Job plot automatic recognition and calculation method based on convolutional neural network

Abstract

The application provides a job plot automatic identification and calculation method based on a convolutional neural network. The application introduces multi-dimensional motion parameters provided by satellite positioning and geographic environment parameters provided by basic geographic data, uses the combination of multiple data sources, greatly improves the job plot identification accuracy, and simultaneously introduces the geographic data parameters, so that the model can automatically adjust the weight according to the spatial distribution of the track data, so that the generality is higher under different geographic conditions and different operation habits in different regions, and the identification accuracy of super-small plots and special-shaped plots is higher.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of satellite positioning identification, and more particularly, to a method for automatically identifying and classifying work plots based on a convolutional neural network. BACKGROUND

[0002] The convolutional neural network is a high-efficiency recognition method developed in recent years and has attracted wide attention.

[0003] In the 1960s, Hubel and Wiesel discovered that the unique network structure of neurons for local sensitivity and direction selection in the cat cortex can effectively reduce the complexity of feedback neural networks when studying the neurons, and then proposed the convolutional neural network (CNN).

[0004] Now, the CNN has become one of the research hotspots in many scientific fields, especially in the field of pattern classification, because the network avoids complex pre-processing of images and can directly input original images, and thus has been more widely applied.

[0005] The new recognition machine proposed by K. Fukushima in 1980 is the first implementation network of the convolutional neural network.

[0006] Subsequently, more researchers have improved the network, and the representative research achievement is the "improved cognitive machine" proposed by Alexander and Taylor, which combines the advantages of various improvement methods and avoids time-consuming error back propagation.

[0007] Generally, the basic structure of the CNN includes two layers, one of which is a feature extraction layer, and the input of each neuron is connected to the local receptive field of the previous layer, and the local feature is extracted:

[0008] First, after the local feature is extracted, the positional relationship between the local feature and other features is also determined;

[0009] Second, the feature mapping layer, each calculation layer of the network is composed of multiple feature mappings, and each feature mapping is a plane, and the weights of all neurons on the plane are equal.

[0010] The feature mapping structure uses a sigmoid function with a small influence function kernel as the activation function of the convolutional network, so that the feature mapping has displacement invariance.

[0011] In addition, since the neurons on a mapping plane share weights, the number of free parameters of the network is reduced.

[0012] Each convolutional layer in CNN is followed by a computation layer for local averaging and secondary extraction, which reduces the feature resolution.

[0013] CNN is mainly used to identify two-dimensional figures that are invariant to displacement, scaling and other forms of distortion, and this function is mainly realized by the pooling layer.

[0014] Since the feature detection layer of CNN learns from training data, explicit feature extraction is avoided when using CNN, and implicit learning from training data is performed. Furthermore, since the weights of neurons on the same feature map are the same, the network can learn in parallel, which is a major advantage of convolutional networks over networks in which neurons are connected to each other.

[0015] Convolutional neural networks have unique advantages in speech recognition and image processing due to their special structure of local weight sharing. Their layout is closer to the actual biological neural network, and weight sharing reduces the complexity of the network. In particular, the feature of directly inputting images of multi-dimensional input vectors into the network avoids the complexity of data reconstruction in the feature extraction and classification process.

[0016] In traditional agriculture, most of the work is done manually, and the degree of automation is not high. For example, in the overall planning and calculation of agricultural cultivated land area, manual division and measurement are generally used, which consumes a lot of labor, and manual measurement has certain errors, which cannot accurately reflect the actual situation of cultivated land area.

[0017] With the development of China's Beidou satellite navigation system and the completion of Beidou-3 system networking, the Beidou system can provide reliable passive positioning services and meet the needs of precision agriculture for high-precision positioning, low-cost, and other needs, with good application advantages and potential, providing important technical support for precision agriculture measurement and monitoring. At the same time, with the increasing intelligence of agriculture, a more real-time and efficient monitoring system can obviously improve the planting effect of farms and regional agriculture. Using satellite navigation system to automatically detect the working state and manage the working tools can help to evaluate the overall efficiency and manage the work progress, and improve the data reliability, providing the possibility for the analysis and identification of agricultural work data.

