The invention relates to a robust target tracking method based on deep learning and multi-scale correlation filtering. The tracking process is divided into a target location part and a scale selection part. In the target location part, the position of a target is located through a convolutional neural network and correlation filtering. In the scale selection part, a scale pyramid is used, and different scales are selected in a matching manner for targets through scale filtering. The multilayer characteristic of the convolutional neural network is taken as a representation model of targets, so the structural and semantic information of targets can be described robustly. Through use of the characteristics of correlation filtering, there is no need to train a classifier online, and the running speed of the algorithm is increased greatly. The idea of scale pyramid is adopted in scale, and correlation filtering matching is performed on targets of different scales to select the optimal scale. The method is of strong robustness to deformation, shading and scale change of targets.

Robust target tracking method based on deep learning and multi-scale correlation filtering

The invention discloses a significant object detection method based on sparse subspace clustering and low-order expression The method comprises the steps of: 1, carrying out super pixel segmentation and clustering on an input image; 2, extracting the color, texture and edge characteristics of each super pixel in clusters, and constructing cluster characteristic matrixes; 3, ranking all super pixel characteristics according to the magnitude of color contrast, and constructing a dictionary; according to the dictionary, constructing a combined low-order expression model, solving the model and decomposing the characteristic matrixes of the clusters so as to obtain low-order expression coefficients, and calculating significant factors of the clusters; and 5, mapping the significant value of each cluster into the input image according the spatial position, and obtaining a significant map of the input image According to the invention, the significant objects relatively large in size in the image can be completely and consistently detected, the noise in a background is inhibited, and the robustness of significant object detection of the image with the complex background is improved The significant object detection method is applicable to image segmentation, object identification, image restoration and self-adaptive image compression

Significant object detection method based on sparse subspace clustering and low-order expression

The invention discloses a SAR (Synthetic Aperture Radar) image segmentation technique based on dictionary learning and sparse representation, and mainly solves the problems that the existing feature extraction needs a lot of time and some defects exist in the distance measurement The method comprises the following steps: (1) inputting an image to be segmented, and determining a segmentation class number k; (2) extracting a p*p window for each pixel point of the image to be segmented so as to obtain a test sample set, and randomly selecting a small amount of samples from the test sample set to obtain a training sample set; (3) extracting wavelet features of the training sample set; (4) dividing the training sample set by using a spectral clustering algorithm; (5) training a dictionary by using a K-SVD (Kernel Singular Value Decomposition) algorithm for each class of training samples; (6) solving sparse representation vectors of the test sample on the dictionary; (7) calculating a reconstructed error function of the test sample; and (8) calculating a test sample label according to the reconstructed error function to obtain the image segmentation result The invention has the advantages of high segmentation speed and favorable effect; and the technique can be further used for automatic target identification of SAR images

SAR (Synthetic Aperture Radar) image segmentation method based on dictionary learning and sparse representation

The invention discloses a single image super-resolution rebuilding method. Based on non-local similarity and a classification half-coupling dictionary learning algorithm, the method comprises a training stage and a rebuilding stage. According to the method, the half-coupling dictionary learning algorithm is used as a framework, training image block sparse domain classification based on mapping errors is introduced, and a heuristic method strategy conducted alternatively through the sparse domain classification and the half-coupling dictionary learning is adopted; sparse domain non-local similarity restriction items are introduced, structural information of training image block space is excavated in the sparse domain so as to rebuild more high-frequency details; the sparse representation algorithm based on non-local restriction is improved to meet the requirements of the half-coupling dictionary learning algorithm overall framework; further, an error compensation mechanism is introduced into the rebuilding stage to further improve the super-resolution rebuilding quality. Compared with the prior art, the method improves rebuilt texture details and forge edge and saw tooth removal, achieves good effects, and achieves best subjective visual effect in the prior art.

Single image super-resolution rebuilding method

The invention provides a target tracking algorithm based on the self-adaptive particle filter and sparse representation. According to the target tracking algorithm based on the self-adaptive particle filter and the sparse representation, the improved self-adaptive particle filter technique is adopted to serve as a tracking algorithm framework, a block sparse representation model is used for establishing an observation similarity model of a target, partitioning of the target is achieved by means of the self-adaptive partitioning technique, a structural sparse column diagram of a current target state is constructed to calculate the observation similarity of the current target state, blocking is detected by means of a blocking detection mechanism, a target / background dictionary template and a target template column diagram are updated to capture the change of the appearance of the target and the change of the environment during tracking, L1 optimization in the sparse representation is achieved by means of the variable-direction multiplicator method, and then the execution speed of the target tracking algorithm is increased. The target tracking algorithm based on the self-adaptive particle filter and the sparse representation has the advantage that the robustness to the conditions of the posture change of the tracking target, the change of the environment and lighting and blocking is strong.

