A method and product for producing road orthophotos using automobile big data

Abstract

The present invention discloses a method and product for producing road orthophoto images using automobile big data, the method comprising: taking the longitude and latitude in the initial frame data of a single vehicle as the origin, constructing an initial frame rectangular coordinate system, dividing the initial frame rectangular coordinate system into four coordinate azimuth intervals of north, east, south and west; determining the driving direction of the road where the vehicle is located according to the position where the longitude and latitude in the subsequent frame data of the single vehicle falls into the coordinate azimuth interval; in a map section, fitting the longitude and latitude data of multiple vehicles on the road in the same driving direction in the automobile big data to synthesize an orthophoto digital road image. The present invention is based on the rear intersection positioning mechanism adopted by the navigation satellite, and the positioning coordinates fundamentally do not have the problem of projection misjudgment, so the obtained road image is the orthophoto image of the road.

Description

Technical Field

[0001] The present invention relates to the field of road imaging technology, and in particular to a method and product for producing road orthophotos using automobile big data. Background Art

[0002] From the physical perspective of observing the road surface, the surveying and mapping methods in the prior art include:

[0003] (1) Ground surveying: Mapping the road shape while standing on the ground, which is equivalent to a viewing angle of zero degrees. The camera on the surveying vehicle is affected by the height of the vehicle body, and the viewing angle is usually very small.

[0004] (2) Aerial photography by aircraft is limited by the flight altitude and the central projection imaging mechanism. Except for the central projection point which is orthographic, the deformation becomes greater towards the edge of the image.

[0005] (3) Remote sensing satellites: The imaging mechanism is basically the same as that of aerial photography. The undulations of the ground will affect the image. For example, when opening the remote sensing image of any navigation map, the tilt of high-rise buildings is clearly visible.

[0006] Aircraft aerial photography and remote sensing satellites are difficult to obtain road information of the national road network due to factors such as cloud cover / mountain / building cover, satellite round trip cycle, and imaging basis (inability to obtain orthophotos). Under the current technical background, it is almost impossible to use surveying vehicles to achieve monthly updates, daily updates (i.e., using surveying vehicles to run the roads once a month or every day) or even dynamic continuous updates of large-scale road surveying. This is especially true for a large number of terminal roads, non-public roads, and unconventional roads.

[0007] In short, due to the physical limitations of the above-mentioned road imaging technologies, coupled with the high implementation costs, there is currently no technology for large-scale production of full-coverage road orthophotos either domestically or internationally. Summary of the invention

[0008] In view of one or more of the problems in the prior art, the present invention provides a method for producing road orthophotos using automobile big data, comprising the following steps:

[0009] S1. Taking the latitude and longitude of the initial frame data of a single vehicle as the origin, constructing the initial frame rectangular coordinate system, dividing the initial frame rectangular coordinate system into four coordinate azimuth intervals of north, east, south and west;

[0010] S2. Determine the direction of travel of the vehicle on the road according to the position where the latitude and longitude of a single vehicle in subsequent frame data falls within the coordinate azimuth interval;

[0011] S3. In a map tile, the latitude and longitude data of multiple vehicles on the road in the same driving direction in the automobile big data are fitted to synthesize the orthophoto digital road image.

[0012] According to one aspect of the present invention, the step S3 further includes the following steps:

[0013] S31. Fitting the orthophoto digital road images under multiple map sections to obtain the orthophoto digital road images of the area to be measured.

[0014] According to one aspect of the present invention, the generation of the coordinate azimuth comprises the following steps:

[0015] SA1. In the above coordinate azimuth interval, the north direction is between 315° and 45°, the east direction is between 45° and 135°, the south direction is between 135° and 225°, and the west direction is between 225° and 315°.

[0016] According to one aspect of the present invention, the process of determining the map tiles comprises the following steps:

[0017] SB1. Multiply the lowest resolution of the decimal latitude and longitude coordinates uploaded by the vehicle by the correction factor and round it to the integer;

[0018] SB2. Name the rounded decimal longitude X, and the rounded decimal latitude Y;

[0019] SB3. Let X = 2^n × Hx + Lx, Y = 2^n × Hy + Ly, where n is a positive integer, Hx, Lx, Hy, Ly are all natural numbers, and Lx and Ly are both less than 2^n;

[0020] SB4. Name the map tile Ln_Hx_Hy.

[0021] Preferably, the correction coefficient is 1000000 or 921600.

[0022] Further preferably, n=16 in the map tiles.