[0018] In the existing process of satellite positioning trajectory recognition and calculation of farmland work, most of the simple recognition is based on the characteristics of single parameters such as work trajectory shape and work trajectory speed, and the reliability is greatly affected by the different geographical conditions and work habits in different regions, and the universality is very low.

[0019] If more complex parameters are introduced to build a more accurate calculation model, an abstract feature construction method needs to be introduced, and the continuous development of deep learning technology makes it possible.

[0020] Therefore, using the convolutional neural network method, more abstract high-level representation attribute categories or features are formed by combining low-level features, and distributed feature representation of open source discovery data is found.

[0021] Therefore, using the method can greatly improve the automatic identification and automatic classification of work plots.

[0022] Therefore, the existing problems are studied and improved, and a convolutional neural network-based automatic identification and calculation method of work plots is provided, which aims to solve the problems and improve the practical value through this technology. SUMMARY

[0023] The present application aims to provide a convolutional neural network-based automatic identification and calculation method of work plots to solve the problems and deficiencies in the background art.

[0024] To achieve the above-mentioned purpose, the present application provides a convolutional neural network-based automatic identification and calculation method of work plots, which is achieved by the following specific technical means:

[0025] The convolutional neural network-based automatic identification and calculation method of work plots comprises the following steps:

[0026] Step 1: Multi-dimensional feature image generation

[0027] As a further optimization of the present technical solution, the convolutional neural network-based automatic identification and calculation method of work plots of the present application mainly completes:

[0028] (1) Using geographic coding to convert vector trajectory data to generate raster data

[0029] According to the latitude and longitude information of the trajectory data in units of vehicles, the trajectory points are encoded (,) using simple geographic coding with the origin as (,) and the spatial resolution as (,), (,), (,). The time sequence points are connected.

[0030] The connection method is a straight line generation algorithm, which calculates the slope of the straight line and then determines whether the straight line pixel grid passes through the straight line according to the straight line equation. The generated trajectory line coding array is Dl.

[0031] The coding array is used to generate a trajectory image, where the pixel position (,) is the geographic coding (,), and the trajectory layer and parameter information layer are generated. The pixel value of the trajectory layer is the number of times the geographic coding is repeated in the coding array, and the pixel value of the parameter information layer is the average value of the sensor carried by the geographic coding trajectory point.

[0032] The pixel value of the generated trajectory image is normalized using log to avoid the interference of long tail data caused by parking and other trajectory stagnation scenarios.

[0033] The generated parameter information layer Pi pixel value is normalized.

[0034] (2) Conversion into raster data using external geographic vector data

[0035] The maximum and minimum longitude and latitude information of the track image is used to intercept the vector layer, and a raster image is established using the scan line method.

[0036] The scan line method is specifically to sequentially scan the lines on the vector graphics with a spatial resolution as an interval from the minimum longitude to the maximum longitude, and if the graphics exist, the block raster value is marked as 1, otherwise as 0, and finally the raster form data of the basic geographic vector is obtained.

[0037] (3) Unified into raster data using external geographic raster data

[0038] The raster of the track image is overlaid on the basic geographic raster data, the corresponding raster on each raster is calculated, the new pixel value is calculated, and finally the raster form data of the new basic geographic vector is obtained, and the pixel value is normalized.

[0039] Step two, using a multi-dimensional convolutional neural network to identify the work plot;

[0040] As a further optimization of the technical solution, the application discloses an automatic identification and calculation method of a work plot based on a convolutional neural network, and the use of a multi-dimensional convolutional neural network to identify the work plot mainly completes:

[0041] (1) The generated two-dimensional track image, two-dimensional basic geographic vector raster, and two-dimensional basic geographic raster data are combined into a new multi-dimensional image, and the number of dimensions of the multi-dimensional image is determined by the number of sensors of the carrier itself and the number of external geographic information data;

[0042] (2) In order to remove the redundancy caused by repeated or similar attributes to calculation, PCA principal component analysis is performed on the multi-dimensional image, and more than 80% of the fields with a contribution rate are combined into a new multi-dimensional image with a dimension of n;

[0043] (3) A deep learning network is used to train and identify the work plot to obtain a segmentation result:

[0044] A 3*3 convolution kernel is used for convolution calculation, the input channel is n, and the output channel is 64; then a 3*3 convolution kernel is used for convolution calculation, the channel number is 64, and finally the ReLU function is activated;

[0045] A maximum pooling layer is used to simplify the image to half of the original size.

