| Animal husbandry plays an extremely important role in the national economy of Inner Mongolia,and the sheep industry also plays an important role in grassland animal husbandry.At present,the treatment methods for sheep delivery in domestic and foreign still remain at the stage of manual discovery and treatment.Operation by person does not guarantee timely and accurate detection of newborn lambs.A slight delay will result in irreversible losses.If we use a computer system with artificial intelligence as its core instead of manual detection,we can avoid many unnecessary losses.This method can reduce the mortality rate of newborn lambs,the manpower consumption will be the same.It’s very helpful to realize the wisdom of the animal husbandry.This paper studies the sheep delivery scene by means of the perspective of computer vision.Training the neural network model by the deep learning method,the newborn lambs in the sheep delivery scene can be detected by the neural network model.This research selected the Faster-RCNN neural network model to complete the detection of newborn lamb in the sheep delivery scene.Due to the lack of large public datasets for sheep delivery scenarios,data set for sheep delivery were gathered by collecting raw data from sheep delivery,screening available experimental data,expanding experimental data,and annotating experimental data.The self-made sheep delivery data set was used to train the Faster-RCNN model based on different feature extraction networks,different coverage domain thresholds,and different detection frame merging algorithms.Through the experimental results and the comparison of the detection visualization results,the Faster-RCNN model with better detection effect on the newborn lamb in the sheep delivery scene was selected.Finally,the Faster-RCNN neural network model with the best detection effect on sheep delivery scene is applied to the actual scene,so as to achieve the effect of detecting the newborn lamb in the real scene.This study proposes the detection of newborn lambs in a sheep delivery scenario based on deep learning.The neural network model was trained by using a self-made sheep delivery scene data set,and comparing the experimental results to select the Faster-RCNN neural network model with better detection effect.Therefore,there will be a fine effect and good value on this detection. |