Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

The Discrimination of Rice Weeds Based on DSP

Author FanDeYao
Tutor YaoQing
School Zhejiang University of Technology
Course Signal and Information Processing
Keywords DSP Rice weed Computer vision Image processing Feature extraction Weed recognition
CLC TP391.41
Type Master's thesis
Year 2011
Downloads 64
Quotes 0
Download Dissertation

The rice weeds bring great harm to the growth of rice; they have strong survivability as they had adapt the rice farming, climate, soil and environment. They compete for nutrients, water, light with rice, and weeds in rice field are easy to contribute to the spread of pests and diseases, the yield and quality of rice are great reduced, caused huge losses to rice.Face the serious damage of weeds in rice field, the mainly methods are manual weeding and chemical weeding at present. The manual weeding is low efficiency, labor-intensive, because the chemical weed can effectively control the weeds, it has become the mainly way of weeding in rice field. At present the method to weed is to spray the herbicide, and the means to spray is the well-distributed spraying. This method, not only improve the cost of agriculture but also damage the quality of field and pollute air. Study indicates that weed isn’t well-distributed, so we should find a variable-controlling spaying method. It is to spray in the region where there are weed patches, and to stop in the region where there are not. In order to do so, the first thing is to realize the detection of weed patches. In this paper we selected the DSP as the main processing chip, construct the rice weed recognition system, and realized the reorganization of rice weed region, marked the weed region in real time, and laid a foundation for the subsequent development of automatic variable spraying system. The main content and results are as follows:(1)Select TI’s TM320DM642 DSP as the main processing chip, constructed the hardware system of rice weed identification, and programmed the rice weed identification algorithm, implemented the rice weed identification and marking in real time.(2)Rearched the methods of extracting the region of interest on the images of rice field, and the color image analysis (2g-r-b), (2G-R-B), the color characteristics of the three color value H can be used to distinguish between green plants and background. For the segmentation of the background of rice fields, combine the color factor of Ext-green and Ext-red were proposed, and application the Otsu method to achieve automatic segmentation of the background of rice fields.(3)Research the image acquisition method of rice field, proposed the algorithm of rice weed segmentation according to the image acquisition method. Realized the identification of the rice weed region, and completed the rice weed region marking and area statistics.(4)In the feature extraction of the several rice weeds (Echinochloaphyllopogon Koss Sagittaria Pygmaea Miq, Ludwigia prostrata Roxb, Marsilea quadrifolia, Monochoria vaginalis Presl), the inhomogeneous quantification histogram was used to extract color characteristic and it enhanced the extraction efficiency and recognition robustness. The gray level co-occurrence matrix was used to extract texture characteristic and it compressed the grayscale of spot image. The computation was reduced by 3/4 and the complexity of feature extraction was also reduced. Three methods including Bayes discrimination, support vector machine and neural network were used to recognize five rice weeds by four feature sets (color set, shape set, texture feature set and all the feature set), the results showed that the BP neural-network got the best result in rice weed discrimination.

Related Dissertations
More Dissertations