[0046] A 3*3 convolution kernel with a channel number of 128 is used for convolution calculation, and finally the ReLU function is activated.

[0047] The maximum pooling layer is used to simplify the image to half of the original size;

[0048] The 2 layers of 3*3 convolution kernels with 256 channels are used for convolution calculation, and finally activated by the ReLU function;

[0049] The maximum pooling layer is used to simplify the image to half of the original size;

[0050] The 2 layers of 3*3 convolution kernels with 512 channels are used for convolution calculation, and finally activated by the ReLU function;

[0051] The maximum pooling layer is used to simplify the image to half of the original size;

[0052] The 2 layers of 3*3 convolution kernels with 512 channels are used for convolution calculation, and finally activated by the ReLU function;

[0053] The maximum pooling layer is used to simplify the image to half of the original size;

[0054] The 2 layers of 1*1*4096 full connection layers are used, and the ReLU function is activated;

[0055] Finally, the recognition result is output by the Softmax.

[0056] Step three, the recognition result is combined with the time sequence relationship to regress into vector data;

[0057] As a further optimization of the technical solution, the application discloses an automatic identification and calculation method of work plots based on a convolutional neural network, and the recognition result combined with the time sequence relationship is mainly used to regress into vector data, which mainly completes:

[0058] (1) The track code D_P is matched with the segmentation result image in step two, and the identification label result information is restored to the track data as attribute information;

[0059] (2) The median filter with a window of 7 is used to process the identification label result attribute of the track data in time sequence, the start and end times of the work plot corresponding to the track are obtained through the difference of the label result attribute and the time interval threshold Tt, the work area vector MP_B is generated according to the track and the work range, and the plot geographic code D_b is generated according to the coding rule in step one.

[0060] Step four, the work plot area calculation of the recognition result.

[0061] As a further optimization of the technical solution, the application discloses an automatic identification and calculation method of work plots based on a convolutional neural network, and the work plot area calculation of the recognition result mainly completes:

[0062] (1) Perform dilation operation on the working surface vector with a width of half of the working width W / 2 to restore the true working surface MP;

[0063] (2) Perform closing operation on the true working surface MP with a width of W / 2 to eliminate the noise points in the plot;

[0064] (3) Extract the outer contour data in the true working surface MP, and smooth the outer contour trajectory using the Douglas-Peucker method, which connects the first and last points of the curve to be processed with a straight line, calculates the distance of all intermediate points from the straight line, and finds the maximum distance value D_max, and compares D_max with the thinning threshold T_dp:

[0065] If D_max is less than T_dp, all intermediate points on the curve are discarded;

[0066] If D_max is greater than or equal to T_dp, the curve is divided into two parts with the point as the boundary;

[0067] Repeat the above process for the two parts of the curve until all points are processed.

[0068] (4) Project all points of the obtained contour in the true working surface MP_B onto the horizontal coordinate, and calculate the trapezoidal area formed by connecting the points in sequence and the points projected onto the horizontal coordinate, wherein the trapezoidal area from left to right connecting points is positive, and the trapezoidal area from right to left connecting points is negative, and the sum of the trapezoidal areas is the total area of the entire figure.

[0069] Due to the use of the above technical solutions, the present application has the following advantages compared with the prior art:

[0070] The present application introduces multi-dimensional motion parameters provided by satellite positioning and geographic environment parameters provided by basic geographic data, uses the combination of multi-source data to greatly improve the recognition accuracy of the working plot, and the introduction of geographic data parameters can automatically adjust the weight of the model according to the spatial distribution of the trajectory data, so that the generality is higher under different geographical conditions and different working habits, and the recognition accuracy of super-small plots and special-shaped plots is higher. BRIEF DESCRIPTION OF DRAWINGS

[0071] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application, and are incorporated in and constitute a part of this application. The embodiments of the application illustrated in the drawings, and their description, are used to explain the application and are not intended to limit the application. In the drawings:

[0072] Figure 1 is a schematic diagram of the method structure of the present application; DETAILED DESCRIPTION

[0073] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0074] It should be noted that, in the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more than two; the terms "upper", "lower", "left", "right", "inner", "outer", "front end", "rear end", "head", "tail" and the like indicate the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore it cannot be understood as a limitation on the present application.

