Research On Classification Method Of Infrared Images Under Sea Surface Background

In recent years,marine distress incidents have occurred from time to time,and it is an urgent and challenging task to be able to carry out rapid and accurate search and rescue of distress targets.In the process of searching and rescuing distress targets on the sea,infrared imaging technology is the main application.Due to the complex and changeable sea environment,it may affect the search and detection of targets,and then affect the efficiency of the search and rescue.And the current sea target detection algorithm is usually applicable in a specific environment.Therefore,in order to better adapt to the image target detection in a variety of complex sea environment,it is necessary to get the sea image according to different sea backgrounds.Furthermore,it can detect and process all kinds of sea surface images in a targeted way,which is helpful to improve the success rate of search and rescue.Obviously,it is urgent and practical to classify the sea image according to different sea conditions.Therefore,this paper mainly studies the scene classification of sea infrared image under different sea conditions.Based on the analysis of the sea surface infrared image and the actual needs of the sea surface search and rescue environment and the design of the target detection algorithm,according to the actual needs of the environment and the target detection algorithm design in the sea surface search and rescue,the different sea surface environment conditions in the sea surface infrared images are divided into five kinds of different scenes,that is,back-lighting environment(BE),large wind wave environment(LWWE),sea fog environment(SFE),sea-sky-line environment(SSLE),and calm sea environment(CSE).According to the extracted features of sea surface infrared images and the combined classification methods,two classification methods are proposed,one is an image classification method based on extracting feature values,and one is an image classification method based on feature vectors.Based on the feature value,this paper proposes a method to calculate the image gray concentration,gray gradient and gray co-occurrence matrix to describe the image features.Through the spatial feature extraction,the characteristic parameters of each type are calculated,and the similarity of the statistical characteristics of the same type of image and the difference of the statistical characteristics of the different types of image are obtained through comparison.Based on the analysis of statistical law and the threshold processing method,the infrared images of sea surface are classified.The results show that the accuracy of this method is low,and it has not reached the ideal classification effect after the threshold is constantly modified.This paper proposes a new feature description representation method based on feature vectors,this paper proposes a new feature descriptor representation method for the sea infrared image,which combines the improved histogram of directional gradient(HOG)texture feature based on image layering and the local contrast(LC)gray feature.Then the sea surface images are classified according to the above five kinds of scenes.Based on the analysis of the gray distribution and small edge information characteristics of different scenes of the sea infrared image,this paper mainly extracts the features of the sea infrared image from two aspects of texture and gray level.In the aspect of texture,in order to reflect more detail information,the original image is divided into base layer image and detail layer image by Gauss filtering method,and then the features of different layers are extracted by the improved hog method to obtain the feature vectors of base layer and detail layer;in the aspect of gray level,the gray level features of the image are obtained by calculating LC of the original image,and finally the gray level features are extracted.Finally,the extracted features are fused to generate a new feature descriptor as the feature expression of the sea surface infrared image.Then,support vector machine(SVM)classifier is used to classify the sea infrared image background.The obtained sea infrared images are divided into training set and test set.The image in the training set is input into SVM classifier for training,and the final discrimination model is obtained,and then the images in the test set are classified and verified.In order to verify the effectiveness of the sea infrared image classification method,different feature extraction methods are used for comparative analysis of multiple groups of experiments,and SVM different kernel functions are used for comparative analysis.Compared with the commonly used feature extraction methods such as Local Binary Pattern(LBP)and Bag of Features(BOF),the experimental results show that the method proposed in this paper can better classify the sea surface images as needed,which reflects the feasibility and effectiveness.