Research On Key Technology Of Weed Detection Based On The Variable Spraying

Corn has an important position in agricultural production in the northeast of China, however, weeds have great harm on the growth of maize. They compete for nutrients, water, space, light with corn and they are easy to contribute to spread of pathogenic bacteria, increase the incidence of pests and diseases go against the promotion of agricultural, therefore, remove weeds on corn farm is very useful.At present, the main method of weeding is chemical weed control. Ordinary spraying machine widely used the way of vulgar and large spraying, not only increase the cost of agriculture but also bring the soil, environment, ecological problems such as pollution, this is not conducive to realize the sustainable development of agriculture, the study of high efficiency, saving energy variable spraying technology is particularly important. In order to realize the variable spraying, field weed identification is the first key point. This paper uses the digital image processing technology, studied about weed identification method of variable spraying technology, the main content of this paper are as follows:(1)In the image pre-processing phase, has carried on the research of based algorithm in the view of the image gray level, denoising, binarization and image segmentation. At each stage, choosing the appropriate weed identification method by comparison and analysis of algorithms.(2)Image edge detection algorithms, there are many ways to detect the edges, and each has its own advantages and disadvantages, because the image obtained under the complex natural conditions, this paper compared with the commonly used method of image edge detection, proposed a fast edge detection algorithm based on fuzzy enhancement, and through many experiments verified the fast effectiveness of the algorithm, overcome the cost time, ignoring fuzziness, image fine point positioning is not accurate, edge information loss and the other disadvantages which is often happened in the traditional edge detection algorithm.(3)The growth of weeds is irregular, in order to facilitate the research, divided the weeds into the inter row weeds and weeds in lines two conditions for recognition. In this paper, adopted to the method based on location characteristics to identify inter-row weeds, first determining crop center line and fill the center line, then distinguish the corn and weeds. In determining the center row crop, Hough transform algorithm is used, according to the actual situation of the corn crop seeds with fixed line width, filling the center line on the basis of line width, realize the distinction between crop and weed, complete inter-row weeds identification.(4)Inline weed identification, for weeds inline, in this paper the identification method is based on shape feature, for the four common weeds, select the main statistical about area, perimeter, long width, circular degree and degree of discrete parameters. Because a single parameter of shape features can not accurately finish the task, this paper collect the 5 characteristic parameters, get the different ranges of different characteristic parameters, then choose different membership functions of different structures, by adopting fuzzy theory in the weed recognition to achieving the identification of inline weeds.In this thesis, both inter row weeds and weeds in lines have been identified, The research of weed identification had guide and practical significance for the development of agricultural modernization, suitable applying of herbicide, and the protection of the ecological environment.