[0075] In addition, the terms "first", "second", "third" and the like are only for descriptive purposes, and cannot be understood as indicating or implying relative importance.

[0076] Meanwhile, in the description of the present application, unless otherwise specified and limited, the terms "connected", "connected" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0077] Please refer to Figure 1 The present application provides a specific technical implementation scheme of a job plot automatic identification and calculation method based on convolutional neural network:

[0078] As shown in the multi-dimensional feature image generation method in Figure 1 The parameter generation is performed in the following three ways:

[0079] (1) Use geographic coding to convert vector trajectory data to generate raster data:

[0080] According to the latitude and longitude information of the trajectory data in units of vehicles, use simple geographic coding with the origin as (,), and the spatial resolution as (,), (,), (,) to encode the trajectory points (,), and connect the time sequence points;

[0081] The connection method is a straight line generation algorithm, that is, after calculating the slope of the straight line, it is judged whether the straight line pixel grid passes through the straight line according to the straight line equation, and the encoding array of the generated trajectory line is Dl;

[0082] The trajectory image is generated using the encoding array, wherein the pixel position (i, j) is the geographic code (i, j), the trajectory layer and the parameter information layer are generated. The pixel value of the trajectory layer is the number of times that the geographic code is repeated in the encoding array, and the pixel value of the parameter information layer is the average value of the sensors carried by the geographic code trajectory point.

[0083] The pixel value of the generated trajectory image is normalized by using log to avoid the interference of long tail data caused by the parking and other trajectory stagnation scenarios.

[0084] The pixel value of the generated parameter information layer Pi is normalized.

[0085] (2) Conversion into raster data using external geographic vector data

[0086] The maximum and minimum longitude and latitude information of the trajectory image is used to intercept the vector layer, and the raster image is established using the scan line method.

[0087] The scan line method is specifically that whether there is a vector graph on the scan line is scanned in sequence at a spatial resolution interval from the minimum longitude to the maximum longitude, if there is a graph, it is marked that the block raster value is 1, otherwise it is 0, and finally the raster form data of the basic geographic vector is obtained.

[0088] (3) Unified into raster data using external geographic raster data

[0089] The raster of the trajectory image is overlaid on the basic geographic raster data, the corresponding raster on each raster is calculated, the new pixel value is calculated, and finally the raster form data of the new basic geographic vector is obtained, and the pixel value is normalized.

[0090] Step two, the multi-dimensional convolutional neural network is used to identify the work plot:

[0091] As a further optimization of the technical solution, the application discloses an automatic identification and calculation method of a work plot based on a convolutional neural network, and the use of a multi-dimensional convolutional neural network to identify the work plot mainly completes:

[0092] (1) The generated two-dimensional trajectory image, two-dimensional basic geographic vector raster, and two-dimensional basic geographic raster data are combined into a new multi-dimensional image. At this time, the number of dimensions of the multi-dimensional image is determined by the number of sensors of the carrier itself and the number of external geographic information data.

[0093] (2) In order to remove the redundancy caused by repeated or similar attributes to calculation, PCA principal component analysis is performed on the multi-dimensional image, and more than 80% of the fields with contribution rate are combined into a new multi-dimensional image with n dimensions.

[0094] (3) The segmentation result is obtained by using a deep learning network type to train and identify the work plot:

[0095] Convolution calculation is performed using 1 layer of 3*3 convolution kernel, the input channel is n, the output channel is 64, then convolution calculation is performed using 1 layer of 3*3 convolution kernel, the channel number is 64, and finally the ReLU function is activated;

[0096] The maximum pooling layer is used to simplify the image to half of the original size;

[0097] Convolution calculation is performed using 2 layers of 3*3 convolution kernel with a channel number of 128, and finally the ReLU function is activated;

[0098] The maximum pooling layer is used to simplify the image to half of the original size;

[0099] Convolution calculation is performed using 2 layers of 3*3 convolution kernel with a channel number of 256, and finally the ReLU function is activated;