[0023] According to one aspect of the present invention, the step S3 further includes a noise removal step:

[0024] SC1. Get the frequency of the first grid in the orthophoto digital road image and read the result into the two-dimensional array in memory;

[0025] SC2. Traverse each pixel in a map tile and make a sliding box with a side length of L centered on the pixel;

[0026] SC3. Traverse the frequency of each pixel in the sliding frame and form a one-dimensional array of length 1×L;

[0027] SC4. Sort the one-dimensional array from small to large, count the low frequency number n, and use the minimum frequency n0 corresponding to when n is greater than or equal to the pre-specified ratio as the basis for distinguishing the noise points of the sliding box;

[0028] SC5. Traverse each pixel in the sliding frame and delete the pixels with a frequency less than or equal to n0 as noise.

[0029] Preferably, 0<n0<20%.

[0030] Further preferably, n0=5%.

[0031] According to one aspect of the present invention, the step S3 further includes the following steps:

[0032] SD1. Selecting an optical remote sensing image located in the same map frame as the orthophoto digital road image, and scaling the optical remote sensing image to be completely consistent with the orthophoto digital road image;

[0033] SD2. Using the orthophoto digital road image as a template, crop and correct the optical remote sensing image to remove all non-road pixels to obtain an orthophoto optical road image.

[0034] Preferably, the step SD2 further includes the following steps:

[0035] SD3. Fit the orthophoto road images under multiple map sections to obtain the orthophoto road images of the area to be measured.

[0036] According to one aspect of the present invention, the present invention provides an image product, wherein the generation process of the image product adopts any of the above-mentioned methods.

[0037] Beneficial effects of the present invention:

[0038] (1) Compared with remote sensing mapping, the present invention is based on the rear intersection positioning mechanism adopted by navigation satellites. The positioning coordinates fundamentally do not have the problem of projection misjudgment, so the obtained road image is the orthophoto of the road. At the same time, the resolution of the present invention is 25 times that of satellite remote sensing mapping. The implementation of the present invention can equivalently obtain a directional remote sensing system covering the entire world and specifically targeting roads by using existing satellite positioning systems such as Beidou, GPS, GLONASS, and Galileo.

[0039] (2) my country has more than one million kilometers of unconventional roads (roads not shown in conventional navigation software, such as roads inside residential areas, sandy roads, roads in western / border areas, etc.). Existing navigation software can only reflect information on conventional roads (or public roads), and the present invention is very helpful for mapping unconventional roads. The present invention can be applied not only to vehicle roads, but also to the production of airways for aircraft and ships.

[0040] (3) Using traditional surveying and mapping technology to survey and map the roads across the country is a huge project with high investment and long cycle. Even using the latest professional navigation map surveying and mapping vehicles to survey and map the entire road network is still a huge project with high investment and long cycle. The present invention is the first to use the high-resolution road image information with an accuracy of up to 0.1 meters contained in the vehicle-mounted satellite positioning data of the super-large model for road surveying and imaging, with low investment and short cycle.

[0041] (4) The present invention correctly identifies and effectively deletes all original latitude and longitude coordinate data sets containing noise through appropriate means under the influence of the first law of geography, human travel habits and other internal mechanisms.

[0042] (5) The present invention provides a new technical path for high-precision mapping of roads over a large area, and provides a secondary utilization method for the widely existing automobile big data. It can improve the high-coverage mapping accuracy of the national road network to the greatest extent, and can maximize the update speed of the national road network map. The present invention is an innovative technology that can truly achieve full coverage imaging and mapping of the national road network. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0044] Figure 1 is a road surface image drawn according to an embodiment of the present invention;

[0045] Figure 2 is with Figure 1 Satellite remote sensing images of the same location;

[0046] Figure 3 is a ground image before denoising according to an embodiment of the present invention;

[0047] Figure 4 is a ground image after noise removal according to an embodiment of the present invention;

[0048] Figure 5 It is a schematic diagram of road slices drawn according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further described in detail below in conjunction with the embodiments. The following embodiments are provided to explain the present invention, but the present invention is not limited to the following embodiments.

[0050] Driven by relevant policies, satellite positioning mobile terminals (on-board terminals) are widely installed on cars. For example, the newly produced new energy vehicles and engineering vehicles in China are required to install on-board terminals and upload data. With the addition of tens of millions of vehicles, the collection accuracy has reached the second level, forming an increasingly large data source. At this stage, the application of data is mainly to monitor vehicle form parameters and report abnormal status of vehicles. But more importantly, each frame message of the on-board terminal contains satellite positioning data. At present, all domestic new energy vehicles implement GB / T 32960 "Technical Specifications for Remote Service and Management System of Electric Vehicles". During driving, the satellite positioning data of the vehicle is continuously reported at intervals of about 10 seconds. With the rapid popularization of passenger new energy vehicles, the roads covered almost include all public roads and non-public roads that can be passed.

