With the increasing level of modern computer technology and the rapid development of artificial intelligence, various industries have begun to adopt intelligent devices in large quantities to reduce the waste of human...
详细信息
With the increasing level of modern computer technology and the rapid development of artificial intelligence, various industries have begun to adopt intelligent devices in large quantities to reduce the waste of human resources. In the field of garbage disposal, sweepers use image acquisition devices to obtain street and road images, and recognize and detect street garbage through image information. The 'You Only Look Once' (YOLO) algorithm is used to deploy and optimize the garbage recognition and detection software. The results showed that in the brightness balanced scene, the research method had a grayscale mean of 2108, with a slightly lower proportion of poorly exposed pixels at 6.7 parts per thousand. By reducing the exposure time, the grayscale mean was lowered to 1978, significantly reducing the proportion of poorly exposed pixels to 12.7%. For high contrast scenes, this research method faced a high exposure pixel ratio of 146.8 parts per thousand, which was 146.8 parts per thousand of the saturation threshold. It increased the gray mean to 1748 and greatly reduced the proportion of poorly exposed pixels to an extremely low level. After single frame calculation, the research method time increased to 1.643 ms, and after stabilization, the research method time was 1.886 ms, with a distance ratio of only 12.2%, which was more stable. Compared with the YOLO real-time garbage detection method, the YOLOv5s+TensorRT proposed in this study could detect garbage in near real time on low-cost embedded devices in terms of FPS or inference speed, with the YOLO real-time garbage detection method exceeding 32fps. In terms of accuracy, YOLOv5+TensorRT optimization could maintain a good accuracy of 99%, and had certain advantages in comprehensive performance, making it easier to promote and apply in practical applications. It can be seen that the methods proposed in the study have solved the limitations of traditional garbage cleaning methods, improved the efficiency and qua
In order to solve the problem of low speed and accuracy during the ticket image information recognition, which is due to uncertain direction of ticket image, a new algorithm based on regionsegmentation was proposed t...
详细信息
In order to solve the problem of low speed and accuracy during the ticket image information recognition, which is due to uncertain direction of ticket image, a new algorithm based on regionsegmentation was proposed to detect and correct the ticket image direction in this paper. In the algorithm, based on the analysis of ticket image, divide the ticket image split into subregions and demarcate it, combine with the histogram statistics of the calibrated region, then find out the difference of eigenvalues between different calibrated regions, determine the direction of the calibrated region, and judge the direction of the ticket image. On this basis, the image geometric transformation is used to correct the direction of the ticket image. The experimental results show that the correct rectification rate of the algorithm is 98.02%, and the algorithm is simple and easy to hardware implementation. It can improve the efficiency of the ticket recognition algorithm.
In order to solve the problem of the vehicle body color extraction and recognition,this paper presents an identification method of the intelligent *** the system,the red and the white paper shells are used as the body...
详细信息
In order to solve the problem of the vehicle body color extraction and recognition,this paper presents an identification method of the intelligent *** the system,the red and the white paper shells are used as the body of the intelligent vehicle independently to distinguish both *** the same time,the existing image segmentation method had been *** the condition of the unique main image color obtained by image sensor,an improved segmentationalgorithm for image color extraction and recognition was proposed for the drawbacks of the existing algorithms,and experiment simulation was conducted by MATLAB *** simulation results reveal that the algorithm not only can effectively extract and identify the body color of the intelligent vehicle,but also greatly improve the speed of the computer.
暂无评论