Abstract
The low-altitude photogrammetry technology of unmanned aerial vehicles (UAVs) is widely used in many fields, but the absence of analysis and research affects the accuracy of its data products. At the same time, low-altitude photogrammetry faces the problem of low elevation positioning accuracy. The network space triangulation adjustment in the beam technique region is considered to eliminate perspective distortion in non-overlapping areas. This paper explains the key technologies of low-altitude photography and remote sensing mapping of UAVs, rectifies the distortion difference of remote sensing images, and then carries out grid division on the image according to the improved APAP (as-projective-as-possible warp) matching method. Next, it solves each grid homography matrix, linearizes the homography matrix, and carries out image matching according to the linearized homography matrix, which can effectively weaken the ghosting phenomenon during image matching. The network space triangulation adjustment in the beam technique region is considered to eliminate perspective distortion in non-overlapping areas. The two measurement areas’ accuracy level is analyzed using digital line drawing and digital orthophoto images (DOIs). Finally, the experimental results indicate that the image matching algorithm proposed in this paper has strong reliability and can substantially increase photogrammetric elevation positioning.
1. Introduction
With the rapid development of the economy, information and digital construction in different fields are also developing at high speed. Therefore, it is necessary to build relevant geographic databases that can be updated in real time, dynamically monitor the changes in land use, and simultaneously create different types of geographic information thematic maps at the same time. Geospatial information system (GIS) is a field that substantially benefits from deploying professional GIS mapping drones because of its potential to collect information that can be utilized for geospatial data visualization. The unmanned aerial vehicles (UAVs) offer features not available in other types of remote sensing and have seen a widespread attention for a wide range of application around the globe [1].
Remote sensing technology has strong low-altitude flight ability, allowing it to efficiently avoid the interference caused by clouds and weather and compensate for flaws caused by the interference of ordinary aerospace photography. It is an essential technical means in remote sensing technology. UAV remote sensing has several benefits over traditional measuring methods and aerial remote sensing, including high timeliness and high resolution, low cost, and low risk. It can build high-resolution images in huge areas, areas where conventional photography is challenging, and areas hit by natural disasters and produce large-scale topographic maps. In UAVs, DEM (digital elevation model) and other digital products are faster and more efficient [2]. UAVs are extensively useful for many industries, such as large-scale mapping and updating, land use dynamic monitoring, and so on, in order to get some geographic information on spatial elements such as land resources and the environment more quickly. At the same time, UAVs play an important role in surveying and mapping small and medium-sized areas, as well as tough terrain. Therefore, this paper explores the remote sensing mapping accuracy of UAV, i.e., low-altitude photogrammetry [3].
Low-altitude unmanned aerial vehicles (UAVs) are used for a mapping system to acquire both the wide-area coverage of remote sensors and the high levels of detail and accuracy of ground surveying. Unlike high-altitude systems in satellites or aircraft, all measuring equipment is located under the UAV, which looks like a helicopter and collects comprehensive data from low altitudes. Although the survey is carried out from the air, the resolution and accuracy are comparable to ground surveying. In addition, the UAV can simply and safely collect data.
The innovations of this paper are as follows:(1)The essential technologies of UAV low-altitude photography remote sensing mapping are discussed. The image is meshed using the improved APAP matching method. The network space triangulation adjustment of the beam method area is analyzed.(2)The accuracy level of the two measurement areas is analyzed using digital line drawing and digital orthophoto image. Finally, the experimental comparison shows that the proposed influence matching algorithm has high reliability and can effectively increase the accuracy of photogrammetric elevation positioning.
The rest of the paper is organized as follows. Section 2 explains the related works of different researchers who have proposed different models of UAVs based on low-altitude photography. Section 3 explains the different applications of UAV low-altitude photography. Section 4 explains the accuracy analysis of remote sensing mapping using UAVs at low altitudes. Section 5 explains the experimental result of two different surveys. Finally, Section 6 draws the conclusion of the paper.
