Simulation Research On A Class Of New Light Field Reconstruction Method And Target Recognition

With the rapid development of modern information technology,the role of target recognition in various fields is increasingly prominent,and the target recognition technology has been considerable development with the continuous efforts of several generations people.As a special form image,the light field has been favored by many scholars.In recent years,a large number of research results about light field reconstruction spewing out.So that not only provides the direction for the acquisition of the light field,and compensate for the lack of recognition in a certain extent.So this article revolves around object recognition and light field reconstruction studies carried out as follows:(1)A new method of light field reconstructing is proposed,which combines with wavelet transform and sparse Fourier transform.In the method,the original image is decomposed into four sub-images by using wavelet transform,and which are reconstructed respectively.Not only reduces the computation complexity,but also the method effectively inhibited the window effect.In addition,the method can effectively improve problem of small frequency leakage in off-grid recovery by separating high frequency and low frequency information.In the end of the thesis,the effectiveness of the algorithm is verified by simulation.(2)The thesis proposes an image recognition algorithm based on light field.Based the target recognition algorithm existing questions,and the light field reconstruction is applied to the image recognition as the feature extraction library.This recognition algorithm can solve the problem which the recognize image is obtained in the different camera angles.Finally,the simulation verifies the effectiveness of the algorithm.(3)A new object recognition context method is proposed in this thesis by introducing the spatial location relationship of objects into the context models.This thesis also analyzes and improves the representation of object spatial relationship through the degree of membership function.The new method can greatly improve the object recognition rate and better keep the consistency of scenes.The effectiveness of the proposed algorithm is verified by testing and comparing with other existing algorithms in actual dataset.