Study On The View Of Robot About Cotton Image Segmentation And Cotton View Navigation

With the development of science and technology, and the promotion and application of high-tech agriculture, agricultural robot is entered into the areas of agricultural production gradually, and it promoting the modernization of agriculture toward the intelligent direction changes. Xinjiang is a major cotton producing province, and the planting area is over one third of the country. So far, the mostly of cotton picking is still rely on manual labor in Xinjiang, it not only increased the farmers investment, but also reduces the efficiency of cotton picking. Thus, the appearance of cotton picker robot could solve the current predicament. This article focuses on the visual of picking robot to study, including the following aspects:(1) On cotton image segmentation. This paper proposes a way to segmenting cotton image which under the HSV color space to using two-dimensional Otsu threshold and geometric characteristics. This method could overcome the interference of light and shadow. And make a comparative analysis of the experimental results with the conventional segmentation methods in the number of bolls, boll size and split time, the result shows that the average number of bolls recognition accuracy is 92.68%, the average area of bolls segmentation accuracy is 94.31%, the time to identify an object of the image is 0.9637 s, it meet the requirements of cotton picker robot for recognition accuracy and real-time.(2) On cotton maturity of judgment. This paper proposes a way to extract the feature parameters which based on genetic algorithms, establishing the discrimination model of maturity to make the determination of cotton maturity. The method by extracting the minimum bounding rectangle of cotton, obtaining the characteristic parameters and Plug it into the discrimination model to get the cotton maturity level. The experiments shows that the discrimination model can judge the maturity level of cotton accurately, and can overcome the effects of wind, extrusion and other external environment.(3) On navigation. This paper proposes two navigations ways which are based on view and touch the bar. View navigation: first dealing with the seeding period of cotton field image, extract the target area based on the chromatic between the ground and cotton seeding, then based on the Hough transform to extract the centerline, finally extract the navigation centre line from establishing a dynamic winder for video image, the extraction results are basically same with the human eye’s visual effects. Touch the bar navigation: for natural cotton field, based on the sensor for parameters, comparing and analysis the character of parameters to get walk the path.Finally, doing a tracking simulation experiment to routes, the results show that with different speeds, track routes can find the pilot path quickly.