5results about How to "Image data is accurate" patented technology

The invention discloses a weld defect detection system and method based on target recognition, and the system comprises an image preprocessing module, a weld defect analysis module and a weld quality analysis module, relates to the technical field of image processing, and solves the problem that data formats and types obtained by different detection methods are different. In order to solve the technical problems of difficulty in effective integration and comprehensive analysis and proneness to erroneous judgment and missing judgment in the prior art, an image preprocessing module carries out graying, image enhancement and denoising processing on a weld seam image, so that the characteristics of weld seam defects can be highlighted, and the definition and contrast of the image are improved; according to the method, more accurate image data is provided for subsequent defect analysis, the image data is combined with historical data, and the relationship among defect types, gray features, defect reasons and the like is established by sorting and analyzing a large amount of historical data, so that the welding seam quality is evaluated more scientifically.

The present invention belongs to the field of medical image processing, and provides a method, system and medium for processing medical endoscopic images based on GPU. The present invention utilizes the powerful parallel processing capability of GPU to significantly improve the speed and accuracy of endoscopic image acquisition, preprocessing, feature point detection, image correction and pathological feature recognition. The image quality is enhanced by executing GPU-accelerated filtering algorithms and histogram equalization, while the GPU-optimized SIFT algorithm effectively improves the accuracy of feature point matching. The estimation of the affine transformation model and the transformation correction of the image further correct the distortion and ensure the accuracy of the image data. Finally, the pre-trained deep learning model is used to implement fast convolutional neural network forward propagation on the GPU, identify pathological features in real time, and assist clinical diagnosis. The efficiency and accuracy of clinical diagnosis are greatly improved, providing doctors with instant diagnostic assistance, thereby improving the quality of medical services.

The invention relates to the technical field of lenses, and discloses a periscopic light splitting lens module, which comprises a light steering assembly, a light splitting assembly, a direction adjusting assembly and a lens module, and is characterized in that the light splitting assembly is used for split light receiving and emitting red light, green light and blue light to a light inlet surface; the direction adjusting assembly is used for emitting incident light to the light splitting assembly and gathering light emitted by the light emitting face of the light steering assembly to the lens module. The lens module comprises a base, a lens assembly installed in the base, a sensor assembly installed on the base and a circuit board installed outside the base. Through the arrangement of the light splitting assembly, red light, green light and blue light in incident light are respectively absorbed and transmitted to the light steering assembly, so that partial waves are lengthened, and more accurate and abundant image data are provided; after light splitting, the propagation speeds of the same wavelength are consistent, and red light, green light and blue light enter and are transmitted at the same time, so that image quality is guaranteed, stray light is reduced, and an IR sheet is replaced to avoid halo phenomenon, uneven color and the like.

The invention relates to the technical field of image processing, and discloses a step exposure and optical coating combined method and system, and the method comprises the steps: collecting instrument panel image information by a camera, calculating a refresh frequency according to a gray difference, and automatically adjusting the opening and closing time of a shutter according to the refresh frequency. The shutter opening and closing time of the camera is kept synchronous with the refresh frequency of the instrument panel; image information of the instrument panel of the instrument is collected again, whether an optical filter in the camera needs to be adjusted or not is judged, and when it is judged that the optical filter needs to be adjusted, the light transmittance of the optical filter is adjusted according to the light intensity; continuously acquiring image information of the instrument panel, and judging whether the exposure time of the camera needs to be adjusted or not; and acquiring the information of the environment where the camera is located, and adjusting the exposure time of the camera. According to the invention, the camera can obtain accurate and clear image data under various illumination conditions.

This invention relates to the field of computer vision, and more particularly to a crankshaft dimension detection method based on computer vision. The method includes: acquiring original crankshaft images from different viewpoints and performing image preprocessing to extract the crankshaft contour; identifying crankshaft contour feature points based on the contour and constructing a crankshaft contour feature point set; calculating preliminary dimensions based on the crankshaft contour feature points; and introducing a geometric dynamic projection correction algorithm to optimize the preliminary dimensions, obtaining optimized crankshaft dimensions. This method solves the problems of inaccurate crankshaft dimension measurement due to projection distortion in traditional multi-view image acquisition processes; the inability of conventional image processing techniques to effectively extract accurate geometric features due to uneven lighting and noise interference on the crankshaft surface; and the failure of traditional methods to fully consider projection distortion at each viewpoint, thus failing to achieve reasonable dimension optimization.