[0051] The applicant discovered that industry-level satellite positioning data contains ultra-high-precision map road information.

[0052] At present, the minimum resolution of longitude and latitude coordinates uploaded by vehicles can reach 1 / 1000000 (longitude / latitude). The longitude and latitude generated continuously during the driving process of the vehicle are continuously reported in a standard format. If the entire earth's surface is extended into a rectangular plane and regularly divided into grids with a side length of 1 / 1000000 (longitude / latitude), each longitude and latitude reported by the on-board mobile terminal is regarded as an exposure of a grid. With the support of huge amounts of data, the ground space resolution of about 0.1m high-resolution road image (on the longitude, for every 1 degree difference in latitude, the actual distance is about 111 kilometers, so 1 / 1000000 latitude is about 0.1m; on the latitude, for every 1 degree difference in longitude, the actual distance is about 111 kilometers × cosθ, where θ represents the latitude of the latitude. Most of my country’s land is located between 20 and 50 degrees north latitude, cos20°≈0.94, cos50°≈0.64, and 1 / 1000000 longitude here is also about 0.1m). For satellite remote sensing mapping, the highest accuracy claimed in the world is currently 0.5m, and under the limitation of the central projection imaging mechanism, only the center of the image can achieve this accuracy. Calculated in terms of area, the present invention (0.1m×0.1m=0.01m 2 ) is the resolution of satellite remote sensing mapping (0.5m×0.5m=0.25m 2 ) is 25 times.

[0053] The positioning mechanism of the present invention is different from that of remote sensing images. The positioning of the present invention is directly based on the resection of satellites (including Beidou, GPS, GLONASS, and Galileo satellites), which can fundamentally avoid the problem of orthophoto errors and ultimately form an absolute orthophoto image of the road. The present invention combines the satellite navigation system with thousands of vehicles to form a directional scanning system specifically for roads, which can form an array image with strict geographic coordinates. To quickly understand the present invention, you can think of the surface of the earth as a piece of copy paper. When a car goes over one point at a time, road information will be formed on the copy paper. Compared with the overall shooting of optical remote sensing, the present invention only shoots the road, which is conducive to accurate identification of the road on the one hand, and on the other hand, it is conducive to removing noise images. Furthermore, the present invention can be applied not only to vehicle roads, but also to the production of airways for aircraft and ships.

[0054] In order to facilitate those skilled in the art to implement the present invention, the related technologies related to the present invention and the upstream and downstream of the present invention are now described. The following technologies are not necessarily interdependent, but can be superimposed and combined, or each can be implemented separately in combination with other existing technologies. Related technologies include Technology A: Driving direction division technology; Technology B: Spatial indexing and map segmentation technology; Technology C: Frequency-based image corrosion and denoising technology; Technology D: Road image and optical remote sensing image synthesis technology; Technology E: Road section segmentation technology based on driving direction; Technology F: Road section vehicle condition investigation technology; Technology G: Percentile data expression technology; Technology H: Overall accuracy control technology from horizontal to edge and vertical to bottom; Technology I: AI global optimization technology; Technology J: Classification optimization technology.

[0055] Technology A: Driving direction division technology

[0056] According to the relative position of the longitude and latitude coordinates of the front and rear frames of the same vehicle, the driving azimuth on each grid is calculated, and the roads are classified into four directions according to the driving azimuth. The specific method is to construct a rectangular coordinate system with the longitude and latitude of the vehicle's current frame as the origin. The vertical axis of the coordinate system coincides with the longitude, with north at the top and south at the bottom, and the horizontal axis coincides with the latitude, with east on the right and west on the left. The coordinate plane is divided into four directions according to the direction angle, [315,45), [45,135), [135,225), [225,315) or (315,45], (45,135], (135,225], (225,315], which represent the four directions of north, east, south and west respectively. According to the position of the next frame of the current frame of the vehicle falling into the plane, the driving direction of the current road grid of the vehicle is uniquely determined and recorded. For the longitude and latitude points that also report the GPS (GPS in a broad sense includes Beidou, GLONASS, and Galileo) azimuth, the driving direction of the current road grid of the vehicle can be uniquely determined according to the GPS azimuth and recorded.

