a new generation of computer vision, namely event-based or neuromorphic vision, provides a new paradigm for capturing visual data and the way such data is processed. Due to a highly novel type of visual sensors used in event-based vision, only a few datasets aimed at visual navigation tasks are publicly available.In this paper, we present and describe the first event-based vision dataset intended to cover visual navigation tasks for mobile robots navigating in different types of agricultural environment. The dataset might open new opportunities for the evaluation of existing and creation of new event-based visual navigation methods for use in agricultural scenes that contain a lot of vegetation, animals, and patterned objects.The new dataset was created using our own custom-designed Sensor Bundle, which was installed on a mobile robot platform. During data acquisition sessions, the platform was manually controlled. The Sensor Bundle consists of a dynamic vision sensor, a LIDAR, an RGB-D camera, and environmental sensors.In total, 21 data sequences in 12 different scenarios for the autumn season are publicly available. Each data sequence is accompanied by a video demonstrating its content and a detailed description, including known issues. The new dataset is mostly designed for Visual Odometry tasks; however, it also includes loop-closures for applying event-based visual SLAM methods.

The Dynamic Visual Sensor is considered to be a next-generation vision sensor. Since event-based vision is in its early stage of development, a small number of datasets has been created during the last decade. Dataset creation is motivated by the need for real data from one or many sensors. Temporal accuracy of data in such datasets is crucially important since the events have high temporal resolution measured in microseconds and, during an algorithm evaluation task, such type of visual data is usually fused with data from other types of sensors. The main aim of our research is to achieve the most accurate possible time synchronization between an event camera, LIDAR, and ambient environment sensors during a session of data acquisition. All the mentioned sensors as well as a stereo and a monocular camera were installed on a mobile robotic platform. In this work, a time synchronization architecture and algorithm are proposed for time synchronization with an implementation example on a PIC32 microcontroller. The overall time synchronization approach is scalable for other sensors where there is a need for accurate time synchronization between many nodes. The evaluation results of the proposed solution are reported and discussed in the paper.

In this work we introduce a new filter that extracts lines of particular widths. By using 1D model of line and edge profiles we demonstrate the problem of line extraction and propose the kernels and technique for processing of responses. We demonstrate the stability of the proposed filter in the presence of background gradients. Though the area of possible applications is broader, we use biometric palm images and an algorithm similar to Non-Halo Complex Matched Filtering to demonstrate that the proposed approach can extract only palm features (palm vein and palm print patterns) discarding any other irrelevant information, such as lighting conditions, contours, etc., simplifying further processing of the feature data.

In this paper we introduce a bank of fast matched filters that are designed to extract gradients, edges, lines and various line crossings. Our work is based on previously introduced filtering approaches like conventional Matched Filtering (MF), Complex Matched Filtering (CMF) and Generalized Complex Matched Filtering (GCMF), and is aimed to speed up the image processing. Filter kernel decomposition method is demonstrated for the latter mentioned (GCMF) but can be similarly applied to any other filters (like MF, CMF, Gabor filters, spiculation filters, steerable MF, etc.) as well. By introducing the mask kernel approximation, we show how to substitute the GCMF with several more computationally efficient filters, which reduce the overall computation complexity by over hundred of times. Acquired Fast GCMF retains all of the functionality of GCMF (extracts the desired objects and obtains their angular orientation), losing in accuracy only about +26 dB in terms of PSNR.