Abstract: A video file processing method for making and outputting an event film according to video files is provided. The method includes: acquiring at least one video file and detecting a plurality of events and an occurring time of each event in the video file; extracting a plurality of event clips including the events from the video file according to the occurring time; obtaining an event film by assembling a plurality of concerned clips in the event clips according to a merging rule; and outputting the event film. In addition, a video file processing device and a monitoring system using the method are also provided.

Abstract: A method and a computer system for object tracking are provided. The method includes the following steps. An image sequence including multiple images with a target object is obtained. A labelling operation corresponding to the target object in first two images and last two images in the image sequence is received to respectively generate four ground truth labels of the target object. A forward tracking of the target object is performed on the image sequence in time series according to the ground truth labels of the first two images to obtain a forward tracking result. A backward tracking of the target object is performed on the image sequence in time series according to the ground truth labels of the last two images to obtain a backward tracking result. The forward and the backward tracking results are compared to accordingly generate a final tracking result of the target object.

Abstract: A method and a computer system for object tracking are provided. The method includes the following steps. An image sequence including multiple images with a target object is obtained. A labelling operation corresponding to the target object in first two images and last two images in the image sequence is received to respectively generate four ground truth labels of the target object. A forward tracking of the target object is performed on the image sequence in time series according to the ground truth labels of the first two images to obtain a forward tracking result. A backward tracking of the target object is performed on the image sequence in time series according to the ground truth labels of the last two images to obtain a backward tracking result. The forward and the backward tracking results are compared to accordingly generate a final tracking result of the target object.

Abstract: A method and a computing device for monitoring an object are provided, where the method includes the following steps. An image sequence captured on a monitored area is received. Whether a target object is within a region of interest (ROI) of the monitored area is determined. If the determination is not affirmative and at least one first moving object is detected, whether the first moving object enters the ROI is determined according to a relative position between the first moving object and a detection boundary as well as whether the first moving object has continuous movements in the ROI. If the determination is affirmative, whether the target object leaves the ROI is determined according to a relative position between at least one second moving object associated with the target object and the detection boundary as well as whether the second moving object has continuous movements in a non-ROI.

Abstract: A video file processing method for making and outputting an event film according to video files is provided. The method includes: acquiring at least one video file and detecting a plurality of events and an occurring time of each event in the video file; extracting a plurality of event clips including the events from the video file according to the occurring time; obtaining an event film by assembling a plurality of concerned clips in the event clips according to a merging rule; and outputting the event film. In addition, a video file processing device and a monitoring system using the method are also provided.

Abstract: A drowsiness detection apparatus is provided. The drowsiness detection apparatus: an image capturing unit, an alarm apparatus, and a drowsiness determination circuit. The image capturing unit is configured to repeatedly capture a plurality of facial images of a user. The drowsiness determination circuit is configured to obtain an eye region from a current image of the facial images, calculate a vertical histogram of the eye region, and determine an eye part from the eye region according to the vertical histogram. The drowsiness determination circuit further compares the pixel height of the eye part with a closed-eye pixel threshold to perform closed-eye detection, and performs drowsiness detection on the user according to the results of the closed-eye detection. When the result of the drowsiness detection indicates that the user is drowsy, the drowsiness determination circuit transmits a control signal to the alarm apparatus to sound an alarm.

Abstract: A method and a computing device for monitoring an object are provided, where the method includes the following steps. An image sequence captured on a monitored area is received. Whether a target object is within a region of interest (ROI) of the monitored area is determined. If the determination is not affirmative and at least one first moving object is detected, whether the first moving object enters the ROI is determined according to a relative position between the first moving object and a detection boundary as well as whether the first moving object has continuous movements in the ROI. If the determination is affirmative, whether the target object leaves the ROI is determined according to a relative position between at least one second moving object associated with the target object and the detection boundary as well as whether the second moving object has continuous movements in a non-ROI.

Abstract: A video editing method, a video editing device, and a video editing system are provided. A target is obtained from an original video. Relative position relations respectively between each frame and a previous frame and a next frame adjacent to it are obtained based on multiple image feature points of each frame. An adjustment process is executed respectively for each frame. The adjustment process includes: obtaining a frame N; identifying the target in the frame N; adjusting the frame N to obtain a new frame, wherein the target is located on a center of the new frame; and modifying the new frame based on the relative position relations to obtain a postproduction frame. A processed video is outputted based on multiple postproduction frames.

Abstract: A video editing method, a video editing device, and a video editing system are provided. A target is obtained from an original video. Relative position relations respectively between each frame and a previous frame and a next frame adjacent to it are obtained based on multiple image feature points of each frame. An adjustment process is executed respectively for each frame. The adjustment process includes: obtaining a frame N; identifying the target in the frame N; adjusting the frame N to obtain a new frame, wherein the target is located on a center of the new frame; and modifying the new frame based on the relative position relations to obtain a postproduction frame. A processed video is outputted based on multiple postproduction frames.

Abstract: A drowsiness detection apparatus is provided. The drowsiness detection apparatus: an image capturing unit, an alarm apparatus, and a drowsiness determination circuit. The image capturing unit is configured to repeatedly capture a plurality of facial images of a user. The drowsiness determination circuit is configured to obtain an eye region from a current image of the facial images, calculate a vertical histogram of the eye region, and determine an eye part from the eye region according to the vertical histogram. The drowsiness determination circuit further compares the pixel height of the eye part with a closed-eye pixel threshold to perform closed-eye detection, and performs drowsiness detection on the user according to the results of the closed-eye detection. When the result of the drowsiness detection indicates that the user is drowsy, the drowsiness determination circuit transmits a control signal to the alarm apparatus to sound an alarm.

Abstract: A hover-control method for multirotor unmanned aerial vehicle (UAV) upcast at any angle to take off is applied to correct at least one tilt angle of the multirotor UAV by dual-feedback schemes. When any of the at least one tilt angle is greater than a corresponding angle limitation, an open feedback scheme is executed to drive a plurality of motors of the multirotor UAV until a predetermined condition is achieved. When all of the at least one tilt angle are not larger than the respective angle limitations, a negative feedback scheme is executed to drive the motors of the multirotor UAV until a hover condition is achieved.

Abstract: A hover-control method for multirotor unmanned aerial vehicle (UAV) upcast at any angle to take off is applied to correct at least one tilt angle of the multirotor UAV by dual-feedback schemes. When any of the at least one tilt angle is greater than a corresponding angle limitation, an open feedback scheme is executed to drive a plurality of motors of the multirotor UAV until a predetermined condition is achieved. When all of the at least one tilt angle are not larger than the respective angle limitations, a negative feedback scheme is executed to drive the motors of the multirotor UAV until a hover condition is achieved.