Artificial intelligence first week homework
1. What are the components of the vision system? Take the robotic arm vision control system as an example to illustrate the components of the vision system.
Answer: The constituent elements are lighting equipment: light source imaging equipment: camera processing equipment: host algorithm software: vision processing system.
Take the robotic arm vision control system as an example. Light source: indoor light or special lighting. Camera: placed on the front of the robotic arm, single camera
Host: Desktop computer or embedded processor Algorithm software: Visual recognition algorithms and programs written in C++ or other languages
2. Try to explain this understanding of the content organization of the course from the perspective of imitating human visual processing flow. Further search on the Internet and find 1-2 books that you think are the most worthy of reference in the learning process (not too thick)
Answer: The human brain processes visual progress in hierarchical ways, and the courses are also carried out with similar hierarchical content. First, learn all kinds of knowledge in image acquisition, luminous flux, irradiance and other knowledge, and then analyze the image after acquiring the image. The preprocessing knowledge is in line with the order of the images acquired by our eyes. Human vision and machine vision are compared with each other to facilitate our understanding of machine vision related knowledge
The image acquisition in the curriculum system simulates the perception of human eyes on the picture; image preprocessing simulates the acquisition and processing of the picture by the retina; feature extraction, edge detection, and image segmentation simulate the human perception of objects in the picture; position estimation, motion estimation, camera Calibration simulates human's understanding of the behavior of objects in the picture; three-dimensional vision simulates human's understanding of the internal connections of the entire picture.
3. What is luminous flux and irradiance? Explain the luminous flux of several common light sources and the irradiance of several common lighting environments.
Answer: Luminous flux refers to the radiant power that the human eye can feel. It is equal to the product of the radiant energy of a certain band per unit time and the relative viewing rate of this band.
Sun: 3.566×1028 lm Candlelight: 12.56 lm Incandescent lamp/Tungsten halogen lamp: 12～24 lm/W Fluorescent lamp and gas discharge lamp: 50～120 lm/W LED lamp: 110 lm/W
Irradiance refers to the radiant flux density projected onto a flat surface. Refers to the radiant energy per unit time and per unit area upon reaching a surface.
Dark night: 0.001-0.02 lux, cloudy indoor: 5-50 lux, sunny indoor: 100-1000 lux, sunny day, direct sunlight: 100000 lux
Suitable for reading: 300～750 lux Standard illumination of home camera: 1400 lux
4. Combined with the color space diagram, briefly describe the physical meaning of each channel in the HSI color space, and illustrate with examples of images.
Answer: Hue H: describes the properties of a pure color
Saturation S: A measure of the degree to which a pure color is diluted by white light
Brightness embodies the concept of colorless light intensity.
In the HSV color model, the hue H uses an angle measurement ranging from 0° to 360°, the saturation S ranges from 0% to 100%, and the lightness V ranges from 0% (black) to 100% (white).
In the double hexagonal pyramid of the HSI color model, the angular range of hue H is [0, 2π], and the H parameter represents color information, that is, the position of the spectral color.
Saturation S is the distance from any point in the color space to the I axis, ranging from 0 to 1. It is expressed as the ratio between the purity of the selected color and the maximum purity of the color. V represents the brightness of the color, ranging from 0 to 1.
5. Explain the basic principles of color image sensor and gamma correction.
Bayer sensor: Use Bayesian filter to make different pixels sensitive to only one of the three primary colors of red, blue, and green. These pixels are intertwined, and then demosaicing interpolation is used to restore the original image.
The purpose of gamma coding for images is to compensate for the characteristics of human vision, so as to maximize the use of data bits or bandwidth representing black and white according to human perception of light or black and white. The linear relationship correction is the same as the image that people feel. There are two steps in the correction process: 1) The first time the image is collected by the sensor and then saved on the disk through γ correction. The common value is 1/2.2, and the input brightness is basically quantified in a linear manner.
2) From image storage to display, it undergoes the second γ correction, in order to make the image conform to human characteristics