Research On Machine Vision Detection System For Grafting Machine

With the development of agricultural equipment intellectualized and refined, machinevision is widely used in agricultural operations as the most useful sensor obtaining informationfor agricultural machinery, Machine vision used in grafting robot can increase the graftingspeed and improve the survival rate of grafted seedlings. According to the informationextraction technical problem of grafting seedling for the cut-pasted grafting operations, thisthesis focuses on the research of the Machine vision detection system of grafting robots.Obtain the data supporting for the construction of machine vision detection system and thedevelopment of the characteristic parameter extraction algorithm for grafting seedling. Selectwatermelon as scion and gourd as stock, and measure their parameters in need under certaingrowth conditions. Analyze the measured data, the ellipticity of watermelon s and gourd scotyledon is0.69and0.56, the ellipticity of watermelon s and gourd s stem is0.75and0.86.According to observing change rule of cotyledon s angle by time in a day, the cotyledon sangle is greater than120°from10:00to15:00, which is suitable for collecting the pictures ofthe cotyledon.A machine vision detection method was presented, which can detect the grafting seedlingsplanted in plug nondestructively and on-line. Based on the size of the plug and of the seedling,the key components of the vision system is chosen reasonably, and a reconfigurable machinevision detection installation is established.On the base of the agricultural technical requirements and the grafting condition ofgrafting robots, the seedlings stem dimensions and the cotyledon information of graftingseedlings must be obtained.By dealing with gourd stem image captured from its front view, the image processingalgorithm for the stem parameters is exploited. The experimental results show that theparameters extraction success rate of stem is99%, which s timeliness meet grafting demand.The relative error s mathematical expectation of the gourd stem s height，long/minor axisbetween the machine measurement and the manual measurement is4.16%、6.57%、 6.48%,which indicate that seedling stems parameter extraction algorithm with highermeasurement accuracy meet the grafting accuracy requirements.By dealing with cotyledon image captured from its vertical view, the image processingalgorithm for the cotyledon parameters is exploited. Due to overlap between adjacentcotyledons, the cotyledons information is partly lost, which increase the difficulty ofidentification of the cotyledons morphology, thereby reduce the accuracy of the parameterextraction. To solve this problem, the geometric parameters extraction algorithm based onellipse fitting cotyledons contour is developed In this paper, which get all the cotyledon contours firstly, then combined and fitted contour segments according to specific criteria,followed by separating the cotyledons, and finally calculate all the parameters of thecotyledons. Due to the check feedback in the ellipse fitting process which could re-fit theequivocal contour segments, the recognition success rate of the cotyledons improved. Thealgorithm can not only stably and efficiently extract the cotyledons parameters of the differentocclusion forms, but also be able to identify the missing grafting seedlings coordinate. Theexperimental results show that the parameters extraction success rate of gourd cotyledons inthe occluded conditions is97.5%,which indicate that seedling cotyledon parameter extractionalgorithm with higher measurement accuracy meet the cotyledons geometrical parametersmeasurement requirements.The studied above show that some achievements about grafting seedlings parametersmeasurement by machine vision are acquired in this paper, which provides sufficient supportfor grafting robot when it automatically grasp and match the grafting seedling.