Online Multi-Object Tracking (MOT) has wide applications in time-critical video analysis scenarios, such as robot navigation1 and autonomous driving2. In tracking by-detection, a major challenge of online MOT is how to robustly associate noisy object detections on a new video frame with previously tracked objects. This paper aims to build technology that can track a movement of people via surveillance cameras that are located in stores, but not only (theoretically, the algorithm may be applicable to the location of the camera at any premises). The algorithm works with a variety of camera angles that allows. The main innovation of the paper is that algorithm SORT has been updated to consider the difference between datasets used on competitions and the real ones. The difference is that recognition is not perfect in data created by the program. People’s contours may be of different size (rectangles corresponding to the same man may differ twice) and some of them may be not recognized. The new metric of proximity called “soft-iou” has been introduced in SORT. We have achieved the accuracy of 95% for the daily number of visitors for one of jewelry retail chains. This level of accuracy allows applying the algorithm in different areas: not only retail stores, but also shopping centers, sports events, performances, traffic in public transport, etc.
Stereo matching is one of the most important computer vision tasks. Several methods can be used to compute a matching cost of two pictures. This paper proposes a method that uses convolutional neural networks to compute the matching cost. The network architecture is described as well as teaching process. The matching cost metric based on the result of neural network is applied to base method which uses support points grid (ELAS). The proposed method was tested on Middlebury benchmark images and showed an accuracy improvement compared to the base method.
The paper is devoted to the problem of correct vergence in the mixed reality systems. There are two main types of mixed reality devices, these are optical see-through and video see-through mixed reality systems. They differ by the way how the device user observes the real-world environment: in the optical see-through mixed reality systems the real world is observed directly and as result the vergence and accommodation conflicts are absent while in the video see-through mixed reality systems there present both vergence and accommodation conflicts. The vergence conflict is caused by the fixed orientation of the device cameras that transmit the image of the real world, and the accommodation conflict is caused by the constancy of the focal length of optical systems that form the images of the real world on the human eye retina from the CCD matrices. This paper presents the research of the possibility to restore the correct convergence of a real-world image formed by video see-through mixed reality systems. The research shows how convergence can be correctly reconstructed using images captured by fixed cameras with parallel axes located at the front side of the mixed reality headset that form images for the left and right eyes. The resulting reconstructed images are almost free from any visual artifacts. Additionally, authors show how errors in scene depth map reconstruction affects the vergence restoration quality.
Optical components are being used more and more extensive in the most areas of production and human life, therefore improving the quality control of optical elements manufacture is high priority task when manufacturing any type of optical element. Every aspect of the production process directly impacts the final product quality and reliability level. In the process of optical elements manufacture, after determining the overall product quality requirements, the next most important thing is to clarify and implement the quality indicators of each product subsystem. So, after the quality requirements of the optical element is determined, the next step that should be done to ensure the quality of final product is measurements and testing work by the professional trained staff. The measurement and test job itself is also a system engineering, because it not only includes the metering and testing hardware (e.g. instruments, equipment, etc.) but also includes software work such as metering management and metering transmission. The quality control of optical components is a process that includes both quality management and the implementation of specific processing and testing methods. Optical component processing plants can use the following methods to improve and control the consistency of the quality of the finished optical elements: learn and use new optical component processing techniques, use numerical control technology, use advanced processing and inspection equipment, proficient in process inspection and finished product inspection methods. In this article we present an example of implementation of the quality control phase in the optical elements manufacture process.
Femtosecond laser has the feature of ultra-short and ultra-fast. In recent years, femtosecond laser has been used extensively in the field of new energy, automobiles, large aircraft, and biomedical fields. The femtosecond laser induced surface structure has the advantages of simple process, fast processing speed and high fineness, and it is one of the most widely used surface structure preparation method. Femtosecond laser induced surface periodic structures have wide range of applications in many areas, such as hydrophobic surface or hydrophilic surface, color effect, surface-enhanced Raman scattering(SERS), and so on. Inspired by micro-nanostructures with unique functions on various of biological surfaces in nature, researchers use femtosecond laser micro-nanofabrication produce technology to prepare controlled periodic micro-nano structures on metals, dielectrics, and semiconductor surfaces. Thus playing a role in improving the surface properties of metals. For example, butterflies and birds have very beautiful colors, superhydrophobicty of lotus leaves, shark skin can reduce frictional resistance, strong adsorption capacity of gecko claws, and so on. The improvement of material surface properties can bring about the realization of new functions and the development of related technology, it can effectively serve the production and life of human beings. This article introduces the development and research status of the surface periodic structure induced by femtosecond laser, and discusses the influence of materials, laser pulse parameters, processing methods and processing environment on the surface periodic structure. The application prospect of femtosecond laser induced surface periodic structure are pointed out, and the challenges it faces at present stage are summarized.
In the long-term production and processing practice, China's optical manufacturing industry has formed a coexistence model with traditional optical classical polishing and high-speed polishing. China’s optical experts are familiar with many optical standards such as the US military standard, German standard, ISO10110, and China national standard, promote optical elements production which can meet the needs of high precision instruments. High efficiency production methods are used to lay a foundation for meeting the market demands of China and other countries. The application of virtual prototyping methods, the emergence of a large number of professional optical specialists, and the popularization of advanced processing and testing equipment have all contributed to the development of China's optical industry. The number, shape and type of the optical components and test requirements are critical to the mass production of the optical components. An optical manufacturer’s sustainable operations not only depend on professional lens design, but also rely on advanced processing and testing equipment, effective cost control management and China's mature industrial chain. In this paper, we analyze the current state and characteristics of China's optical industry from the aspects of component design, processing methods, optical standards, processing and inspection equipment, and cost control. The cost information is valid for China. We propose a way to optimize a mass production cost which is based on the complexity of the optical system. As result it makes it possible to produce complex optical systems in China. To demonstrate the efficiency of the proposed method of analysis, we present examples of the lenses produced.
In now days the devices of augmented and mixed realities are widely used in all areas of the human live. However, these devices still have the visual perception conflicts. These conflicts may result in a discomfort which can lead to a headache or other health problems. The source of this conflict is disagreement between the expected and seen images. In this work we propose a way for an analysis of the visual perception problems of these devices. The virtual prototyping approach is used to synthesize and compare the physically correct images of one seen using the mixed reality device with the expected one.
Mass-production of optical and opto-electronic devices is very critical to the number of the elements, their shape and type. Physical prototyping for this optical system depends not only on the lens quality and quantity, but also on the camera housing and the sensor. In our research we’ve studied the reasons of optical systems’ complexity and found out that it depends on optical system structure. After that a lens is optimized, tolerances are calculated and physical prototyping is implemented. In our case physical prototyping means a trial lot production. It was found, that virtual prototyping for different starting points can help to find the solution for mass-production, which will be then used for its manufacturing. This approach includes calculation of all elements of the camera lens from the first surface of a lens to the sensor, including 3D environment model. Depending on the type of the designed optical system, the virtual prototype modeling is included into a merit function for further optimization in the automatic lens design software. We can estimate a reduction of visual contrast of the image of the camera lens in the conditions similar to the conditions of the real device operation or show dependence of the contrast reduction in time of the operation. We implemented such an approach as a virtual parametric model of the lens camera device. We show a plot of the dependence on the cost of mass production of the mobile phone camera from the number of components. The cost information is valid for China.