Segmentation of Egg Breeder under Complex Background
Image segmentation is the premise of image recognition, good segmentation effect is conducive to improve the accuracy of image recognition. In order to automatically identify the behavior of egg-breeders, the authors collected images of birds in self-breeding cages. Four kinds of methods were adopted respectively to segment the objects, such as threshold segmentation (Otsu, iteration method, minimum error method), edge detection (Roberts operator, Sobel operator, Prewitt operator, LOG operator), region segmentation (region generation). Then the running time and segmentation effects of each algorithm were compared. Based on that, this paper proposed a model based on histogram of oriented gradient (HOG) feature for segmentation under complex background, where radial basis function (RBF) was used to map the training samples to high-dimensional eigenspace, and C_SVM was used to train the classifier. According to the test, precision ratio was 94.8%, recall ratio was 95.7% and the accuracy rate was 95.3%. The test results showed that the algorithm achieved better segmentation effect. This method will provide useful reference for image segmentation in complex environment.