Search

This paper proposes an integrated vision based obstacle detection algorithm based on Background subtraction technique, optical flow, Grey Level Co-occurrence Matrix features, Color Histogram Features and Hu Moments for Micro Aerial vehicles (MAVs). Raspberry Pi 3 camera module is mounted on the MAV to acquire the image frames with obstacles. Obstacle detection algorithm is developed using Python programming language. In the feature extraction phase, Haralick Features, Color Histogram Features and Hu Moments Features were extracted from the training image frames. The extracted features are used to train the SVM classifier. In the testing phase, trained object recognition model is used to detect the object in the test image frame. Support vector machines with various kernels such as the linear kernel (LSVM), Gaussian kernel SVM (GSVM), Sigmoid kernel SVM (SSVM) and Polynomial kernel SVM (PSVM) have been employed. Experimental results demonstrate that the SVM with gaussian kernel-based yields the highest accuracy for the classification of object class and non-object class from the testing images. Finally, dynamic obstacles are detected by using the Optical Flow Lukas Kanade tracking method. [ABSTRACT FROM AUTHOR]

The Experiment illustrates the working of Lidar based obstacle detection and avoidance system for Micro-Aerial Vehicle by interfacing the Lidar sensor to Micro-Aerial vehicle to estimate the distance and the relative velocity of the obstacle detected. LiDAR measures distance accurately and has high sensitivity towards change in the distance. The obtained input data from the lidar sensor assists the system in making judgments to proceed in a free path while avoiding obstacles. The Lidar sensor works on Time of flight principle which provides multiple inputs required for estimation. The motors mounted on Micro-Aerial vehicle are been controlled using L293D motor driver after receiving the input from the Lidar sensor. [ABSTRACT FROM AUTHOR]