Research on Target Intelligent Detection of Port Container Keyhole
In order to solve the problems of low contrast of information,obscure details and difficulty in identifying image information caused by external light environment in container images,proposes a new image enhancement algorithm,which focuses on improving the dark part details of the image by using nonlinear transformation and converting the original RGB model to the HSV model for equalization.Using the test data set provided by the port in the actual project,a wide range of experiments are carried out with the trained model.The experiment explores the effect of image enhancement algorithm on the performance of convolutional neural network.The contrast limited adaptive histogram equalization,gamma correction,Laplace transform and original image are compared with the proposed algorithm,and the 10-fold cross is used to verify the accuracy rate and recall rate of the proposed algorithm,which are greater than other algorithms.The paired T test is used to compare and analyze the differences of each index among the algorithms.And the results show that the proposed algorithm has better effect than other algorithms.
万方数据
维普