[0100] The maximum pooling layer is used to simplify the image to half of the original size;

[0101] Convolution calculation is performed using 2 layers of 3*3 convolution kernel with a channel number of 512, and finally the ReLU function is activated;

[0102] The maximum pooling layer is used to simplify the image to half of the original size;

[0103] Convolution calculation is performed using 2 layers of 3*3 convolution kernel with a channel number of 512, and finally the ReLU function is activated;

[0104] The maximum pooling layer is used to simplify the image to half of the original size;

[0105] After 2 layers of 1*1*4096 full connection layer, the ReLU function is activated;

[0106] Finally, the recognition result is output by Softmax.

[0107] Step three, vector data regression of grid recognition result:

[0108] As a further optimization of the technical solution, the present application is based on the working plot automatic identification and calculation method of convolution neural network, and the recognition result is combined with the time sequence relationship to regress into vector data, mainly completing:

[0109] (1) The track code D_P is matched with the segmentation result image of the second step, and the recognition label result information is restored to the track data as attribute information;

[0110] (2) using the median filter with a window of 7 to process the identification label result attribute of the trajectory data in time sequence, obtaining the start and end time of the trajectory corresponding to the work plot through the difference of the label result attribute and the time interval threshold Tt, and generating the work face vector MP_B according to the trajectory and the work range, and simultaneously generating the plot geographic code D_b according to the coding rule in step one.

[0111] Step four, work plot area calculation:

[0112] As a further optimization of the technical solution, the work plot automatic identification and calculation method based on the convolutional neural network of the present application mainly completes the following:

[0113] (1) performing dilatation operation on the work face vector with a width of half of the work width W / 2 to restore the real work face MP;

[0114] (2) performing closed operation on the real work face MP with a width of W / 2 to eliminate the noise points in the plot;

[0115] (3) extracting the outer contour data in the real work face MP, and smoothing the outer contour trajectory using the Douglas-Peucker method, which connects the first and last points of the curve to be processed with a straight line, calculates the distance of all intermediate points from the straight line, and finds the maximum distance value D_max, and compares D_max with the thinning threshold T_dp:

[0116] If D_max is less than T_dp, all intermediate points on the curve are discarded;

[0117] If D_max is greater than or equal to T_dp, the curve is divided into two parts with the point as the boundary;

[0118] Repeat the above process for the two parts of the curve until all points are processed.

[0119] (4) projecting all points of the obtained real work face MP_B contour onto the horizontal coordinate, and calculating the trapezoidal area formed by sequentially connecting the points and the points projected onto the horizontal coordinate, wherein the trapezoidal area from left to right connecting points is positive, and the trapezoidal area from right to left connecting points is negative, and the sum of the trapezoidal areas is the total area of the entire figure.

[0120] In summary: the present application introduces multi-dimensional motion parameters provided by satellite positioning and geographic environment parameters provided by basic geographic data, uses the combination of multiple sources of data to greatly improve the identification accuracy of work plots, and the introduction of geographic data parameters can automatically adjust the weight according to the spatial distribution of trajectory data, making it more widely used in different geographical conditions and different work habits, and the identification accuracy of super-small plots and special-shaped plots is higher.

