Automatic Extraction Of Crop Objects On Color Image In Complication Background

The greenhouse crop automatic recognition is one of the most important subjects in the fields of computer-vision and pattern recognition. Automatic extraction of crop objects is very important for crop growth parameters measuring, the simulation system of both green house control and crop growth process and undestroyed measuring.To the characters of greenhouse background image, an efficient algorithm is proposed to extract the crop object from green house background image. In order to improved the dynamic range and contrast of the image, histogram equilibrium and median filter are used to image strengthened and filter. So the detail of the image become clearer and the image is easier being segmented.In order to automatic segment the color image efficiently, two color spaces are used: HSV color space and CIE L~*a~*b~*color space. In the HSV space, H channels show image hue, we use histogram concept on H channels. Form the valley number we can get the main color part of the image. And this number will be used in next step. When we find the main color part of image, the fuzzy clustering is use to get segment the color image in the CIE L~*a~*b~*color space. In this method, the input are color image and the number for cluster, then the output is the center of clustering .By calculate the distance from the green color value(it can be got form a stand color block) to different clustering center, the closed one is the cluster we wanted. Then the green crop are found in the image. The crop object of greenhouse are automatic extracted successful.The experiments show that this method could automatic extract crop object from complex greenhouse back ground image. The method is efficient, accurate and validity.