2. Related Work
In the aspects of UAV remote sensing image matching and stitching, image matching is the key to realizing stereo measurement. The influence of matching accuracy, reliability, and algorithm adaptability is important to research content. The authors in [4] believe that deep learning is the most effective method for properly interpreting remote sensing images. However, the scale of remote sensing image is relatively large. Large images must be cut into small-scale image for training and prediction, and the use of different prediction strategies will impact the effect of prediction. This research primarily discusses several prediction techniques for semantic segmentation of remote sensing images. It also evaluates the prediction strategies using the training convolution network and DenseASPP model on international photogrammetry and remote sensing datasets. The results show that using prediction enhancement technology can effectively improve the segmentation accuracy by 2% compared with non-prediction enhancement, but it will increase the energy consumption. The authors in [5] adopted aerial photography technology to conduct topographic mapping of the coal mine area in view of the outdated mine topographic map in the coal mine area of a province, which cannot fulfill the application demands of mine ecological restoration and management as well as geological prevention and control. The surveying and mapping results reveal that aerial photography surveying technology has a good application impact in mine map surveying and mapping, which not only minimizes the workload of fieldwork but also reduces the cost of surveying and mapping and achieves reasonably high map accuracy. The use of aerial photography data in subsequent topographic map development is relatively convenient, improving topographic map production efficiency. However, there is an issue of poor comprehensive utilization of topographic maps. The authors in [6] used a photogrammetric image matching method based on terrain simulation to address the poor matching reliability of position-based image matching in photogrammetry. The mapping area simulated by the program is utilized to collect sample points on the ground. The imaging model clarifies the image matching relationship by collecting sample points and relevant data, and the full-automatic triangulation supported by image matching is measured based on the accurate matching relationship. The results of the experiments show that this method can successfully tackle the problem of unreliable image matching and enhance image accuracy, but the process is far too complicated. The authors in [7] proposed that with the advancement and development of information mapping technology and big data application, UAV tilt photogrammetry technology can be effectively applied to topographic map mapping, allowing the flexibility and convenience of UAV tilt photogrammetry as well as simplifying the data processing process. The creation of the three-dimensional virtual model using tilt photography can effectively make up for the problems related to photogrammetry occlusion so that relevant technicians can collect topographic elements through naked eyes indoors and only occasionally supplement the prediction of adjustment and mapping, thus effectively reducing the field workload and improving the efficiency of surveying and mapping work. However, due to the complex process, this method leads to high production costs.
3. Preprocessing of UAV Low-Altitude Photography and Remote Sensing Mapping
Different errors can affect the process of low-altitude photography using UAVs and remote sensing to improve mapping accuracy. The most common is tilt angle or lens distortion. There are different methods to remove distortion, which are discussed in detail.
3.1. Distortion Correction
Lens distortion is a systematic error when the focal length of the lens is fixed. Many UAVs used for low-altitude photogrammetry are equipped with non-measured digital camera azimuth elements, which can cause errors in the process of installation and debugging, making it easy to distort the lens, resulting in optical distortion errors in low-altitude photography and remote sensing mappings, such as radial distortion and eccentric distortion [8, 9]. There are two ways to remove the distortion difference in a remote sensing image: the direct method and the indirect method, which are shown in Figures 1 and 2.
The direct method shown in Figure 1 starts with the original remote sensing image, uses camera inspection to proofread the factors affecting the remote sensing image, and corrects the corresponding image point coordinates on the corrected remote sensing image using the camera distortion correction model to avoid the gray value of a remote sensing image from changing. As illustrated in Figure 2, the indirect technique corrects the remote sensing mapping by calculating the image point coordinates of the relevant points on the remote sensing mapping after correction and combining them with the gray interpolation method [7, 8].
