Visual recognition for urban traffic data retrieval and analysis in major events using convolutional neural networks

Accurate and prompt traffic data are necessary for the successful management of major events. Computer vision techniques, such as convolutional neural network (CNN) applied on video monitoring data, can provide a cost-efficient and timely alternative to traditional data collection and analysis methods. This paper presents a framework designed to take videos as input and output traffic volume counts and intersection turning patterns. This framework comprises a CNN model and an object tracking algorithm to detect and track vehicles in the camera’s pixel view first. Homographic projection then maps vehicle spatial-temporal information (including unique ID, location, and timestamp) onto an orthogonal real-scale map, from which the traffic counts and turns are computed. Several video data are manually labeled and compared with the framework output. The following results show a robust traffic volume count accuracy up to 96.91%. Moreover, this work investigates the performance influencing factors including lighting condition (over a 24-h-period), pixel size, and camera angle. Based on the analysis, it is suggested to place cameras such that detection pixel size is above 2343 and the view angle is below 22°, for more accurate counts. Next, previous and current traffic reports after Texas A&M home football games are compared with the framework output. Results suggest that the proposed framework is able to reproduce traffic volume change trends for different traffic directions. Lastly, this work also contributes a new intersection turning pattern, i.e., counts for each ingress-egress edge pair, with its optimization technique which result in an accuracy between 43% and 72%.

Virtual histology of archaeological human deciduous prenatal enamel through synchrotron X-ray computed microtomography images

Virtual histology is increasingly utilized to reconstruct the cell mechanisms underlying dental morphology for fragile fossils when physical thin sections are not permitted. Yet, the comparability of data derived from virtual and physical thin sections is rarely tested. Here, the results from archaeological human deciduous incisor physical sections are compared with virtual ones obtained by phase-contrast synchrotron radiation computed microtomography (SRµCT) of intact specimens using a multi-scale approach. Moreover, virtual prenatal daily enamel secretion rates are compared with those calculated from physical thin sections of the same tooth class from the same archaeological skeletal series. Results showed overall good visibility of the enamel microstructures in the virtual sections which are comparable to that of physical ones. The highest spatial resolution SRµCT setting (effective pixel size = 0.9 µm) produced daily secretion rates that matched those calculated from physical sections. Rates obtained using the lowest spatial resolution setup (effective pixel size = 2.0 µm) were higher than those obtained from physical sections. The results demonstrate that virtual histology can be applied to the investigated samples to obtain reliable and quantitative measurements of prenatal daily enamel secretion rates.

Matters of choosing the pixel size in topographic aerial photography

The existing regulatory documents on photogrammetric works are technologically outdated. They neither take into account the peculiarities of aerial photography with digital cameras, the navigation equipment used and modern image processing methods, nor regulate the technique of calculating the pixel size on the ground. In order to select the pixel size in the terrain for aerial photography with topographic requirements concerning to the results, the method of multivariate analysis of the input data is proposed. It is supposed to ensure the minimum pixel size on the ground according to such criteria as the accuracy of the aerial triangulation results, the accuracy of building a digital elevation model for orthotransformation, the possibility of the objects interpretation with a specified minimum size and consideration of camera exposure parameters. To determine the accuracy criteria, we used formulas for pre- calculation of spatial phototriangulation accuracy with multiple choice of parameters. Examples of pixel size selection in the terrain at designing aerial photography by an amateur camera for the purposes of large-scale mapping are considered. Conclusions on the necessity of solving the issues of selecting parameters of large scale aerial photography, taking into account multiple input data and used aerial survey equipment are made.

Mapping Gaps in Sugarcane by UAV RGB Imagery: The Lower and Earlier the Flight, the More Accurate

Imagery data prove useful for mapping gaps in sugarcane. However, if the quality of data is poor or the moment of flying an aerial platform is not compatible to phenology, prediction becomes rather inaccurate. Therefore, we analyzed how the combination of pixel size (3.5, 6.0 and 8.2 cm) and height of plant (0.5, 0.9, 1.0, 1.2 and 1.7 m) could impact the mapping of gaps on unmanned aerial vehicle (UAV) RGB imagery. Both factors significantly influenced mapping. The larger the pixel or plant, the less accurate the prediction. Error was more likely to occur for regions on the field where actively growing vegetation overlapped at gaps of 0.5 m. Hence, even 3.5 cm pixel did not capture them. Overall, pixels of 3.5 cm and plants of 0.5 m outstripped other combinations, making it the most accurate (absolute error ~0.015 m) solution for remote mapping on the field. Our insights are timely and provide forward knowledge that is particularly relevant to progress in the field’s prominence of flying a UAV to map gaps. They will enable producers to make decisions on replanting and fertilizing site-specific high-resolution imagery data.

