An Entropy Navigation Method Based On HOG Descriptor

The aerospace is one of the important indicators of national science and technologylevel is developed and national defense is powerful. There are many common navigationmethods after continuous development in navigation field. Among them, the visualnavigation based on insect is a hot key object of study, but also the future developmenttrend of navigation. The hardware requirements of visual navigation is relatively low, itneed only a simple visual sensors. Besides, it requires no external signals for navigationand can get the distance from the camera by calculating the real objects in the environmentto avoid obstacles in order to reach the autonomous navigation purpose.Histogram of Oriented Gradients is a local area descriptor through the local area ofcomputational and statistical image gradient histogram. It is based on the image of thelocal cell and the local cell are got normalization operation which is very useful tocharacterize the image’s local information for the reason that it remain the optical andgeometric invariance.In this paper, on the basis of existing laboratory project’s scientific research I madesome improvement and propose a navigation method base on the HOG entropy flow.Firstly, according to the concept of entropy flow combined with the power of HOG Iredefine the calculation of the entropy map and show the detailed steps of constructedentropy diagram. Secondly, we got the flow of information in the navigation process basedon the entropy flow which is calculated by entropy map, and set entropy flow as an inputparameter into the six-parameter motion estimation model for global motion estimation inorder to get the navigation motion estimation parameters. However the use of least squaresmethod will introduce the noise data, so the finally step is to use the Kalman filter toobtain the optimized motion estimation parameters.We simulated the navigation methods in the MATLAB platform after the previoustheoretical research. At first, we show the comparison of the original image and the imageprocess by HOG descriptor. Then we gave the effect of entropy flow navigation by usingHOG. Through the analysis of the result, we found that using HOG descriptor can get more navigation information compared with general information entropy to describe theimages’change information so as to reach the purpose of accurate navigation.