XDMOM: A Real-Time Moving Object Detection System Based on a Dual-Spectrum Camera.

A low-cost and power-efficient video surveillance system, named XDMOM, is developed for real-time moving object detection outdoors or in the wild. The novel system comprises four parts: imaging subsystem, video processing unit, power supply, and alarm device. The imaging subsystem, which consists of a dual-spectrum camera and rotary platform, can realize 360-degree and all-day monitoring. The video processing unit uses a power-efficient NVIDIA GeForce GT1030 chip as the processor, which ensures the power consumption of the whole system maintains a low level of 60~70 W during work. A portable lithium battery is employed to supply power so that the novel system can be used anywhere. The work principle is also studied in detail. Once videos are recorded, the single-stage neural network YOLOv4-tiny is employed to detect objects in a single frame, and an adaptive weighted moving pipeline filter is developed to remove pseudo-targets in the time domain, thereby reducing false alarms. Experimental results show that the overall correct alarm rate of the novel system could reach 85.17% in the daytime and 81.79% at night when humans are monitored in real outdoor environments. The good performance of the novel system is demonstrated by comparison with state-of-the-art video surveillance systems.

A Preliminary In Vitro Study of 3D Full-Field Strain Distribution in Human Whole Premolars Using Digital Image Correlation.

Full-field measurements can provide a more complete description of the behavior of human whole tooth under load. To that end, in vitro experiments were carried out to measure the full-field buccal surface strains of human premolars free of caries and abrasion using digital image correlation (DIC). Experimental results show that both the value field and the orientation field of strains can be observed exactly, both of which contain a wealth of information. Furthermore, the strain distributions between the crown and the root of specimens were significantly different (p < 0.001). An interesting observation was a watershed at the cementoenamel junction (CEJ) which separates the orientation field of strains into two distinct parts; the watershed was also observed in the value field of strains in some specimens whose geometries changed obviously at the CEJ. Another interesting observation was that the minor strains increased linearly from cervical to apical regions in the root cementum. Experimental results also support the viewpoint that mechanisms of non-carious cervical lesions (NCCLs) may in part be due to the changing orientation of tensile strains, as well as their magnitude, and they also support the hypothesis that occlusal force can contribute to root fractures.

Low-density point eating algorithm for surface reconstruction from dense scans.

We present a low-density point eating algorithm for surface reconstruction from dense scans. First, the density map for each scan is estimated and the boundary densities are down-weighted. Subsequently, the poorly scanned low-density overlapping points are eaten up based on a user-specified threshold. Finally, the overlapping areas are thinned by using the moving least-squares operator and the homogeneous points are weighted averaged. The new algorithm can extract smooth but detailed point set surfaces that are as close as possible to the ground truth. The good performance of the new algorithm is demonstrated by comparison with several advanced surface reconstruction algorithms.

Exposure-fusion-based dot-grid image acquisition and recognition for sheet metal strain analysis.

Dot-grid images are usually captured for grid strain analysis during sheet metal forming. Due to the strong reflective characteristic of the metallic surfaces, the recorded dot-grid images often have poor quality, low positioning accuracy, and low recognition rate. Therefore, an exposure-fusion-based dot-grid image acquisition and recognition approach is proposed. First, multiple dot-grid images are captured at different exposure levels. Subsequently, the recorded multi-exposure dot-grid images are fused into a new high-quality dot-grid image based on exposure fusion technology. Finally, a dot-grid image recognition procedure is developed to detect the dot-grids in the new dot-grid image. Both synthetic and real dot-grid images were tested to verify the performance of the novel approach. When synthetic dot-grid images were tested, the maximum positioning error was up to 6.044 pixels if they were recognized in the traditional way, whereas the maximum positioning error was reduced to 0.132 pixels if the novel approach was adopted. When real dot-grid images were tested, the lowest recognition rate is only 50.52% if they were recognized in the traditional way. Nevertheless, the recognition rate can reach about 91% if the novel approach was employed. These experimental results show the superiorities of the novel approach.

Guide to quickly build high-quality three-dimensional models with a structured light range scanner.

In literature, there are many reports about how to build high-precision, high-speed, and/or flexibly structured light range scanners. Whereas, there are few papers reported about how to implement the scanning to reach the full potential of the scanners when digitizing various objects. In this paper, some scanning strategies adopted by the structured light range scanner XJTUOM are introduced. In order to build high-quality three-dimensional (3D) models for various applications, such as 3D inspection, if necessary, the object surface is prepared in advance by an application of coating spray. Then, a precise 3D coordinate control network (CNN) is established to control the overall measurement accuracy. When the sensor is adjusted to face the object, a visual measuring volume, which can guide the scanning, is automatically aligned to the established CNN. Meanwhile, to maintain the local region scanning accuracy, simple rules are developed to check the low-quality regions in each scanning. Finally, the advantages and limitations of these scanning strategies are discussed in detail. We hope our work will be helpful in order for others to make their own scanning plans with similar optical devices at hand.

Density adaptive trilateral scan integration method.

We present a trilateral scan integration method to extract a desirable single-layer, smooth, detailed point set surface. A density adaptive overlapping areas detection algorithm is developed to detect overlapping points in raw scans. Subsequently, a variant trilateral filter is designed to shift the overlapping points to a single layer surface as well as reduce noises and outliers. Finally, a mean-shift clustering scheme is employed to gather corresponding points iteratively, and the local maxima model replaces members in the stable cluster, thereby removing redundancies. Experimental results demonstrate the superiority of the new method.