High-Precision image stitching Based on Computer Vision

Abstract: Image Stitching is an important part in the process of measuring large-size images. The accuracy of stitching directly affects the accuracy of the measurement results. This paper introduces the basic components and principles of the visual detection system, and focuses on the high-precision and large-scale image stitching Based on Computer Vision. Key words: Computer Vision; image stitching; image acquisition; Detection System machine vision technology is the most promising new technology in the field of precision testing technology, it uses electronics, optical detection, image processing, and computer technology to introduce machine vision into industrial inspection.

Abstract:Image Stitching is a very important part in the process of measuring large-size images. Its Splicing accuracy directly affects the accuracy of the measurement results. This paper introduces the basic components and principles of the visual detection system, and focuses on the high-precision and large-scale image stitching Based on Computer Vision.
Keywords:Computer Vision; image stitching; image acquisition; Detection System

Machine Vision is the most promising new technology in the field of precision testing technology.ElectronicsTechnology, optical detection, image processing, and computer technology, machine vision is introduced into industrial inspection to achieve rapid measurement of the three-dimensional size or position of an object, it has outstanding advantages such as non-contact, fast speed, and good flexibility, and has an important application prospect in modern manufacturing. With the increasing requirements on measurement accuracy, the positioning method based on mechanical motion accuracy cannot meet the requirements, while the high-precision Image Stitching Technology no longer relies on mechanical motion accuracy, instead, we use features of the overlapping parts of adjacent images to determine their relative positions. This method removes the restriction of mechanical precision and greatly improves the visual measurement accuracy [1]. At the same time, with the improvement of precision and width requirements in image measurement, the same image cannot meet both requirements.SensorIn relative motion mode, multiple images are taken as the basis for measurement, stitching, and image stitching. All merged images are mapped to the same image coordinate system and displayed in the same image space.
1. Composition of Computer Vision Detection System
The computer vision detection system consists of the following parts by function: (1) Image Information AcquisitionModule(2) image information processing module; (3) system control module; (4) X-Y moving object platform module.
Figure 1 shows the structure of the computer vision system. The working process of the system: the system control module to the X-Y platform control part of the single-chip microcomputer a command, requires the X-Y platform to move to the corresponding position, after arriving at the location, feedback a signal to the system control part, after receiving the information, the system control module communicates with the image acquisition module and processing module to complete system operations. System control module and single-chip microcomputer passRS232The system control module communicates with the image acquisition module and receives image information. The computer must be equipped with a dedicated image acquisition card and I/O module. In this system, the image acquisition card uses cognex MVS-8600, image acquisition card and the camera through the dedicated15The core cable is connected, and the I/O module is based onPCIThe parallel input and output ports of the bus.

2. Working Principle of the detection system
The workflow of the computer vision detection system consists of three parts: Image Information Acquisition, image information processing, and mechanical and electrical system execution detection results. In addition, parameters can be set and adjusted in real time through the man-machine interface according to system requirements. The working principle of the detection system is shown in figure 2. When the system is running normally, when the detected sample is installed on the detection platform with a clip, the operator manually enters the command to start detection to the system, the system will first perform pre-calculation based on the settings of the image parameters to be collected, determine the location parameters that the platform needs to locate During the image collection process, and then automatically move to the first image collection location, the system sends a position signal to the image acquisition module. The image acquisition system receives the signal and collects the images in the current field of view. After the acquisition is complete, the system sends a complete image acquisition signal to the Controller, the Platform moves to the next collection location again. Repeat the above work until all the images are collected, the system enters the image processing module for processing, and displays or generates alarms based on the detection results. During the entire operation, the image acquisition module and image processing module also need to continuously perform real-time image monitoring. This operation is temporarily suspended only when the image is collected. Each time an image is collected, it returns to the real-time monitoring interface of the image.