Adaptive Video Caption Detection and Extraction Method Based on Color Filtering Principle

Based on the principle of color (RGB) filtering, an improved adaptive video caption detection and extraction method is proposed. Firstly, the principle of the color filtering algorithm used in the video caption detection and extraction method is analyzed, and then the algorithm is improved adaptively according to the caption pixel size to filter the noise. Finally, experiments verify the effect of this method in extracting subtitles from video. The experimental results show that the accuracy of detecting and extracting subtitles in color video is as high as 99.3%.

Simulation study on the performance of time-over-threshold based positioning in monolithic PET detectors

The vast majority of PET detectors in the field today is based on pixelated scintillators. Yet, the resolution of this type of detector is limited by the pixel size. To overcome this limitation one can use monolithic detectors. However, this detector architecture demands specific and high-speed detector readout of the photodetector array. A commonly used approach is to integrate the current pulses generated by every pixel but such circuitry quickly becomes bulky, power consuming and expensive. The objective of this work is to investigate a novel readout and event positioning scheme for monolithic PET detectors, based on Time-over-Threshold (ToT). In this case, we measure the time that the pulse is above a certain threshold through a comparator. The pulse widths are used for event positioning using a mean nearest neighbour approach (mNNToT). For energy determination one integrating multiplexed channel is foreseen. We evaluate the positioning accuracy and uniformity of such a ToT detector by means of Monte Carlo simulations. The impact of the threshold value is investigated and the results are compared to a detector using mean nearest neighbour with pulse-integration (mNNint), which has already proven to allow sub-mm resolution. We show minimal degradation in spatial resolution and bias performance compared to mNNint. The highest threshold results in the worst resolution performance but degradation remains below 0.1 mm. Bias is largely constant over different thresholds for mNNToT and close to identical to mNNint. Furthermore we show that Time-over-Threshold performs well in terms of detector uniformity and that scattered photons can be positioned inside the crystal with high accuracy. We conclude from this work that ToT is a valuable alternative to pulse-integration for monolithic PET detectors. This novel approach has an impact on PET detector development since it has the advantage of lower power consumption, compactness and inherent amplitude-to-time conversion.

PLANAR SCINTIGRAPHY IMAGE DE-NOISING USING COIFLET WAVELET

The planar scintigraphic image usually has poor resolution and contains noise. This noise can be removed using the coiflet wavelet method so that the image quality gets better. This coiflet wavelet method is a noise reduction method based on frequency analysis. The planar scintigraphy image is the reconstructed image of the gamma radiation count data (phantom with the Cs-137 source in it). The original image is 15×15 pixel. Before the de-noising process, the image went through an interpolation process, which is to increase the pixel size of the image. The original image enlarged to 70×70, 480×480, and 1200×1200 pixel. After de-noising with coiflet wavelet, the image quality is measured based on MSE and PSNR parameters. The resulting images are quite good, with MSE values are close to zero and PSNR values of more than 60 dB. The smaller the MSE and the bigger the PSNR, is getting the better the image quality. In this study, the results show that the 1200×1200 pixel image has the best quality. It means that the image enlargement process has a good effect on the de-noising process, especially if the original image has a low resolution.

Detection and Erasing Scribble Blackboard System Based on Hough-Transform Method Using Camera

Teachers of a secondary school in Semarang city are often exposed to blackboard chalk dust. This will give a significant impact if it occurs at a fairly frequent intensity and a long period of time. An automatic scratch detector and designing a device that can erase the blackboard is a preventive step that can reduce the long-term impact of chalk dust. The scratch detector uses a circle shape parameter as a mark of dirty position that needs to be erased. Light can affect the system performance. The system works properly at light intensities ranging from ± 160 to ± 200 lux. Testing the threshold value proves that the system can detect circles in the range of 40 - 55. The pixel size which is detected by the camera was 640x480 will allow the system to divide the blackboard into 9 mapping areas. The mapping area is differentiated into 9 sections so that the x and y coordinate positions of the blackboard dirty spot can be determined. A mechanical execution will erase the top and bottom areas according to the position of the detected mapping area. The success of scratch detector reaches 81.8%..

Discrete Environment-Driven GPU-Based Ray Launching: Validation and Applications

In this work, the Discrete, Environment-Driven Ray Launching (DED-RL) algorithm, which makes use of parallelization on Graphic Processing Units, fully described in a previous paper, has been validated versus a large set of measurements to evaluate its performance in terms of both computational efficiency and accuracy. Three major urban areas have been considered, including a very challenging scenario in central San Francisco that was used as a benchmark to test an image-ray tracing algorithm in a previous work. Results show that DED-RL is as accurate as ray tracing, despite the much lower computation time, reduced by more than three orders of magnitude with respect to ray tracing. Moreover, the accuracy level only marginally depends on discretization pixel size, at least for the considered pixel size range. The unprecedented computational efficiency of DED-RL opens the way to numerous applications, ranging from RF coverage optimization of drone-aided cellular networks to efficient fingerprinting localization applications, as briefly discussed in the paper.