Research On Corn Seeding Stage Weeds Detecting Technique Based On DSP

At the conclusion of this article at home and abroad in the field machine vision-based feature extraction and identification of weeds based on the results of research on China's major agricultural crops of field corn, feature extraction and identification of weeds were studied. In this study, as the main processing chip DSP, set up the hardware and software platform for image processing, and for corn weed seedling characteristics, the corresponding algorithm, the recognition rate of 95.0%, and automatic follow-up weeding of agricultural systems to lay a certain amount of technical base. The main thesis and conclusions of the study are as followings:(1) Selected TI's TMS320DM642 DSP as the main processing chip, the company's union reached SEED-VPM642 as a platform for image processing hardware and software into the system initialization, image acquisition and storage, image processing, image display four modules to complete the image processing system design.(2) Introduced the use of DSP methods of extracting the region of interest, and the color image analysis (2g-r-b), (2G-R-B), the color characteristics of the three color value H can be used to to distinguish between green plants and background. Indoor Relative light intensity in the relatively stable condition, the use of (2G-R-B) in terms of color feature is simple, fast processing speed(3) The use of (2G-R-B) gray-scale image, and through experiments chosen adaptable, fast iterative adaptive thresholding method for image segmentation threshold can be very good to be green from the background in differently. Overlap is not serious for the plant by the number of de-noising and morphological operation; you can get a complete picture of the leaves.(4) For the bending phenomenon of improved leaf blade length and width for the algorithm, the gray-scale image and contour image of the shape of the leaves extracted parameters, and an analysis of dimensionless parameters can be drawn from the results: length and width moment invariants can be a corn, narrow-leaf weeds, broad leaf weeds in three main categories distinguish between plants; circular degree can narrow leaf weeds in the area from the three separate categories; and three types of rectangular degrees serious overlap with the rectangular shape is not a valid degree characteristics. (5) Divided 180 samples into training set and test set, defined by orthogonal experiments the structure of BP neural network, using the trained BP neural network validation experiment showed that the recognition rate of training set was 99.2%, test set recognition rate was 95.0%.