[0057] The core function of the driving direction division technology is to add an indicator with clear classification significance to the indifferent longitude and latitude coordinate points. Driving direction is the most common feature in the vehicle-road community. It has nothing to do with the vehicle category, but is strongly related to the vehicle's working conditions. For example, on steep slopes, the speed and fuel consumption (up to dozens of times) of vehicles in the upward and downward directions are very different. By processing the longitude and latitude coordinate information collected from the vehicle through the driving direction division technology, it is possible to determine in the navigation system whether the vehicle is in an uphill or downhill state, and then provide predictive reminders for vehicle driving. Driving direction division technology is also an important support for further distinguishing lanes, because most vehicles will not go in the opposite direction, and the vehicle coordinate information can reflect the lane information in which it is located. This technology can also be linked to traffic management rules in subsequent applications.

[0058] Technology B: Spatial indexing and map tiling technology

[0059] Globality is one of the essential characteristics of maps. Mapping of specific roads in local areas is only meaningful in special application scenarios, such as unmanned mining trucks. In order to achieve the globality of maps, maps need to be indexed. Map framing is to make the size of topographic maps of various scales consistent, facilitate surveying and mapping, avoid re-surveying and missing surveys, and divide the survey area into several small blocks according to certain rules. This technology is a map framing technology solution for road orthophotos formed by combining map framing theory and computer data characteristics.

[0060] Based on the minimum resolution of the longitude and latitude coordinates uploaded by the vehicle, which can reach 1 / 1000000 (longitude / latitude), this technology first multiplies the longitude and latitude coordinate values ​​(decimal degrees) reported by the vehicle by 1000000 and then rounds them up (in specific implementation, rounding does not necessarily have to be multiplied by 1000000, for example, it can also be multiplied by 921600 and any other value that can round the longitude and latitude coordinates). For example, 115°30' east longitude is rounded to 115.500000×1000000=115500000. 115500000 is converted to binary as 00000110 11100010 01100011 11100000, which is stored in 4 bytes in total (1Byte=8bit). Furthermore, the 4 bytes of longitude are divided into a high double byte and a low double byte. The high double byte (00000110 11100010) is converted to decimal 1762, and the low double byte (01100011 11100000) is converted to decimal 25568. The decimal longitude is named X, the decimal value converted from the high double byte of longitude is named Hx, and the decimal value converted from the low double byte of longitude is named Lx. Then the following equation exists between X, Hx, and Lx:

[0061] X=65536×Hx+Lx(1)

[0062] Taking 115°30' east longitude as an example, 115500000=65536×1762+25568, where X=115500000, Hx=1762, Lx=25568.

[0063] Similarly, there is the following equation relationship between the decimal latitude Y and the decimal value Hy converted from the high double-byte latitude and the decimal value Ly converted from the low double-byte latitude:

[0064] Y=65536×Hy+Ly(2)

[0065] This technology uses Hx_Hy as the basic standard framing scheme for road images. The full name of the map is "Ln_Hx_Hy", where Ln represents the level of the framing, Hx represents the high-order word of longitude, and Hy represents the high-order word of latitude. The framing method in formulas (1) and (2) is named "L16_Hx_Hy", where L16 represents the width of the map, which is 2^16=65536 basic grids. In the L16 framing method, since 1 / 1000000 (longitude / latitude) is approximately 0.1m, each basic grid is approximately equal to 0.1m on the ground. The area of ​​each standard framing is approximately equivalent to 65536×0.1m×65536×0.1m=42949672.96m on the ground. 2 , which is about 43km 2 . Based on the total land area of ​​my country, which is about 9.6 million square kilometers, the total number of maps in China is about 230,000. In the specific implementation, n in Ln can be any positive integer. For example, L15 represents a map width of 2^15=32768 basic grids, which is equivalent to the enlarged map of L16; L17 represents a map width of 2^17=131072 basic grids, which is equivalent to the reduced map of L16.

[0066] Since computers need to convert decimal data into binary operations and finally convert it into decimal data for feedback to users, if decimal data can be associated with binary data, the computer processing power can be greatly improved. The Ln framing method uses 2^n as the framing basis, which can establish a strong association with computer binary, thereby achieving efficient storage and operation, making it easier for computers to process large amounts of data. Among them, when n=16, the Ln framing method can establish a strong association with the lowest resolution of longitude and latitude coordinates 1 / 1000000 (longitude / latitude), further improving the computer's ability to process coordinate data. This technical point splits the four-byte longitude and latitude values ​​into two double-byte words, high and low, and uses the high-order words of longitude and latitude as the framing basis, which fully utilizes the characteristics of computer binary data and is deeply compatible with the classic GIS four-fork spatial indexing technology.

