The goal of the thesis is to make an image of an entire row of fruits (in this case, tomatoes) in a greenhouse farm, which will be captured using a 360-degree camera. The picture produced will be utilized for various reasons, such as counting and monitoring. To begin, this thesis will review the basics of computer vision and introduce essential issues. The features of the 360-degree video, as well as their technological specs, will be discussed next. The OpenCV library may now be used to evaluate the data collected in the greenhouse in the form of 360-degree video. It is possible to begin video processing and picture stitching to get the desired outcome with all of that knowledge. However, the fisheye lens causes significant distortion, necessitating extra procedures to undistort the image using methods such as cube mapping. There are also flaws in determining the video's speed, which will result in an undesired outcome. This issue may be solved by using dynamic stitching, which calculates the movement speed in real-time. All of the above techniques have resulted in a few different algorithm implementations. An assessment utilizing a generated video with all the controlled parameters is used to quantify the mistakes caused by several variations of the algorithm in order to choose the optimum technique. The pixel differences technique delivers the best result with a decent speed after a lengthy testing period. Furthermore, future enhancements for best practices in picture capture and processing for this project will be offered.
Anotace v angličtině
The goal of the thesis is to make an image of an entire row of fruits (in this case, tomatoes) in a greenhouse farm, which will be captured using a 360-degree camera. The picture produced will be utilized for various reasons, such as counting and monitoring. To begin, this thesis will review the basics of computer vision and introduce essential issues. The features of the 360-degree video, as well as their technological specs, will be discussed next. The OpenCV library may now be used to evaluate the data collected in the greenhouse in the form of 360-degree video. It is possible to begin video processing and picture stitching to get the desired outcome with all of that knowledge. However, the fisheye lens causes significant distortion, necessitating extra procedures to undistort the image using methods such as cube mapping. There are also flaws in determining the video's speed, which will result in an undesired outcome. This issue may be solved by using dynamic stitching, which calculates the movement speed in real-time. All of the above techniques have resulted in a few different algorithm implementations. An assessment utilizing a generated video with all the controlled parameters is used to quantify the mistakes caused by several variations of the algorithm in order to choose the optimum technique. The pixel differences technique delivers the best result with a decent speed after a lengthy testing period. Furthermore, future enhancements for best practices in picture capture and processing for this project will be offered.
The goal of the thesis is to make an image of an entire row of fruits (in this case, tomatoes) in a greenhouse farm, which will be captured using a 360-degree camera. The picture produced will be utilized for various reasons, such as counting and monitoring. To begin, this thesis will review the basics of computer vision and introduce essential issues. The features of the 360-degree video, as well as their technological specs, will be discussed next. The OpenCV library may now be used to evaluate the data collected in the greenhouse in the form of 360-degree video. It is possible to begin video processing and picture stitching to get the desired outcome with all of that knowledge. However, the fisheye lens causes significant distortion, necessitating extra procedures to undistort the image using methods such as cube mapping. There are also flaws in determining the video's speed, which will result in an undesired outcome. This issue may be solved by using dynamic stitching, which calculates the movement speed in real-time. All of the above techniques have resulted in a few different algorithm implementations. An assessment utilizing a generated video with all the controlled parameters is used to quantify the mistakes caused by several variations of the algorithm in order to choose the optimum technique. The pixel differences technique delivers the best result with a decent speed after a lengthy testing period. Furthermore, future enhancements for best practices in picture capture and processing for this project will be offered.
Anotace v angličtině
The goal of the thesis is to make an image of an entire row of fruits (in this case, tomatoes) in a greenhouse farm, which will be captured using a 360-degree camera. The picture produced will be utilized for various reasons, such as counting and monitoring. To begin, this thesis will review the basics of computer vision and introduce essential issues. The features of the 360-degree video, as well as their technological specs, will be discussed next. The OpenCV library may now be used to evaluate the data collected in the greenhouse in the form of 360-degree video. It is possible to begin video processing and picture stitching to get the desired outcome with all of that knowledge. However, the fisheye lens causes significant distortion, necessitating extra procedures to undistort the image using methods such as cube mapping. There are also flaws in determining the video's speed, which will result in an undesired outcome. This issue may be solved by using dynamic stitching, which calculates the movement speed in real-time. All of the above techniques have resulted in a few different algorithm implementations. An assessment utilizing a generated video with all the controlled parameters is used to quantify the mistakes caused by several variations of the algorithm in order to choose the optimum technique. The pixel differences technique delivers the best result with a decent speed after a lengthy testing period. Furthermore, future enhancements for best practices in picture capture and processing for this project will be offered.
