ing at the problems that the stacked sparse denoising auto-encoder(SSDA) is difficult to train on image denoising, such as slow convergence rate and poor universality, an adaptive image denoising model based on stacked rectified denoising auto-encoder is proposed. The rectified linear units is used as a network activation function to alleviate the phenomenon of gradient dispersion. Joint training with the residual learning and batch normalization to accelerate convergence speed. In order to solve the problem of noise poor universality of the new model, it is necessary to carry out the multi-channel parallel training, and make full use of the potential data feature extracted by the network to find the optimal channel weights, and learn to predict optimal column weights via training weight prediction model for realizing adaptive image denoising. The experimental results show that the proposed algorithm is not only better than the SSDA in the convergence effect, but also adaptively processing the non-participating training noise, and has better universality, compared with the current methods of BM3D and SSDA.
.tracking based on correlation filter has become a research hotspot currently. The traditional tracking model trained from circular correlation is sensitive to the pixel arrangement of the target and is difficult to adapt object deformation, but it has good robustness of variety in color of illumination and similar color interference. However, the model based on spatial reliability can adapt to the deformation by establishing the spatial confidence map as the random field constraint of the correlation filter, but it has less robustness for the color change. In order to exert the superiorities of the two tracking methods, the concept of directional reliability is innovative presented and a set of the optimization strategies is proposed to achieve optimal translation estimation of the two tracking models in both the
ing at the problem of decreasing the recognition efficiency of multi-class Support Vector Machines (SVM) in the detection and classification of lane arrow markings, an improved method for a simple SVM classifier which is applied to realize the multi classification of arrow markings by custom binary encoding for results is proposed. The Harris corner coarseness is detected for the arrow markings region of interest (ROI), and the pseudo corners are screened by improved FAST-9 (features from accelerated segment test-9) algorithm. According to the location of the largest two corners of the ordinate in the final corner set, the recognition area is obtained. The SVM classifier is trained by invariant moments. And the multi classification with one SVM classifier is realized via the binary encoding for results. The algorithm is tested on 500 real images obtained from the real shot, and the recognition rate is superior to 96.8%. The results show that the proposed method does not need inverse perspective transformation. A simple SVM classifier can realize the multi classification of arrow markings, and the accuracy and operation efficiency of arrow marking recognition can be improved effectively.
.