[0067] The size of the points divided by this technology is very suitable for multi-threaded parallel processing. In practice, it can increase data processing efficiency by about 100 times. In terms of hardware, a medium-performance server CPU can be easily configured with more than 100 cores / threads. The well-known parallel computing technology used by relevant technical personnel can achieve a nearly proportional speed increase compared to a single process. In terms of software, taking the commonly used trajectory point road positioning algorithm as an example, under GIS technology, a complex nonlinear algorithm is used, while this technology converts it into a simple linear technology, and ultimately into the most efficient direct addressing for computers.

[0068] Technology C: Frequency-based image erosion denoising technology

[0069] Affected by factors such as road construction, vehicle parking, and occlusion by buildings or plants, the longitude and latitude data uploaded from the mobile terminal on the vehicle inevitably contain noise, and there are a large number of burrs on the edges around the main roads. Erosion operation is a basic operation in digital image processing, which can effectively remove edge burrs. The standard corrosion algorithm is only for binary black and white images, and only considers spatial relationships without considering frequency relationships. This technology uses the number of repeated occurrences of grid points as an important basis for determining whether it is noise, that is, in a specified spatial area, such as a standard partition defined by spatial indexing and map partitioning technology (technology B), the higher the frequency of the point, the more likely it is to correspond to the real road grid, and the lower the frequency of the point, the more likely it is to be noise.

[0070] When implementing this technology, you can first preprocess the specified map data to obtain the frequency of the first grid, and read the result into the two-dimensional array in the memory. Traverse each pixel, make a sliding frame with a side length of L with the pixel as the center, traverse the frequency of each pixel in the sliding frame, and form a one-dimensional array with a length of 1×L, calculate the total frequency N in the sliding frame, sort the one-dimensional array from small to large, and then sequentially accumulate the number of low frequencies n. When n is greater than or equal to the pre-specified ratio (such as 5%), the corresponding minimum frequency n0 (such as 5%) is used as the basis for distinguishing the noise points of the sliding frame, traverse each pixel in the sliding frame, and pixels with a frequency less than or equal to n0 (such as 5%) are regarded as noise. Figure 3 and Figure 4 This is a ground image obtained through GPS coordinates in automobile big data. The black pixels in the image are noise images. Figure 3 is the original image before denoising, and n0 is 0%. Figure 4 This is the denoised image with n0 set to 5%.

[0071] Compared with conventional noise removal technology, this technology focuses on detailed features such as the geometric relationship between adjacent longitude and latitude coordinates and the logical relationship between speed and displacement. The operation of this technology is extremely simple and efficient, and it is very suitable for large-scale data processing. The image corrosion denoising technology in the prior art uses the number of hit points or the proportion in the sliding window as the basis for noise discrimination. For the noise points in the low-frequency noise area of ​​the original image, because there are many hit points in the sliding window and the proportion is high, they will be mistakenly regarded as valid points and retained, thereby weakening the noise removal effect. The frequency-based image corrosion denoising technology retains the advantages of spatial operations and makes full use of the data characteristics of the satellite positioning terminal: more than 90% of the longitude and latitude are concentrated in the grid below 10%. This technology is equivalent to the 19 rule, that is, the concentration is higher than the 28 rule. In theory, this technology extends the corrosion algorithm from binary images to grayscale images, and the mechanism establishes a connection between low-frequency points and noise points. Relying on the existing domestic automobile big data resources, the inventor has basically completed the road imaging of the national road network.

[0072] Technology D: Road image and optical remote sensing image synthesis technology

[0073] When the user of this technology has available optical remote sensing image resources, the image cropping and correction algorithm provided by this technology can be used to achieve the synthesis of road images and optical remote sensing images. The preferred steps may include:

[0074] SD11. Based on the standard framing provided by the spatial index and map framing technology (Technique B), prepare the optical remote sensing image of the corresponding area and scale the optical remote sensing image to the same pixel number as the standard framing;

[0075] SD21. Use the road image denoised by frequency-based image erosion denoising technique (Technique C) as a template to crop and correct the optical remote sensing image to remove all non-road pixels.

[0076] Figure 1 Road images drawn using longitude and latitude coordinate information from automobile big data. Figure 1 The original image size is 2048*2048 pixels. Figure 1 Each pixel in represents a coordinate value actually uploaded by the vehicle terminal, that is, Figure 1 The distance between adjacent pixels in the . Figure 1 The image drawn reflects a part of a typical overpass, which is different from the actual road section ( Figure 2 ) and can be used for cropping and correction of optical remote sensing images.

