Another related approach relies on the use of a conditional particle filter to detect persons in motion .These previous methods do not use 3D information as input for their calculations, a restriction that limits their use for security applications, as they only detect moving objects at a predetermined height. This shortcoming was released by Tanner and Hartmann  by using a single time of flight (ToF) indoor camera. In the same line, Swadzba et al.  were able to track dynamic objects to reconstruct a static scene by using a ToF camera and a 6D data representation consisting of 3D sensor data and computed 3D velocities. Other options combine a 2D LIDAR scanner with a vertical servo to obtain 2.5D data of the environment (range images or point clouds). Using this combination, Ohno et al.
 were able to eliminate the moving objects from the scans of static scenes by comparing collision distances in the same area. More recently, Moosmann and Fraichard  have proposed a method consisting of deriving a dense motion field based exclusively on range images for performing object-class independent trajectory estimations.However, none of the previous approaches use full environment range images to effectively detect and track multiple dynamic objects from multiple robots. Herein, we develop two methods to detect moving objects from a robotic platform using range images. The first one is intended for static platforms and the second for dynamic ones. Furthermore, detection based on this type of data is followed by an effective tracking process using the generated dynamic objects lists.
1.2. Tracking of Dynamic ObjectsThere are several possibilities for tracking dynamic objects using a single robot. One of the most successful approaches  uses parameters, such as size and position, in a blob segmentation algorithm to characterize each detected object. These blobs are managed by creating a movement hypothesis with specific position and velocity data for each object. Each hypothesis is stored and updated with the estimated position and velocity of the objects, as well as with a weighting probability of the actual tracking of the moving object.In multi-robot systems, the information generated by each robot must be combined to enable better tracking. Stroupe et al.
 proposed two-dimensional Gaussian distributions to represent each observation of the object and a statistical procedure based on the Bayes rule and Kalman filters to combine two measurements. Another cooperative target tracking approach as proposed by Wang et al.  Carfilzomib consists of a distributed Kalman filter to estimate the target position. Mazo et al.  proposed a hierarchical algorithm to locate and track a single dynamic object from data provided by a two-robot system.