A Recognition Method of Oil Depot Objects in Remote Sensing Images

Abstract

The invention discloses a method for identifying an oil depot target in a remote sensing image. Firstly, the saliency of the phase spectrum of the entire scene is calculated, and all regions of interest in the scene that may contain the target are extracted according to the saliency of the phase spectrum; in the feature extraction, local regression is adopted The kernel model calculates the local structural features of the region of interest point by point, and generates a feature descriptor that can describe the target structure; in the target detection stage, the cosine similarity is used as similarity measurement to calculate the similarity between the region of interest and the sample image of the oil depot , and use the feature descriptor's ability to distinguish positive and negative samples and the characteristics of the similarity surface to construct a decision network with adaptive capabilities, obtain the preliminary results of target detection through the decision network, and remove redundant preliminary results through the non-maximum value suppression algorithm , to obtain the final target detection result; the general detection method of the oil depot target in the remote sensing image proposed by the present invention has a good effect on multi-scale and multi-view target recognition.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image target recognition, and more particularly relates to a method for recognizing an oil depot target in a remote sensing image. Background technique [0002] High-resolution remote sensing images provide a wealth of detailed information, making it possible to identify various specific targets; however, noise interference, seasonal weather, shadows, light intensity, occlusion and other factors will cause fluctuations in the structure and texture information of internal details of the target , which brings difficulties to the recognition of high-resolution images. [0003] The methods for target detection of oil depots in remote sensing images in the prior art include: target detection methods based on deep learning, target recognition and detection methods based on prior knowledge, and target detection methods based on models; It has a high dependence on the number of samples, and in mos...

AI Technical Summary

Problems solved by technology

[0002] High-resolution remote sensing images provide a wealth of detailed information, making it possible to identify various specific targets; however, noise interference, seasonal weather, shadows, light intensity, occlusion and other factors will cause fluctuations in the structure and texture information of internal details of the target , which brings difficulties to the recognition of high-resolution images

[0003] The methods for target detection of oil depots in remote sensing images in the prior art include: target detection methods based on deep learning, target recognition and detection methods based on prior knowledge, and target detection methods based on models; and the number of samples have a very high dependence, and in most cases, the recognition of remote sensing image oil depot targets can only provide simple data image sources; the target recognition method based on prior knowledge is to use the prior knowledge of the target such as aircraft The priori features such as the mean, variance, and invariant moments of the target are used as the basis for judging the position of the target. It is necessary to accurately express the characteristics of the target, and a decision-making method with adaptive ability is required. When the prior knowledge expression is not accurate enough or the decision-making method is not enough In the case of perfection, the accuracy of target detection is low; the model-based method is to extract target features through a large number of experiments, mark the model parameters of the target to generate hypotheses and predict the target characteristics, and measure the background or target model in actual use. Then match with the predicted characteristics and reach a certain degree of similarity, that is, it is considered to be the target; this model-based target detection method has high requirements for the accuracy and fault tolerance of the modeling, which is very important for remote sensing images in complex scenes. The identification of oil depot targets is very difficult

Images