[0077] This technology can realize the data fusion of longitude and latitude information of navigation satellites (Beidou, GPS, GLONASS, Galileo) and optical remote sensing satellites in the same WGS84 coordinate system. From the perspective of navigation satellites, surface object information can be added to each longitude and latitude point. From the perspective of optical remote sensing satellites, high-precision image registration and higher-resolution image interpolation can be achieved. In the process of remote sensing image processing with existing technologies, orthorectification and coordinate correction are required. This technology can provide refined reference for remote sensing image correction, making remote sensing image processing more accurate.

[0078] Furthermore, inspired by the synthesis of navigation coordinates and optical remote sensing images proposed by this technology, it is easy to think of the on-board forward-facing camera device that is standard on new energy vehicles. Through the foreground and background separation technology of conventional video processing (separating temporary vehicles, pedestrians, construction, etc. on the roadside), the real optical image map of the ground of the entire road network can be obtained.

[0079] Technology E: Road section segmentation technology based on driving direction

[0080] The basic idea of ​​road section segmentation is to continuously segment the road into multiple road sections at equal intervals along the direction of the road. Since the direction of the road can be any direction between 0 and 360 degrees in the rectangular plane coordinates, the road cross-section line usually displayed is not a horizontal line or a vertical line, which is not convenient for large-scale rapid calculation on the computer. The essence of road slicing is to regard spatially continuous road grid points in the same latitude band (north-south roads) or the same longitude band (east-west roads) as the same road section. Based on this technology, large-scale national road slicing can be implemented by user programming. This technology combines the actual characteristics of China's roads that are mainly north-south or east-west, and the basic algorithm flow on the computer includes:

[0081] SE1. Based on the driving direction division technology (technology A), the road direction is divided into four directions of east, south, west and north at 90 degrees, corresponding to the four azimuth intervals of [315,45), [45,135), [135,225), [225,315) or (315,45], (45,135], (135,225], (225,315) on the rectangular plane;

[0082] SE2. Determine the direction of the vehicle according to the coordinate data of at least two points continuously uploaded by the vehicle terminal, and determine the direction of the road according to the direction of the vehicle;

[0083] SE3. Using the longitude and latitude coordinate information in the automobile big data, count the longitude and latitude coordinates of multiple vehicles in the same driving direction in the map slice to be investigated;

[0084] SE4. For the image to be processed, open a two-dimensional matrix of the same size in the computer memory;

[0085] SE5. For roads in any direction of east, south, west or north, draw a horizontal line on the basic grid with a grid interval of 2^n;

[0086] SE6. Traverse all grids from left to right along the horizontal line. The first point encountered is regarded as the left boundary point of the section, until a blank grid is encountered as the right boundary point of the section.

[0087] like Figure 5 It reflects a northbound road slice with longitude starting from 104.893480 degrees and ending at 104.893501 degrees, and latitude being 31.261430 degrees (in order to facilitate the understanding of the present technology by those skilled in the art, the slice only traverses one row of grids, and in actual production, all grids are traversed with a grid interval of 2^n). In the figure, the length of the map sheet is 104893505-104893475=30 basic grids, and the width of the map sheet is 31400000-30000000=1400000 basic grids, each basic grid is equivalent to 1 / 1000000 (longitude / latitude), and the grid interval is 2^n. In actual operation, different grid interval values ​​can be selected depending on the accuracy of section segmentation. For example, when n=4, that is, a road slice is made every 16 basic grids, that is, when the bandwidth of the longitude or latitude band is 16 / 1000000 degrees, it is equivalent to the width of the road section slice being 1.6 meters. The national road network is estimated to be 5.4368 million kilometers. Through this technology, about 3.4 billion road sections will be equivalent.

[0088] Technology F: Road section vehicle operating condition survey technology

[0089] The various vehicle operating indicators in the vehicle message, such as vehicle speed, torque, and rotation speed, are regarded as a sampling survey of the section. In the manual section survey, because the section is fixed, only the actual measured value of the section can be observed, and the change cannot be observed.

[0090] Taking vehicle speed as an example, manual cross-section surveys can only observe average speed, but not acceleration. On a normal driving road, acceleration is strongly correlated with the slope of the road, and acceleration can invert the road slope. Using the characteristics of continuous sampling, automobile big data can successfully solve the cross-section acceleration observation problem.

[0091] Taking fuel flow (in liters per hour) as an example, it is impossible to obtain the fuel flow and change when the vehicle passes the current section through manual section survey. However, fuel flow and change are strongly correlated with energy conservation and emission reduction, and are the key data basis for PCC predictive energy-saving technology. Automotive big data can solve this problem.

