Ion Berechet

Advanced algorithms for identifying targets from a three-dimensional reconstruction of sparse 3D ladar data

There is a considerable interest in the development of new optical imaging systems that are able to give threedimensional images. Potential applications range across medical imaging, surveillance and robotic vision. Identifying targets or objects concealed by foliage or camouflage is a critical requirement for operations in public safety, law enforcement and defense. The most promising techniques for these tasks are 3D laser imaging techniques. Their principles are to use movable light sources and detectors to collect information on laser scattering and to reconstruct the 3D objects of interest. 3D reconstruction algorithm is a major component in these optical systems for identification of camouflaged objects. But 3D reconstruction must take into account sparse collected data i.e. concealed objects and reconstruction algorithms must solve a complex multi-parameter inverse problem. Therefore the inverse problem of recovering the surface three-dimensional shape function from intensity data is more challenging. The objective of our paper is to present a new algorithmic approach for the generation of 3D surface data from 3D point clouds corresponding to reconstruction algorithm. This algorithmic approach is based on research of automatic minimization of an energy function associated with a sparse structure of 3D points. The role of this type of algorithmic data-driving process is to complete the incomplete 3D image at satisfactory levels for reliable identification of concealed objects.

Reflective Imaging Solved by the Radon Transform

This letter is concerned with imaging a 3-D scene from a set of 2-D laser images of backscattered intensity without prior knowledge. The interaction between an electromagnetic wave and a medium can be understood and modeled in different ways. In the configuration considered here, the interactions result in a 3-D projection of the scene. After inspection in the measured signal, the reflection data appear as a sum of point spread functions with or without angular limitations. The approach proposed here consists thus in considering reflection data as an incomplete data set of Radon kind used in conventional computerized tomography. Under this approximation, reconstruction techniques, such as the well-known filtered back projection or the Feldkamp algorithm, can be exploited. The advantage of this approach is to provide isosurfacic reconstructions of the scene from reflective data performed with real-time computation times and without prior information. Simulation results on a real laser tomographic data set attest of the strength and of the relevancy of the proposed method.

3D laser imaging for concealed object identification

This paper deals with new optical non-conventional 3D laser imaging. Optical non-conventional imaging explores the advantages of laser imaging to form a three-dimensional image of the scene. 3D laser imaging can be used for threedimensional medical imaging, topography, surveillance, robotic vision because of ability to detect and recognize objects. In this paper, we present a 3D laser imaging for concealed object identification. The objective of this new 3D laser imaging is to provide the user a complete 3D reconstruction of the concealed object from available 2D data limited in number and with low representativeness. The 2D laser data used in this paper come from simulations that are based on the calculation of the laser interactions with the different interfaces of the scene of interest and from experimental results. We show the global 3D reconstruction procedures capable to separate objects from foliage and reconstruct a threedimensional image of the considered object. In this paper, we present examples of reconstruction and completion of three-dimensional images and we analyse the different parameters of the identification process such as resolution, the scenario of camouflage, noise impact and lacunarity degree.

Optical 3D imaging and visualization of concealed objects

This paper gives new insights on optical 3D imagery. In this paper we explore the advantages of laser imagery to form a three-dimensional image of the scene. 3D laser imaging can be used for three-dimensional medical imaging and surveillance because of ability to identify tumors or concealed objects. We consider the problem of 3D reconstruction based upon 2D angle-dependent laser images. The objective of this new 3D laser imaging is to provide users a complete 3D reconstruction of objects from available 2D data limited in number. The 2D laser data used in this paper come from simulations that are based on the calculation of the laser interactions with the different meshed objects of the scene of interest or from experimental 2D laser images. We show that combining the Radom transform on 2D laser images with the Maximum Intensity Projection can generate 3D views of the considered scene from which we can extract the 3D concealed object in real time. With different original numerical or experimental examples, we investigate the effects of the input contrasts. We show the robustness and the stability of the method. We have developed a new patented method of 3D laser imaging based on three-dimensional reflective tomographic reconstruction algorithms and an associated visualization method. In this paper we present the global 3D reconstruction and visualization procedures.

Scattering computation for 3D laser imagery and reconstruction algorithms

This paper addresses non-conventional three-dimensional imaging with laser systems which explores the advantages of laser imagery to form a three-dimensional image of the scene. In this paper, we present the 3D laser scattering simulation of objects hidden behind porous occluders, such as foliage or camouflage. The physics based model presented in this paper is designed to provide accurate results but also to include all the electromagnetic interaction mechanisms with the different elements of the scene. A 3D laser cross-section computer model is used to develop reconstruction algorithms to obtain a high-resolved three-dimensional image. Synthetic images of three-dimensional objects are based on extraction of laser backscattered signals. But 3D reconstruction must take into account sparse collected data and reconstruction algorithms must solve a complex multi-parameter inverse problem. The objective of our paper is also to present new algorithmic approaches for the generation of 3D surface data from 3D sparse point clouds corresponding to our reconstruction algorithm. The role of this type of algorithmic process is to complete the 3D image at satisfactory levels for reliable identification of concealed objects. Identifying targets or objects concealed by foliage or camouflage is a critical requirement for operations in public safety, law enforcement and defense.