Three-dimensional Spatial Localization Based on Binocular Vsion

In order to solve the problems of low accuracy and unstable system performance existing in binocular vision alone, this paper proposes a threedimensional space recognition and positioning algorithm based on binocular stereo vision and deep learning algorithms. First, a binocular camera for Zhang Zhengyou calibrated by several adjustments, calibration error will eventually set at 0.10pixels best, select and SAD in block matching algorithm in the algorithm, the matching point of the search range reduction, mitigation data for subsequent experiments burden. Then input the three-dimensional spatial data calculated by using the binocular ”parallax” principle into the Faster R-CNN model for data training, extract and classify the target features, and finally realize real-time detection of the target object and its position coordinate information. The analysis of experimental data shows that when the best calibration error is selected and the number of data training is sufficient, the algorithm in this paper can effectively improve the quality of target detection. The positioning accuracy and target recognition rate are increased by about 3%-5%, and it can achieve faster fps.

The Application of Visual Sensors in the Artistic Design of College Campus Landscape

The research purpose is to solve the problems of low efficiency, low accuracy, and high cost of traditional environmental landscape mapping and landscape volume measurement methods in the artistic design of college campus landscape and make up the defects that the traditional campus monitoring is vulnerable to adverse weather, which results in low monitoring accuracy. Primarily, a binocular stereo vision measurement based on Scale Invariant Feature Transform (SIFT) matching algorithm is proposed, which can realize accurate collection of environmental spatial information and measurement of landscape volume without contact in the process of campus landscape design. Then, the visual monitoring system of college landscape based on the Internet of Things (IoT) is constructed to realize real-time monitoring and early warning of human damage to campus landscape. The proposed method is verified by actual measurement of different objects and simulation experiments using simulation software. Ultimately, the application of visual sensors in artistic design of college campus landscape is analysed by literature analysis. The results show that (1) the error of the improved binocular stereo vision measurement designed here is 52.32% and 59.69% lower than that of the traditional measurement method when measuring the same object with different volumes and the volumes of different objects, respectively, which indicates that the measurement accuracy of the new method is higher. (2) The proposed landscape visual monitoring method based on IoT improves the image recognition accuracy by 21% compared with the traditional digital image monitoring method. The average recognition time is shortened by 12 ms, which ensures the accuracy and improves the recognition efficiency. (3) Through the analysis of existing literature, it is found that the binocular stereo vision sensor can be used to monitor the whole process of landscape construction in real time. The sensor can be combined with social networks, mobile terminals, and physiological monitoring equipment to comprehensively analyse and evaluate people’s preference for campus landscape. The proposed method has broad application prospects in campus landscape design, construction, and maintenance. The research purpose is to provide important technical support for the improvement of the overall image of the college campus and even the city for the design of landscape environment and the technical upgrading of maintenance work in the college campus.

Pose Estimation under Visual Sensing Technology and Its Application in Art Design

The study is aimed at solving the problem of large measurement errors caused by the binocular camera in traditional 3D art design, which leads to inaccurate 3D information of the target. The contour information extraction in the process of human motion pose reconstruction is easily affected by the noise in the image. Therefore, a binocular stereo vision system is built first and it integrates image acquisition, camera calibration, and image processing. The dedistortion method is used to process the image because it can reduce errors. Second, a three-dimensional human motion pose reconstruction model is implemented, the Gaussian template is used to remove the noise in the image frame, and the change detection template (CDM) is used to solve the problem of background “exposure” and “occlusion.” Finally, simulation experiments are designed to verify the system and model designed. Since the research on the application of pose estimation based on visual sensing technology in art design is still blank, such research has great significance and provides a reference for the research in the field. The literature analysis is used to expound and analyze the application of pose estimation based on visual sensing technology in visual communication design and environmental art design: (1) although the binocular stereo vision system causes some errors in the measurement, the overall error is controlled within 2% and the accuracy is high, which proves that it can be applied to the acquisition of three-dimensional information of the target in art design; (2) there is a high degree of fitting between the video sequence data created by the three-dimensional human motion pose reconstruction model designed and the real motion data, which indicates that this method has high accuracy in processing image sequences and the feasibility of applying it to human pose reconstruction in three-dimensional art design is high; (3) through the analysis of the existing literature, it is found that most of the current visual-based attitude assessment studies are carried out by using network cameras combined with computers, and the quality of the obtained images is low. The combination of binocular stereo sensor and attitude estimation technology can be applied to the design of advertising, animation, games, and packaging, making the behavior of virtual characters in animation and games more vivid. The combination provides convenience for the collection of environmental spatial information and object attitude information, the formulation of a design scheme, and real-time monitoring of construction in environmental art design. The purpose of this study is to provide an important theoretical basis for the technical upgrading of art design.

An iterative refinement method of point cloud for binocular vision 3D reconstruction

This paper introduces a new binocular stereo deep learning network based on point cloud, which can realize higher precision point cloud reconstruction through continuous iteration of the network. Our method directly carries out point cloud processing on the target, calculates the difference between the current depth map and the real depth, estimates the loss according to the predicted point cloud and the information of the dual view input image, and then uses the appropriate loss function to iteratively process the point cloud. In addition, we can customize the number of iterations to achieve higher precision point cloud effect. The proposed network basically achieves good results on KITTI data set.