[0092] Similarly, engine output torque and engine speed parameters are very important for vehicle operation, especially intelligent driving.

[0093] Technology G: Percentile data expression technology

[0094] Under big data conditions, especially for urban roads with high-density vehicles, the same road section will quickly accumulate thousands of message records. This technology uses cumulative quantiles to represent individual indicators on each section. Taking vehicle speed as an example, for example, all vehicle speed information obtained on the road section over a period of time can be summarized, and the driving speed in the 15% to 85% quantile speed range on the road section is used as the safe driving speed range, which can be used to directly guide vehicle safe driving. And the experience-level medium-speed driving can be directly quantified as the speed range of a specific road section. The so-called "experience level" refers to the common experience of countless drivers "medium-speed driving", and the specific number of kilometers per hour of medium speed can only be experience. The 15%-85% quantile speed range is a definition in traffic engineering. The basis is: vehicles traveling in this speed range have the lowest accident rate.

[0095] Technology H: Overall precision control technology from side to side and from top to bottom

[0096] If the road grid is divided into rectangular blocks, and the road sections within the rectangular blocks are segmented and the section conditions are counted, there will be serious edge effect problems, which will be shown as obvious deviations between the roads at the edge of the region and the actual situation when mapping. This technology adopts a horizontal to edge and vertical to bottom solution. For example, the national road network can be regarded as a whole, and the grid interval value of the road section segmentation in the road section segmentation technology based on the driving direction (Technology E) is used as the bandwidth. For north-south roads, for example, a latitude band with a bandwidth of 1.6 meters can be selected as a sub-whole to run through the country from west to east. For east-west roads, a longitude band with a bandwidth of 1.6 meters can be selected as a sub-whole to run through the country from south to north, fundamentally overcoming the edge effect problem of the block processing strategy.

[0097] Technology I: AI global optimization technology

[0098] By pre-classifying the vehicles, we can produce the most fuel-efficient section survey results, the highest transportation efficiency section survey results, etc. These results are equivalent to the optimal solution for manual driving at the section level. Based on this, the optimal solution for manual driving of spatially continuous road sections can be used as the pre-input of the AI ​​large model to further obtain the global optimal solution. A typical application example is: starting from the current road section, the speed of several subsequent road sections is smoothed, and the proposed results can be used as the speed control parameters for the vehicle to drive more smoothly on a certain section of road. Taking the kinetic energy change as an example, the car starts from the garage, performs a transportation task, and returns to the garage. It is regarded as a transportation process. The entire transportation process is divided into one-second time intervals. The vehicle's kinetic energy per second can be directly calculated according to the kinetic energy formula in physics. The arithmetic sum of the kinetic energy changes of the car in adjacent seconds is equal to zero. Half of the absolute value of the sum of the kinetic energy changes of the car in adjacent seconds can be used to approximate the total energy output of the car in a transportation process, and the energy output directly comes from the consumption of on-board energy. Important off-cycle energy-saving technologies in the modern automobile industry, such as PCC technology, ensure that the speed of the adjacent sections is smooth as much as possible while ensuring the medium speed range of the current section, that is, to make the speed curve smoother on the coordinate plane with time as the horizontal axis and vehicle speed as the vertical axis. The calculation results of automobile big data show that the absolute value of the total kinetic energy change decreases by about 30% by using the sliding average of the speed before and after 5 seconds.

[0099] Technique J: Classification Optimization Technique

[0100] In a broad sense, the road is one of the factors with a clear classification effect in the complex complex of people, cars, roads and environment. Similarly, the factors with a clear classification effect in the automobile industry are as follows: longitude and latitude, corresponding to the common sense that cars are road transportation tools; driving direction, corresponding to the common sense of driving on the right, prohibiting reverse driving, etc.; vehicle speed, corresponding to the common sense of medium-speed driving, maximum speed limit, etc.; fuel consumption per 100 kilometers, corresponding to the common sense of throttle size; from the speed down, the relationship between indicators such as fuel consumption per 100 kilometers, torque, speed, intake volume, total current, total voltage, etc. and the specific model is getting closer and closer, and the transportation tasks performed by the vehicle are getting closer and closer. This requires that the technical personnel in this field should first complete the vehicle classification in the specific implementation, and then introduce the road section survey implementation of the classified operation data of the same type of vehicles. It should be emphasized that in the specific implementation, it is best to first complete the steps of road section segmentation technology based on driving direction (technology E), and then introduce conventional algorithms such as interpolation in specific vehicle applications so as not to cause obvious deviations from the actual road conditions.

