Research On Infrared Multi Weak Targets Detecting And Tracking Algorithm Based On MFPDE And GM-CPHD Filter

Due to the wide application of infrared imaging technology, the target detection and tracking issue based on infrared imaging has always been a hot research topic. Infrared imaging makes use of the heat radiation information of the target itself, when the target is far away, and the target imaging size is small, the energy that the infrared detector receives is very weak, then it's hardly to detect the target. When the environment surrounding this kind of the weak small target is complex, the target may completely submerged in the background. Furthermore, if there are multiple targets which need to detect and track, it becomes an extremely difficult task.Firstly, an modified fourth-order partial differential equations algorithm is proposed in this paper for the problem of background suppression of the infrared image. The anisotropic diffusion method based on the fourth-order partial differential equations can preserve the image edge information commendably, and simultaneously smooth the background. By improving the conditions of the diffusion coefficient, the area in infrared image where's grayscale distribution is uniform or gradient equal can be suppressed or removed, then, the complex background can be removed, and the contrast ratio of the target signal can be equivalently improved. Because the method does not need to iterative, the time complexity is low.The simulation results verified the effectiveness of the proposed algorithm.Then, for the problems of the number and the position of multiple weak targets in infrared image sequence are uncertain, and the target is vulnerable to the clutter interference, an multi-target tracking algorithm based on the Gaussian mixture cardinalized probability hypothesis density filter is proposed. The cardinalized probability hypothesis density filter can jointly propagates the posterior intensity and the posterior cardinality distribution of the multi-target, it's Gaussian mixture implementation method assumes that each target follows a linear Gaussian dynamical model, and then the mean value of the state, the weight and the covariance of each target can be treated as a Gaussian term. The multi-target state can be obtained by predicting, updating, pruning and merging the Gaussian terms, and the multi target number can be obtained by predicting and updating the cardinality distribution. Thus the state and the number of multiple weak targets can be estimated.At last, the proposed modified fourth-order partial differential equations algorithm and the Gaussian mixture cardinalized probability hypothesis density filter are joined together. After processing of the modified fourth-order partial differential equations, the adaptive threshold segmentation algorithm is used to get the position and number of candidate target information. By feeding the candidate targets information into the Gaussian mixture cardinalized probability hypothesis density filter to estimate the real-time states and the number of the multiple weak targets.Simulation results show that the proposed method has a good performance to detect and track time-varying multi weak targets in infrared image with complex backgrounds, and it's not necessary to consider the data association problem, so, it has low computational complexity.