Principle of UAV camera

In the process of orthophoto or tilt modeling, some knowledge of photogrammetry and tilt modeling will be involved. By the way, I looked up some information and later found it necessary, so I listed it a little systematically. Finally, I provide several modeling routes for your reference, hoping to help you.

The ratio of the length of overlapping parts of adjacent photos on the same route to the side length of photos. In short, it is the overlap between photos on the same line.

The ratio of the length of overlapping parts of adjacent photos taken along two adjacent routes to the length of photos. In short, it is the overlap between lines of a photo.

There are certain requirements for the overlapping rate of photos when making orthophoto maps. Photos have at least 60% title and horizontal overlap rate, which can ensure that three photos have overlapping parts. This is to meet the needs of stereo measurement and mosaic of photos in areas with relatively flat terrain. When the terrain fluctuates greatly, the overlap rate should be increased when setting.

In the process of aerial photography, because the stability of unmanned aerial vehicle is not as good as that of manned aircraft, it is easily affected by high-altitude wind, which leads to the drift of the route, and the flight trajectory is no longer as straight as traditional aerial photography, which leads to the bending of the route. The so-called route bending is to splice the aerial photos of a route according to the ground image. The main points of each photo are not in a straight line, but appear as a zigzag line.

Route curvature: the percentage of the ratio of the maximum curvature vector of the route to the length of the route. Required route curvature

The included angle between the connecting line of adjacent main points and the connecting line of the border in the same direction in the photo. Require photo rotation angle

Knock on the blackboard! Knock on the blackboard! Knock on the blackboard!

To sum up:

1. It is suggested that the overlapping rate of photos should be above 60%, and the greater the topographic relief, the higher the overlapping rate.

2. Route curvature

3. The photo rotation angle is less than 6.

In addition, the requirements of automatic modeling software for building 3D models are higher than those of ordinary aerial photography, and the images need to have at least 70% of the course and lateral overlap.

This situation is a conventional aerial flight, and the flight can be completed by directly setting the route according to the required height and overlap.

Setting the overlap directly according to the requirements, the ground can meet the overlap requirements, but with the increase of floors, the image overlap of the roof will decrease, and insufficient roof overlap will lead to loopholes or obvious stretching of the 3D model, resulting in poor quality. Therefore, it is necessary to recalculate the overlap degree according to the highest building height in the mission area as the overlap degree set by flight aerial photography.

It is known that H is the height of aerial photography, H is the height of buildings, and α is the image angle.

The corresponding field length (width) of the photo: S= 2 H? tanα

W is the overlapping set of flight:

The overlapping parts at the top of the building are:

Therefore, in order to ensure the overlap of all features in the aerial photography area, the ground overlap w is set as:

It is known that the take-off point height is H, α is the image angle, the aerial flight height is H, and the ground overlap is W, so it is necessary to set the field height to H-H. ..

The overlap required for high-altitude takeoff is:

It is known that the takeoff point is lower than the ground height of the mission area by L, and α is the image angle. If the aerial flight height is required to be H and the ground overlap is W, the field height needs to be set to H+L..

Overlap is set to:

Analysis of practical problems:

Suppose the mission requires an aerial photography altitude of 200 meters, the ground overlap is 75%, and the take-off point is 50 meters below the ground in the mission area.

The calculation formula shows that:

Therefore, when taking off below the ground in the mission area, it is necessary to set the field height to 250 meters and the overlap degree to 80%, which can meet the requirements of aerial photography height of 200 meters and ground overlap degree of 75%.

Spatial resolution: spatial resolution is also called ground resolution, and the former is also called image resolution in terms of image resolution; The latter is the ground corresponding to the former. In short, spatial resolution is the size of the smallest feature that can be distinguished in detail in remote sensing images. The ground resolution of remote sensing images refers to the actual range corresponding to the size of each pixel on the ground, that is, the size of the ground is equivalent to the size of a pixel. Take TM image as an example, one pixel in the image represents 30 meters on the ground.

Image resolution = map distance/pixel

Scale = map distance/actual distance

Ground resolution = actual distance/pixel

Dots per inch (DPI)= pixels/map distance

Scale = 1: (ground resolution *(DPI/0.245))

The model accuracy of oblique photography is generally three times the resolution of photos and three times the ground resolution of orthophoto images generated by photos. If the resolution of the generated orthographic image is 3cm/ pixel, the model accuracy is basically 8- 15cm. Why not 9 cm accuracy? But a range, because the ground will fluctuate under any circumstances, and because of uncontrollable factors such as wind, can not guarantee that the resolution of the photo is fixed.

Formula: The accuracy of oblique photography model = three times the resolution of orthographic projection of the same project.

In fact, many people will basically convert after reading the above knowledge. This is just an example.

With the ratio of 1: 1000, the corresponding ground resolution means that 1cm on the map corresponds to 1000cm on the ground.

1 cm = 0.397008 inch

According to the 72dpi shot by DJI UAV, one inch contains 72 pixels, so 1 cm contains 0.3937008 * 72 = 28.44576 pixels.

The corresponding relationship is 28.3464576, corresponding to 1000cm on the map.

The resolution is: 1000/28.5776 = 35. 58666 . 68668686866

1: 1000 scale corresponds to a ground resolution of 35.2 cm, which is close to 0.36 m.

Then the accuracy of the aerial model is required to be 0.36m, and the corresponding aerial resolution is 0.12m. That is to say, the photos taken in aerial photography modeling should reach the accuracy of 12cm or more.

Here only provides reference for the route of oblique photography modeling.

Does the difference between good model quality and fine model necessarily lie in hardware? It doesn't exist. . . .

The S route mentioned here refers to the conventional five-route setting, and it is also the safest route for a single-lens UAV to collect tilt photography model data. They are an orthogonal projection route, one in each of the four directions of southeast and northwest. This method is more suitable for shooting large-area scenes.

Surrounding, as the name implies, is to shoot a circular flight around the area to be modeled, and let the camera aim at the main body of modeling to shoot. This route method is especially suitable for shooting a single building or landmark, with good three-dimensional reconstruction effect and few images needed. Take DJI as an example, if the area or building is not too big, a battery can meet it.

The finer the model, the higher the ground sampling density GSD. When the camera parameters are fixed, the lower the flying height, the higher the ground accuracy and the more detailed the results of model reconstruction. The higher the flying, the larger the collected area, the lower the flying, the higher the accuracy of the model and the better the modeling effect.

References:

1. Yu Guangrui, Wang Zhichao, Zhang Kunpeng, Sun Lijun. Research on the Application of Route Optimization Design of Surveying and Mapping UAV [J]. Beijing Surveying and Mapping, 20 15, (04):46-48+70.

2. Star City. Analysis of UAV image overlap based on simplified SIFT algorithm [J]. Journal of Harbin Engineering University, 2012,33 (02): 221-225.

3. Cui, Lin Zongjian, Research on 3D Modeling Method of Remote Sensing Images of Large Overlapping UAV [J]. Surveying and Mapping Science, 2005, (02):36-38+4.