Measurement Points Optimization And Body Size Measurement For Sheep Based On Improved C-V Model

In animal husbandry farming,sheep body size parameters and weight are the main parameters to judge sheep growth level and productivity analysis,also their measurement work plays a key role in the process of experiment.Currently,the most favoured way of measuring sheep is manual measurement.Owing to direct contact between survey crew and sheep,it appears the stress response of sheep easily.The current situation leads to increase measurement error in the actual measurement process.To solve the above issues,it is necessary to use the image processing technology to acquire sheep body size parameters and weight in the mode of non-contact and non-stress.The research can be categorized to three aspects:sheep image segmentation under complex background,optimal ranging of measurement points and sheep weight estimation,which is beneficial to realize sheep measurement with non-contact method.The main contents are as follows:(1)To solve the problem that the object was difficult to segment from complex background in the sheep images and to take into account the illumination interference,the improved C-V active contour model based on YCbCr space was proposed.Three color components in YCbCr space were used for sheep image analysis to get Cb component to improve C-V model.It is proved that the improved C-V model can achieve interactive and semiautomatic image segmentation under complex background.(2)A method of sheep measurement points extraction based on the inflection point was proposed,which is to consider the different postures of sheep in the image acquisition process and then to generate accurate measurement points.At first,according to sheep image segmentation,the sheep contour was extracted.Then,the sheep measurement points were attained extracting the inflection point and calculating the contour sequence,to achieve the identification of seven measurement points.In the end,the measured parameters based on measurement points compared with the actual parameters.Although it has a certain error,the mean relative error is less than 3%.(3)Body size parameters were attempted to build the linear and nonlinear models of sheep weight for predicting.The experimental results show that using RBF network can demonstrate the predictive validity,and the mean relative error is 7.61%.