Recognition And Localization Of Tomato Clusters In Shed Environment Based On Binocular Vision

At present,in the research of various kinds of agricultural picking robots,the identification and positioning of single fruit is mostly studied,and the identification and positioning of string or spike fruits and vegetables is less studied.In order to meet the demand of string tomato picking in greenhouse environment,reduce the cost of agricultural production and relieve the seasonal pressure of picking,this paper takes string tomato in greenhouse environment as the main research object.(1)In this paper,the noise reduction of the image of tomato string collected under different shooting conditions in the greenhouse environment was studied.This paper analyzes the noise characteristics of tomato images,analyzes and compares the noise reduction algorithms,and uses a better edge-preserving bilateral filtering algorithm combined with an improved adaptive median filtering algorithm to remove different kinds of noise.Taking PSNR and SSIM as the index,the experiment proves that this algorithm can remove noise and has better edge-preserving ability.(2)In order to improve the recognition accuracy of subsequent string tomato image shape features,image sharpening algorithm and wavelet enhancement algorithm were used to enhance the image edge contour information.At the same time to reduce the effects of light intensity characteristics of color segmentation,respectively using gray level transformation,histogram equalization and homomorphic filter algorithm for image illumination normalization processing,results showed that the gray level transformation can make the image based on the requirements’ dim’or’ lights’,homomorphic filtering algorithm can enhance image contrast,softly and after processing the image contour more clear.(3)The results show that the improved OTSU algorithm in this paper has a good effect on the segmentation of fruit image by using the Canny edge detection,mathematical morphological edge extraction and improved OTSU algorithm,respectively,based on the color feature information of the tomato and using the hue segmentation algorithm based on the CIE table color system which is more efficient than the color difference method.(4)Coordinate system transformation,camera imaging model,3d ranging principle of binocular vision system are analyzed,and the camera calibration based on Matlab is completed,which has the external conditions positioning research.The results show that this method can meet the requirement of operation under the condition of error.