Get acquainted with the essential topics of image processing in computer vision and summarize its basic information.
Focus on 360 degrees cameras with an emphasis on their technical specifications.
Analyze the data obtained in the real environment of a specific hydroponic greenhouse in the form of video recordings from a 360 degrees camera.
Based on the previous analysis, propose a suitable algorithm for preparing images for use in subsequent machine vision algorithms, most essentially CNN.
Code several versions of the proposed algorithm and try their suitability on the real data.
Evaluate the obtained results, identify problems, and suggest best practices for image capturing, outline directions for future image processing developments in this area.
Zásady pro vypracování
Get acquainted with the essential topics of image processing in computer vision and summarize its basic information.
Focus on 360 degrees cameras with an emphasis on their technical specifications.
Analyze the data obtained in the real environment of a specific hydroponic greenhouse in the form of video recordings from a 360 degrees camera.
Based on the previous analysis, propose a suitable algorithm for preparing images for use in subsequent machine vision algorithms, most essentially CNN.
Code several versions of the proposed algorithm and try their suitability on the real data.
Evaluate the obtained results, identify problems, and suggest best practices for image capturing, outline directions for future image processing developments in this area.
Seznam doporučené literatury
Forsyth, David, and Jean Ponce. Computer Vision: A Modern Approach. 2nd ed, Pearson, 2012.
Cameron, Joshua, et al. 360 Essentials: A Beginner’s Guide to Immersive Video Storytelling. Ryerson University Library, https://pressbooks.library.ryerson.ca/360essentials/. Accessed 7 Oct. 2021.
Greene, Ned. "Environment Mapping and Other Applications of World Projections.” IEEE Computer Graphics and Applications, vol. 6, no. 11, 1986, pp. 21–29, doi:10.1109/MCG.1986.276658.
Kweon, Gyeong-Il, and Choi, Young-Ho. "Fisheye Lens for Image Processing Applications.” Journal of the Optical Society of Korea, vol. 12, no. 2, June 2008, pp. 79–87, doi:10.3807/JOSK.2008.12.2.079.
Kopf, Johannes, et al. "Locally Adapted Projections to Reduce Panorama Distortions.” Computer Graphics Forum, vol. 28, no. 4, 2009, pp. 1083–89, doi:10.1111/j.1467-8659.2009.01485.x.
Seznam doporučené literatury
Forsyth, David, and Jean Ponce. Computer Vision: A Modern Approach. 2nd ed, Pearson, 2012.
Cameron, Joshua, et al. 360 Essentials: A Beginner’s Guide to Immersive Video Storytelling. Ryerson University Library, https://pressbooks.library.ryerson.ca/360essentials/. Accessed 7 Oct. 2021.
Greene, Ned. "Environment Mapping and Other Applications of World Projections.” IEEE Computer Graphics and Applications, vol. 6, no. 11, 1986, pp. 21–29, doi:10.1109/MCG.1986.276658.
Kweon, Gyeong-Il, and Choi, Young-Ho. "Fisheye Lens for Image Processing Applications.” Journal of the Optical Society of Korea, vol. 12, no. 2, June 2008, pp. 79–87, doi:10.3807/JOSK.2008.12.2.079.
Kopf, Johannes, et al. "Locally Adapted Projections to Reduce Panorama Distortions.” Computer Graphics Forum, vol. 28, no. 4, 2009, pp. 1083–89, doi:10.1111/j.1467-8659.2009.01485.x.
Přílohy volně vložené
-
Přílohy vázané v práci
ilustrace, tabulky
Převzato z knihovny
Ne
Plný text práce
Přílohy
Posudek(y) oponenta
Hodnocení vedoucího
Záznam průběhu obhajoby
Student presented the results of his master's thesis. The student covered all the main goals of his work and presented a comprehensive application of image processing application. Thesis supervisor's and oponent's reviews were read with questions from the thesis oponent. The student answered all of the presented questions. State exam comitee brought up following questions:
1) doc. Šenkeřík: Do you have the source code uploaded onto the github with a link in the thesis
2) dr. Viktorin: Can you see a situation in which there is a rock on the metal rails in the greenhouse with controlled environment?
The student answered both questions and satisfied the comitee's requirements.
The main issue raised by the comitee was that the source codes are not attached with the work.