3.2. Wallis Transformation
Wallis filtering can enhance the local contrast in remote sensing images, increase the image contrast, and minimize the noise in the image. After correcting the distortion difference, the remote sensing image can be transformed into a gray image, which is improved by Wallis’s filtering transformation. Furthermore, Wallis filtering can enhance the reliability and accuracy of remote sensing mapping matching results by enhancing the details of scale texture in remote sensing images and improving the feature points and accuracy extracted by the Harris operator. The Harris corner detector is a corner detection operator extensively used in computer vision algorithms to extract corners and infer image features.
Wallis filter transformation is achieved by mapping the gray value and variance value from a remote sensing picture to a given image [10, 11]. Wallis filter plays an essential role in low-contrast remote sensing images and the uneven effect of variance. Because the filter will effectively introduce the smoothing operator during the calculation of the local gray value and variance of the remote sensing image, it can also suppress the noise in the image, enhance the fuzzy texture in the remote sensing image, and effectively improve the quality of the remote sensing image while improving the remote sensing image information. Formulas (1) and (2) represent the Wallis filter transformation:where represents the gray value of the original remote sensing image point , represents the gray value of the remote sensing image after the point is changed by Wallis filtering, , , parameter represents the multiplicative coefficient, parameter represents the additive coefficient, represents the gray value of the pixel neighborhood in the remote sensing image, represents the gray variance of a pixel neighborhood image in the remote sensing image, represents the target value of the mean value of the remote sensing image, represents the target value of the variance of the remote sensing image, represents the constant of the contrast expansion of the remote sensing image, most of which are , and represents the coefficient of the brightness of the remote sensing image, most of which are .
4. UAV Low-Altitude Photogrammetry Accuracy Analysis of Remote Sensing Mapping
The accuracy of low-altitude photography and remote sensing can be improved by an image matching algorithm based on improved APAP and triangulation in the light beam region.
4.1. Photographic Remote Sensing Image Matching Algorithm Based on Improved APAP
An improved APAP photogrammetric remote sensing mapping matching algorithm is suggested when combined with the above preprocessing of remote sensing images. The parameter model of the remote sensing image is utilized as a sample, and the first remote sensing image is estimated using a random selection of image sample points. The probability of selecting different matching points is approximately the same. When the distance between selected sample points is established, the matching stability of the initial remote sensing image is increased [12, 13]. The parameters of unreasonable remote sensing images are eliminated. During the initial image validation process, the number of matched points in remote sensing images is set according to the remote sensing image threshold . The initial model is eliminated if the applicability of to the matched point model in remote sensing images is less than the confidence interval . For image matching, the slope of the line connecting the matching points of two remote sensing images is used as a constraint. The conversion angle is derived from the parameter model when two remote sensing maps are matched.
The slope between matching points for remote sensing mapping is represented as
The following formula is used to calculate the slope of any matching point in the remote sensing image:where and represent the coordinates of the first pair of remote sensing mapping matching points. If is less than the threshold , then the correct matching point is indicated, and if is greater than the threshold , then it will indicate the wrong matching point [14, 15].
The local matching of remote sensing images is matching of image distortion. Ghosting or distortion is common in matching a remote sensing image with complex textures, but they can produce a smooth transition to the remote sensing image and effectively minimize the ghosting and distortion during linearization. The polynomial is linearized without affecting the matrix characteristic structure of the polynomial. If M image feature points are included in the P field, any point Q in the image presents linearization in the P field, which can be expressed aswhere point and point represent two-dimensional points of the plane, represents the weight factor, represents the Jacobian matrix, which conforms to the model of affine transformation, , and represent the offset. The linearization formula is expressed as
On the basis of publicity (6), the linear weighting formula is given aswhere represents the weight factor, and the matching point that corresponds to the center is given the same weight value. In order to solve this problem, the weight is determined by pseudo-centroid weighting, and the average value of remote sensing image matching points is taken as the centroid, which can be expressed as
Use and replace to obtain the weight formula, expressed as
The weighting of formula (10) is more robust and can effectively avoid the instability of value [16, 17].