Target tracking algorithm based on self-adaptive particle filter and sparse representation

The invention discloses an image super-resolution reconstruction method based on dictionary learning and structure similarity, mainly solving the problem that a reconstructed image based on the prior art has a fuzzy surface and a serious marginal sawtooth phenomenon The image super-resolution reconstruction method comprises the following implementation steps of: (1) acquiring a training sample pair; (2) learning a pair of high/low-resolution dictionaries by using structural similarity (SSIM) and K-SVD (K-Singular Value Decomposition) methods; (3) working out a sparse expression coefficient of an input low-resolution image block; (4) reestablishing a high-resolution image block Xi by using the high-resolution dictionaries and the sparse coefficient; (5) fusing the high-resolution image block Xi to obtain a high-resolution image X'I subjected to information fusion; (6) obtaining a high-resolution image X according to the high-resolution image X'I; and (7) carrying out high-frequency information enhancement on the high-resolution image X through error compensation to obtain a high-resolution image subjected to high-frequency information enhancement A simulation experiment shows that the image super-resolution reconstruction method has the advantages of clear image surface and sharpened margin and can be used for image identification and target classification

Image super-resolution reconstruction method based on dictionary learning and structure clustering

The invention discloses a concealed-target tracking method based on on-line learning, and mainly solves the problems in the prior art that when partially concealed, a target is subject to template drift, and when completely concealed, the tracking fails. The method comprises the following steps: (1) inputting a first frame of a video and manually marking a target to be tracked; (2) establishing and initializing a random forest detector and a Kalman filter; (3) detecting the target by the random forest detector and tracking the target through a Lucas-Kanade tracker; (4) determining the position of the target according to the detection result of the random forest detector and the tracking result of the Lucas-Kanade tracker; (5) judging whether the target is partially concealed or not; (6) judging whether the target is completely concealed; (7) on-line updating; (8) forecasting the position of the target by adopting the updated Kalman filter; (9) displaying the tracking result; and (10) repeating the step (3)-(9) till the video image finishes. The concealed-target tracking method has the advantage that the tracking is accurate when the target is partially or completely concealed, and can be used in the human-computer interaction field.

Concealed-target tracking method based on on-line learning

Hough forest-based video target tracking method

The invention discloses a method for semi-automatically detecting a road target in high-resolution remote sensing images, belonging to the field of the remote sensing image processing technology and aiming at mainly solving the problems that the prior method fails to accurately detect the road when the images with complex background, the road widths, the road surface markers and the gray values are different greatly The method comprises the following steps: selecting a 2N*2N window according to a starting seed point and a initial direction which are inputted manually; carrying out the histogram equalization and the preliminary gaussian filtering for the images in the window; extracting the straight road boundary line by using the wavelet transform and the Hough transform, and calculatingthe coordinates in the central points of the road; correcting the coordinates in the central points of the road by using a template matching method based on the gray comparison, and correcting the road direction by using the obtained information on the coordinates in the central points of the road; and connecting iterated coordinates in the central point of the road in sequence to obtain the central line of the road, and outputting the central line of the road The method can be used for processing the remote sensing images with complex background, effectively detecting the road when the roadwidths, the road surface markers and the gray values are different greatly, accurately positioning the central line of the road and semi-automatically detecting the road target

Method for semi-automatically detecting road target in high-resolution remote sensing images

The invention discloses an SAR (Synthetic Aperture Radar) image segmentation method based on parallel sparse spectral clustering, relating to the technical field of image processing and mainly solving the problem of limitation of segmentation application of large-scale SAR images in the traditional spectral clustering technology. The SAR image segmentation method comprises the steps of: 1, extracting features of an SAR image to be segmented; 2, configuring an MATLAB (matrix laboratory) parallel computing environment; 3, allocating tasks all to processor nodes and computing partitioned sparse similar matrixes; 4, collecting computing results by a parallel task dispatcher and merging into an integral sparse similar matrix; 5, resolving a Laplacian matrix and carrying out feature decomposition; 6, carrying out K-means clustering on a feature vector matrix subjected to normalization; and 7, outputting a segmentation result of the SAR image. The invention can effectively overcome the bottleneck problem in computation and storage space of the traditional spectral clustering technology, has remarkable segmentation effect on large-scale SAR images, and is suitable for SAR image target detection and target identification.

Zhu Huming

Papers

Image super-resolution reconstruction method based on dictionary learning and structure clustering

Concealed-target tracking method based on on-line learning

Hough forest-based video target tracking method

Method for semi-automatically detecting road target in high-resolution remote sensing images

SAR (Synthetic Aperture Radar) image segmentation method based on parallel sparse spectral clustering