Recent approaches for contactless fire- and smoke detection rely widely on evaluating the content of video-streams, provided by for example video-cameras monitoring a scene. Different features for fire and smoke are calculated using information extracted from the video-stream. The final decision - if smoke or fire is represented in a video frame or not - is then made by merging the result of each single feature to an overall decision. In this thesis a novel approach to detect fire and a novel approach to detect smoke by using two different sensors is proposed. The first sensor is a RGB camera, providing a colored video-stream, while the second sensor is a 3D-sensor which calculates the distance between the sensor itself an objects located in a scene. Hardware, like the Microsoft Kinect, already combine both sensors. Results show that fire itself leads to significant changes in the depth-image provided by the 3D sensor, while the presence of smoke leads to changes in the color-stream only. Therefore, a method is presented, where the information of the RGB color-sensor is continuously analyzed if specific features like the color of moving pixels and their statistical analysis indicate a presence of fire in a scene. If a potential fire-region is recognized by the color-sensor, the implemented method for the depth-sensor verifies the potential fire-area (detected by the RGB sensor only) regarding its plausibility. This combined approach, by fusing RGB and depth information, leads to a decrease of the false-positive rate (fire is falsely detected in a frame) of 97.46%, compared to the detection rates by using the information of the color-sensor only. Due to the fact that smoke does not lead to a significant change in the stream provided by the depth-sensor, for smoke-detection a method using the RGB sensor only is implemented. The results show that fire-detection using e.g. a Kinect enhances the detection of real-fire events significantly, due to the fact that false-positives are effectively reduced.