4.2. Triangulation in Light Beam Region
UAV low-altitude photography processing relies heavily on aerial triangulation encryption. It primarily extracts the connection points and a small number of ground control points through influence matching. It also collects the coordinates of the exterior orientation elements and encryption points of each remote sensing picture and integrates the remote sensing image of the whole measurement area into the unified coordinate system. The range of the light beam is between the infrared (with longer wavelengths) and ultraviolet (with shorter wavelengths). Visible light has wavelengths of 400–700 nanometers (nm), corresponding to frequencies of 750–420 terahertz (with shorter wavelengths).
The triangulated graphic in Figure 3 depicts the measuring area of the beam technique in the remote sensing image. The beam of the remote sensing image is used as the adjustment unit, and the collinear equation is utilized as the adjustment equation when changing the beam method area [18]. Figure 3 shows how different beams are used to rotate and translate in the measurement area so that the common points and lines in the remote sensing image meet, and the area is added to the coordinate system of the control points in the measurement area [19].
Some of the specific contents are as follows:(1)A remote sensing image determines the approximate value between the outer orientation element and the connection points coordinates.(2)The error equation is described by a collinear equation using control points and encrypted coordinates in various remote sensing images. The collinear equation is a mathematical model composed of the adjustment based on the beam method, which is expressed by where and represent the plane coordinates of the remote sensing image; , , and represent orientation elements of photogrammetric remote sensing image; , , and represent the spatial coordinates of photography; , , and represent the spatial coordinates of the object; point space, , and , represent the directional cosine formed by elements of outer orientation of remote sensing image. The left side of the remote sensing image is a non-linear function of unknown values [20]. It is necessary to linearize the collinear equation. Calculate the partial differential by using the elements of the outer orientation of the remote sensing image and the ground coordinates of the encrypted points [21]. As observation values, the coordinates of remote sensing images can be utilized. The error equations of all control points and encryption points are listed and expressed using the following formula: Write it as a matrix: Formula (13): The constitutive normal equation is expressed as(3)Remote sensing image points in areas of low-altitude photogrammetry are based on modified formula (15). In remote sensing images, the number of unknown coordinates of encrypted points in remote sensing images is usually larger than the number of exterior orientation elements. The cyclic partitioning method can solve the external orientation elements of remote sensing images.(4)Solve the geodetic coordinates of encrypted points by the spatial intersection of multiple images based on the orientation elements of the exterior orientation of each remote sensing image.
Field observation and detection can assess the measuring area network for adjustment accuracy. The coordinate solution values of checkpoint ground may be solved [22], and the coordinate value difference of real measurement can be estimated scientifically, based on the calculation of image external orientation elements and checkpoint coordinates [16]. The plane and elevation errors of the checkpoint are calculated by the following formula:where represents the error of the image checkpoint, represents the difference between the actual measured coordinates and the calculated values in the field of the image checkpoint, and represents the number of checkpoints. The above procedure investigates the accuracy of remote sensing mapping for low-altitude photogrammetry of unmanned aerial vehicles.
5. Experimental Result
An experimental study is conducted to collect two low-altitude remote sensing images of the experimental area to validate the reliability of the research on mapping accuracy of low-altitude remote sensing for unmanned aerial photography proposed in this paper. Table 1 shows the parameter of unmanned aerial photography. Zone 1 has six flights with 75 images, and zone 2 has four flights with 40 images.
5.1. Accuracy Analysis of Remote Sensing Mapping for Different Digital Products
PS RTK is used to evaluate the planar checkpoints of topographic maps in the survey area and the elevation checkpoints of specific coordinates. Plane checkpoints refer to relatively evident places such as corners on roads and houses, while elevation checkpoints refer to unique signs on roads and intersections inroads. The accuracy of topographic maps and digital orthophoto maps of the survey region may be assessed by measuring checkpoint coordinates. The measured ground objects and landmarks serve as plane and elevation checkpoints in the remote sensing picture. A detailed analysis of the mapping scales for various measurement regions is combined with the relevant requirements in the photogrammetric specifications. Table 2 and Figure 4 illustrate the outcomes of the analysis.