[0101] The basic logic of the present invention is based on a large number of vehicle positioning coordinate points, converting the longitude and latitude coordinates that only have positions but no shape or size into grids that have both positions and shapes and sizes. Since each grid only represents an area of ​​0.01 square meters, sufficient data volume is a basic prerequisite. The basic knowledge and skills of various computers used in the present invention are regarded as the necessary prerequisite capabilities of the users of the present invention.

[0102] As a digital product, road orthophotos are the cornerstone for the construction of the subsequent continuous road survey system, the link for introducing AI large model technology, an important reference for training enterprise-level data and vehicle-level data, and an important basis for converting vehicle operating conditions into road operating conditions. They have a cornerstone status.

[0103] It should be noted that the terms used in the present invention are only for describing specific embodiments, rather than limiting the scope of the present invention. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular, but may also include the plural. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of more restrictions, the elements defined by the sentence "include one..." do not exclude the presence of other identical elements in the process, method or device including the elements.

[0104] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

[0106] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for producing road orthophotos using automobile big data, characterized in that: The following steps are involved: S1. Taking the latitude and longitude of the initial frame data of a single vehicle as the origin, constructing the initial frame rectangular coordinate system, dividing the initial frame rectangular coordinate system into four coordinate azimuth intervals of north, east, south and west; S2. Determine the direction of travel of the vehicle on the road according to the position where the latitude and longitude of a single vehicle in subsequent frame data falls within the coordinate azimuth interval; S3. In one map tile, fit the latitude and longitude data of multiple vehicles on the road in the same driving direction in the automobile big data to synthesize the orthophoto digital road image; The step S3 further includes the following steps: SD1. Selecting an optical remote sensing image located in the same map frame as the orthophoto digital road image, and scaling the optical remote sensing image to be completely consistent with the orthophoto digital road image; SD2. Using the orthophoto digital road image as a template, crop and correct the optical remote sensing image to remove all non-road pixels to obtain an orthophoto optical road image.

2. The method for producing a road orthophoto according to claim 1, characterized in that: The step S3 further includes the following steps: S31. Fitting the orthophoto digital road images under multiple map sections to obtain the orthophoto digital road images of the area to be measured.

3. The method for producing road orthophotos according to claim 1, characterized in that: The generation of the coordinate azimuth comprises the following steps: SA1. In the above coordinate azimuth interval, the north direction is between 315° and 45°, the east direction is between 45° and 135°, the south direction is between 135° and 225°, and the west direction is between 225° and 315°.

4. The method for producing a road orthophoto according to claim 1, characterized in that: The process of determining the map partitioning includes the following steps: SB1. Multiply the lowest resolution of the decimal latitude and longitude coordinates uploaded by the vehicle by the correction factor and round it to the integer; SB2. Name the rounded decimal longitude X, and the rounded decimal latitude Y; SB3. Let X = 2^n × Hx + Lx, Y = 2^n × Hy + Ly, where n is a positive integer, Hx, Lx, Hy, Ly are all natural numbers, and Lx and Ly are both less than 2^n; SB4. Name the map tile Ln_Hx_Hy.

5. The method for producing road orthophotos according to claim 4, characterized in that: The correction factor is 1,000,000 or 921,600.

6. The method for producing road orthophotos according to claim 4, characterized in that: n=16 in the map tiles.

7. The method for producing road orthophotos according to claim 1, characterized in that: The step S3 also includes a noise removal step: SC1. Get the frequency of the first grid in the orthophoto digital road image and read the result into the two-dimensional array in memory; SC2. Traverse each pixel in a map tile and make a sliding box with a side length of L centered on the pixel; SC3. Traverse the frequency of each pixel in the sliding frame and form a one-dimensional array of length 1×L; SC4. Sort the one-dimensional array from small to large, count the low frequency number n, and use the minimum frequency n0 corresponding to when n is greater than or equal to the pre-specified ratio as the basis for distinguishing the noise points of the sliding box; SC5. Traverse each pixel in the sliding frame and delete the pixels with a frequency less than or equal to n0 as noise.

8. The method for producing road orthophotos according to claim 7, characterized in that: 0＜n0＜20%。 9. The method for producing road orthophotos according to claim 8, characterized in that: n0=5%。 10. The method for producing road orthophotos according to claim 1, characterized in that: The step SD2 further includes the following steps: SD3. Fit the orthophoto road images under multiple map sections to obtain the orthophoto road images of the area to be measured.

11. An image product, characterized in that: The process of generating the image product adopts the method described in any one of claims 1 to 10.

Images