Theory knowledge of Shadow area detection

A. Shadows

It is because light from the light source points to the background is blocked by the object, but the light intensity in the scene does not change the surface texture feature structure of the background, because the intensity of incident light obtained in the shaded area is weakened, Therefore, the pixel value of the shaded area is smaller than the pixel value of the area without shadows (prior knowledge in the shadow detection algorithm). The shadow decreases the pixel value of the background, but the brightness does not change, and the gray variance in the shadow area of remote sensing image is generally small with other non-shadow regions, and the gray values between different regions have strong consistency. The gray mean and variance of the shaded area are not very obvious because of the surface properties projected by the shadow. In the remote sensing image, the target of higher ground is strong, the reflection is usually strong, occupy the higher gray level, and the shadow area is lower than the direct sunlight, the light is darker and the gray level is low, so the contrast of the image is the highest in the area of the target shadow's boundary line. The shadow edge to retrieve the grayscale value of the shaded area in the image with the shadow Edge .

Two The characteristics of the shadow in the color space.

1.RGB Color Space

rgb Span style= "font-family: ' The song Body '; > The color space uses the most frequently. rgb color space, also known as rgb mode, based on the three-color principle, is superimposed by the three primary colors (red, green, blue) in different proportions, forming rgb any color in space f rgb color space, specific 2.2

650) this.width=650; "Src=" http://s5.51cto.com/wyfs02/M00/7C/BC/wKiom1bWz6uiMaMoAABj Y9xsm_w440.jpg "title=" image 1.jpg "alt=" Wkiom1bwz6uimamoaabjy9xsm_w440.jpg "/>

Figure RGB color model

RGB space is sensitive to brightness, the three primary colors will change with the brightness changes, and the RGB color display is the overlay process, so the RGB Space is mainly used to describe the various types of electronic equipment screen can show the color, not suitable for the production of any color print or image color processing.

2.HSI Color Space

Hsicolor space as a main color-oriented model, with HueH, saturationSand StrengthIthree types of basic characteristics of the different colors, and people's visual system to perceive the color pattern is more similar, see Figure2.3. Where the angular size of the disc represents the HueH, which are related to the wavelength of light, to distinguish different color categories; The horizontal distance between the center and the cone surface indicates saturation .S, which is related to the purity of color, describes the proportion of white light added to a solid color, so that the same class of colors show different grades; the length of the hypotenuse of the cone indicates the strengthI, which is related to the intensity of light and reflects the brightness and grayscale of the image. HSIthe establishment of the color model is based on two conditions: (1) The intensity component is independent of the color information of the image;2the hue component and saturation component are strongly related to how the visual system senses color. These two factors makeHSIcolor model is very advantageous to the processing of image color, as well as the detection and research of color characteristics.

The link between HSI and RGB color space can be attributed to the following:the HSI space separates the luminance components in the RGB space first, The chroma component is then parsed into hue and saturation, and the angular vector is used to refer to the tonal component. The RGB color space can be transformed into the HSI space by geometric deduction , and the concrete process is shown in the formula (2-3).

650) this.width=650; "Src=" http://s5.51cto.com/wyfs02/M01/7C/BA/wKioL1bW0DTAN4gfAAA8_pxFHSM193 . jpg "title=" image 2.jpg "alt=" Wkiol1bw0dtan4gfaaa8_pxfhsm193.jpg "/>

HSI color Model

3.C1C2C3 Color Space

C1C2C3 color space as a color invariants, by gevers and smeulders[37] based on RGB Color space is first proposed, the space can be converted directly from the RGB Space, the conversion formula is:

650) this.width=650; "src=" Http://s2.51cto.com/wyfs02/M00/7C/BA/wKioL1bW0D_DOyxLAABT_1TO4Z0806.jpg "title=" Picture 3.jpg "alt=" Wkiol1bw0d_doyxlaabt_1to4z0806.jpg "/>

(2-12)

Wherein , r,g,b each pixel points in the image of the red, green, blue channel luminance value,c1,c2,c3 channel value by R,G ,theB -component value is transformed by the inverse tangent function, and these three channels also belong to the typical color invariants.

The C1C2C3 color space is A proportional model between the RGB components of a pixel, which is characterized by the three color components of the space that follow the object's surface reflectivity and sensor changes. This color change is the best non-linear change for shadow detection because the c1c2c3 color space is obtained through nonlinear changes and is only sensitive to reflected light .

Theory knowledge of Shadow area detection