As you can see from Table 2 and Figure 4, the accuracy of digital line maps in unmanned aerial photography is easy to achieve in topographic maps at a scale of 1 : 2000. When all other photographic parameters remain unchanged, the photographic heights from 670 to 1050 meters can meet the requirements for the accuracy of digital line maps. There are 64 real survey elevation checkpoints in the field in surveyed area 1, which is taken as an example. Among the elevation checkpoints, 16 checkpoints with high accuracy are selected as fitting points to fit the elevation correction model. The remaining checkpoints are thoroughly checked both before and after the elevation correction. The accuracy of checkpoints before and after elevation fitting in different survey areas is shown in Table 3 and Figure 5.
The elevation accuracy has greatly increased after regional elevation fitting, as shown in Table 3 and Figure 5.
5.2. Digital Orthophoto Map
The coordinate values of the checkpoints are measured on the digital orthophoto map using the plane checkpoints, and the mean square error is determined to evaluate the accuracy of the digital orthophoto map. The remote sensing mapping accuracy analysis is carried out using the relevant adjustment in the digital orthophoto map, as shown in Table 4.
As demonstrated in Table 4, the accuracy of a digital orthophoto map can meet the requirements of mapping accuracy at various scales without considering the impact of topographic relief and ground feature projection distortion compared to the digital line drawing.
5.2.1. Comparative Analysis
The time comparison between the photographic remote sensing image matching method based on the enhanced APAP described in this paper and the traditional remote sensing image matching algorithm is shown in Figure 6.
From the beginning of image matching, the traditional remote sensing image matching time is slower than the method suggested in this paper, as shown in Figure 6. As the image matching logarithm increases, the matching speed becomes slower and slower. However, the proposed method matching time is sluggish as the image logarithm increases, but it is still faster than the traditional matching method. Figure 7 shows the reliability comparison between the photographic remote sensing image matching method based on improved APAP and the traditional image matching method. This shows that using the photographic remote sensing image matching method based on the enhanced APAP suggested in this paper can increase the accuracy of remote sensing mapping.
The traditional remote sensing image matching method, as shown in Figure 7, has poor reliability in image matching, with an accuracy of more than 80% occurring just once. At the beginning of image matching, the reliability of the photographic remote sensing image matching method based on enhanced APAP suggested in this research is relatively high. However, it has strong reliability, despite minor fluctuations. When used in conjunction with UAV low-altitude photogrammetry, it can significantly increase the accuracy of remote sensing mapping to improve the efficiency of digital products.
6. Conclusion
With the gradual acceleration of informatization and digitization in multiple industries, the demand for information and data has increased, especially in the surveying and mapping industry, which has growing demands for real-time and accurate high-resolution remote sensing pictures. Unmanned aerial vehicles (UAVs) have rapidly become the most common method of acquiring geographic information and spatial data because of their benefits, such as high timeliness and high resolution. It is widely used in many fields, including meteorological monitoring and geodesy. The experimental results show that steady imaging is important for a low-altitude UAV aerial photography system. Due to large rotation, the image quality will be reduced, and image distortion will be severe. It will also result in an unequal distribution of image ground resolution, irregular overlap between images, and stereo model inconsistencies. Therefore, it is important to analyze and study the accuracy of remote sensing mapping of UAV low-altitude photogrammetry. For particular observation objectives, a small UAV system with specified sensors will be installed in the future. This approach is anticipated to make it easier and safer to grasp geographic characteristics in-depth.
Data Availability
The data used to support the findings of this study are included within the article.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this article.