wavelength of 574nm. The research about the impact of each component reagents of the colorimetric reagent was carried out using UV spectrophotometer. The best ratio of the colorimetric reagent components was optimized by using the Response Surface Methodology. Firstly, the Tris/HCl buffer solution and the phenol red Tris/HCl-buffer solution were mixed by 4:1 volume ratio to form the phenol red color reagents, then the 5mL AOT/isooctane reverse micelle solution with the concentration of 40mmol/L was added to the phenol red color reagents of 0.1mL to constitute a reverse micelle reagent colorimetric reagents. During the fatty acid content of rice was detected by the use of the reverse micelles colorimetric reagent, the change of the color value of colorimetric reagent was detected by using the color sensor, and then the corresponding relation was set up between the changes of the color value of colorimetric reagent and the value of the fatty acid content of rice, and the fatty acid content of rice was detected rapidly based on the change of the color value. The results of the experiments proved that the method can achieve an accuracy of 80.00 %, an error of 5%, and be able to meet the requirements of the rapid detection by simple operations and less times.2. Studied the rapid detection method of the moisture content in rice by the use of Near-Infrared instrument based on the GB assay of the rice moisture content. The rice samples were tested using the Near-Infrared instrument, and gained the diffuse reflectance optical detector values of the near infrared lights at the various wavelengths, and the moisture content of rice samples was detected using the GB method at the same time. Then the correlation between the diffuse reflectance optical detector values and the moisture content of rice was analysised using statistical software. Make the detection values of diffuse reflectance light at the nine wavelengths with the higher correlations as input values, and then a BP neural network would be built to detect the water content of rice, and compared with the test results of the near infrared regression equation detector. The test results show that using the BP neural network to detect the water content of rice with the input of the near-infrared diffuse reflectance optical detection values can achieve an accuracy of 96.67%.3. Studied the rice-shaped detection methods based on machine vision technology. Designed the rice appearance quality static detection system composed of detect boxes, detect units, light source, CCD image sensor, image acquisition card, stereomicroscope, camera and computer operating software. The special pre-processing algorithm were used to deal with the images of rice samples to obtain binary image of rice samples, then the area, perimeter, length, width and other characteristics of 16 parameters were extracted based on the rice-shaped characteristics of the rice, and treated the extracted characteristic parameters by using the main component analysis method, and the 3 former principal components would general all characteristics of the rice-shaped and the cumulative contribution rate was 93.36%. The whole rice and broken rice were detected by using the BP neural network which was built with the preceding three principal components as the characteristic values, the test results showed that the accurate identification rate of the whole-rice is 96.25%, and the broken rice is 92.75%.4. Studied rice processing accuracy by using gray-gradient co-occurrence matrix algorithm. The images of the samples of rice were preprocessed, and the gray-scale matrix and the gradient matrix of rice sample were extracred, and the texture feature values of rice processing precision were extracted and described according to the gray-gradient occurrence matrix. Then the processing precision of rice samples were detected by discriminate and analysis, fisher step by step and discriminate functions array. The experimental results showed that the accuracy rate using this method to detect four different kinds of samples of rice was 93.00%.5. Studied rice chalkiness based on the maximum between-class variance method (OTSU Law). The gray image of samples of rice was obtained. The threshold t of rice's gray-scale images by using OTSU Law was calculateed; the rice's gray-scale images according to the t value were cut up, and the chalky whiteness of rice was calculateed. Test the samples of four types of rice, judgment threshold t of chalky white and non-chalky white is 130, if t is less than or equal 130 and the chalkiness degree is more than 50%, it is the non-chalky rice, otherwise is chalkiness rice. By inspecting, the exact rate of using this method to detect the four kinds of rice chalkiness samples is 97.50%.6. Studied the detection algorithm of the rice appearance color based on the prior network. The target rice grain image was obtained by dealing with the images of rice; the extraction region of rice color value at the abdomen of rice grains was determined, the color value extraction region was divide into five parts according to the area, respectively the R, G, B color Eigen value of each region was extracted; using the color space conversion, the color value was converted from RGB model into HSI model, and the color characteristic value of rice was expressed by using five color values Hi. According to the color characteristics of rice, the forward neural network was built to detect the color of rice. The experimental results show that accuracy rate of color detection by using this method to detect different processing level's samples of rice can be achieved 92.00%.7. Designed rice appearance quality automatic detection system. Rice appearance quality automatic detection system was made up of vibration conveyor system, decline in trough, photoelectric sensors, Camera System, detection system platform and so on. Vibration conveyor system separated samples of rice and transmitd them in a forward line, then the rice samples accelerated along the decline in trough and slided into the photoelectric sensor at the end of decline in trough which the photoelectric sensor has response. The computer systems detected the photoelectric sensor's response and control camera system to get images of rice. Grain swaped out from the photoelectric sensor, and falled over on the conveyor belt of detect system platform, camera obtained images of rice when the rice run to the bottom of camera and sended the images to the software system for analysis and detection.8. Exploited the general detection system of rice quality based on Visual C++6.0. The rice quality inspection system consistd of Rapid detection of fatty acid content of rice subsystem, Rice water content for rapid detection subsystem and Rice appearance quality automatic detection subsystem, which could detect the fatty acids, Moisture content and Appearance quality of rice samples. The system was simple, and could analysis quickly with higher accuracy, would meet the requirements of rapid detection of quality rice and provide basis for the automatic detection of rice quality.