Pursuit-evasion game (PEG) problems are a type of dynamic differential games that received a lot of attention thanks to the ability of this framework to articulate many real-life applications such as in military, aerospace and mobile robotics. Several techniques are used to solve such games, but recently techniques relying on deep reinforcement learning (DRL) gained traction, in particular DRL techniques adapted for problems with continuous action spaces such as Deep Deterministic policy gradients (DDPG). This paper explores the case of a one versus one pursuit-evasion game in a constrained game area, using two twin delayed DDPG (TD3) agents that are trained simultaneously from scratch via self-play only. The simulation results show that agents were performing better than other conventional methods such as Non-linear Model Predictive Control (NMPC).

Convolutional models may effectively identify persons quickly through automatic face analysis. A convolutional neural network architecture known as the CNN cascade structure uses numerous deep convolution layers to extract hierarchical characteristics from the input image. Cascade modeling has several drawbacks, including high computational costs and complexity, difficult training, little feedback, accumulation of errors, sensitivity to model order, and challenging interpretation. In order to improve model performance due to these many issues, we propose adding a dynamic learning rate (DLR). It entails gradually modifying the learning rate in response to the model’s performance during training. The loss of training or validation error is a concern with cascade models. The CNN cascade structure and the DLR technique are combined in this paper to present a novel deep-learning methodology for reliable face recognition as a biometric security solution. The proposed cascading CNN model based on DLR was assessed on 3D face images from the MIT CBCL database. It allows achieving a higher accuracy of up to 99.65% with 5 epochs.

The image processing arose from the idea of the necessity to replace the human observer by a machine. The interest of this paper is to replace the medical image by information interpretable. Usually, experts have manually performed to count the cell nuclei biopsy samples, one by one. This method ensures that accuracy is achieved in the final diagnosis delivered by pathologists, but the time until the patient is notified can vary from weeks to months depending on the laboratory resources. Cancer developing speed is also a limiting factor, so the sooner the disease is discovered the better and quicker the patient can start with the treatment or preparations for surgery can be arranged. Promptness in cancer recognition increases the chances to overcome this illness that affects every year more and more men as the world population's life expectancy increases. So, for this reason, it has proposed an automatic method. To return the more reliable and fast diagnosis, we applied a method based on tools and algorithms. The chain of this processing is begun with the segmentation to separate the various constituent zones the image. Secondly, we have the step of detecting the edges of the prostatic cells as well their center. Finally, we have the step of counting where we are going to find a score for the diagnosis.

Abstract Dramatic technology progress in data manipulation induced several attempts of baleful and illegal processing. In this regard, several protection techniques, including cryptography, steganography, and watermarking have been used to avoid the illegal production and distribution of data online. On the other side, the attacks are still a major problem that limits the effectiveness of watermarking techniques. The geometric attacks involve displacement of pixels. Therefore, they induce synchronization errors between the original and attacked image that complicates the watermark recovery process. So in order to solve the de‐synchronization problem, we propose to use a robust meshing technique between different geometrical data of the image to correct the geometric distortion. In this work, we propose a novel geometrically invariant digital image watermarking approach based on the geometrical feature of the cover image and the Defragmented Delaunay Triangulation on respecting the three constraints of watermarking approaches (imperceptibility of embedded watermark, embedding capacity, and robustness). The aim idea of this work is to propose a blind, imperceptible, and robust digital image watermarking scheme based on Defragmented Delaunay Tessellation and Weber's law. The defragmented triangulation provides a best synchronization of the embedded data after attacks application. The Weber's law is used to propose an auto‐thresholding algorithm to compute the optimal embedding gain factor for each watermark's bit in order to ensure the imperceptibility of the hidden data. Firstly, the invariant features in the host gray scale image are extracted by using the canny edge detector. Then, the Delaunay tessellation of the saved keys points set is generated and defragmented to select the optimal robust triangles for watermark embedding. Simulation results illustrate the imperceptibility of the embedded data and the robustness of the proposed approach against intentional and unintentional geometrical attacks are presented in the finally section. Copyright © 2017 John Wiley & Sons, Ltd.