[0121] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for automatic identification and calculation of work plots based on convolutional neural networks, comprising the following steps: Step one: multi-dimensional feature image generation; Step two: multi-dimensional convolutional neural network identification of work plots; Step three: vector data regression of grid identification results; Step four: work plot area calculation; The step one: multi-dimensional feature image generation is mainly completed in the following way: (1) Use geographic coding to convert vector trajectory data to generate grid data According to the latitude and longitude information of the trajectory data of the vehicle as a unit, using the origin as (X0, Y0), the space resolution as (T X ,T Y ), (T X / 2, T Y / 2), (T X / 4, T Y / 4), the trajectory points are encoded D P (D X , D Y ), and the time sequence points are connected; The connection method is a straight line generation algorithm. After calculating the slope of the straight line, it is determined whether the pixel grid passes through the straight line according to the straight line equation. The generated trajectory line coding array is D1; Generating a trajectory image using an encoding array D1 wherein the pixel position (P X , P Y ) is a geo-encoding (D X , D Y ), a trajectory layer and a parameter information layer P i are generated; the pixel value of the trajectory layer is the number of repetitions of the geo-encoding in the encoding array, and the pixel value of the parameter information layer is the average value of the sensors carried by the trajectory point of the geo-encoding; Generated trajectory image Pixel values using log normalization P t = log 10 p t / log 10 P max To avoid long tail data interference caused by trajectory stagnation scenarios such as parking. The generated parameter information layer P i The pixel value is normalized using P g = (P g -P min ) / (P max -P min ) (2) Convert external geographic vector data to grid data Using trajectory images The maximum and minimum latitude and longitude information of the vector layer is intercepted, and a raster image is established using a scan line method. The scan line method is specifically from the minimum longitude to the maximum longitude in turn with a spatial resolution of T w If there is a vector graphic on the interval scan line, the graphic is marked as the block raster value of 1, otherwise as 0, and finally the raster form data P of the basic geographic vector is obtained v ; (3) Use external geographic grid data to unify into grid data The trajectory image P l is overlaid onto the base geographic grid data G g , and the corresponding grid P XY on G g is calculated, and a new pixel value P XY = (∑ i ∑ j G g (i, j)) / (i*j) is calculated, and finally a new base geographic vector grid form data P g is obtained, and the pixel value P g is normalized using P g = (P g -P min ) / (P max -P min ). The step two: multi-dimensional convolutional neural network identification of work plots is mainly completed in the following way: (1) the generated two-dimensional trajectory image two-dimensional basic geographic vector grid P v , two-dimensional basic geographic grid data P g , the superposition combination is a new multi-dimensional image P m At this time, the multi-dimensional image P m The dimension is determined by the number of sensors on the vehicle itself and the number of external geographic information data; (2) In order to remove the redundancy caused by repeated or similar attributes, the multi-dimensional image P m is subjected to PCA principal component analysis, and the field combination with a contribution rate of more than 80% is taken as a new multi-dimensional image with a dimension of n. (3) Use deep learning network training to identify work plots to obtain segmentation results: Use 1 layer of 3*3 convolution kernel for convolution calculation, input channel is n and output channel is 64, then use 1 layer of 3*3 convolution kernel for convolution calculation, channel number is 64, and finally pass through ReLU function activation; Use the maximum pooling layer to simplify the image to half of the original size; Use 2 layers of 3*3 convolution kernel with a channel number of 128 for convolution calculation, and finally pass through ReLU function activation; Use the maximum pooling layer to simplify the image to half of the original size; Use 2 layers of 3*3 convolution kernel with a channel number of 256 for convolution calculation, and finally pass through ReLU function activation; Use the maximum pooling layer to simplify the image to half of the original size; Use 2 layers of 3*3 convolution kernel with a channel number of 512 for convolution calculation, and finally pass through ReLU function activation; Use the maximum pooling layer to simplify the image to half of the original size; Use 2 layers of 3*3 convolution kernel with a channel number of 512 for convolution calculation, and finally pass through ReLU function activation; Use the maximum pooling layer to simplify the image to half of the original size; Pass through 2 layers of 1*1*4096 fully connected layer, and pass through ReLU function activation; Finally, output the recognition result through Softmax; The step three: vector data regression of grid identification results, which is mainly completed in the following way: (1) Compare the trajectory coding D1 with the segmentation result image of step two, and restore the recognition label result information to the trajectory data as attribute information; (2) using the median filter with a window of 7 to process the identified label result attribute of the trajectory data in time sequence, obtaining the start and end times of the trajectory corresponding to the work plot through the difference of the label result attribute and the time interval threshold Tt, and generating a work face vector MP_B according to the trajectory and the work range, and simultaneously generating a plot geographic code D according to the coding rule in step one P ; The step four: work plot area calculation, which is mainly completed in the following way: (1) Perform dilation operation on the work face vector with a width of half of the work width W / 2 to restore the real work face MP; (2) Perform closed operation on the real work face MP with a width of W / 2 to eliminate noise points in the plot; (3) Extract the outer contour data in the real work face MP, and smooth the outer contour trajectory using the Douglas-Peucker method. The Douglas-Peucker method connects the first and last points of the curve to be processed with a straight line, calculates the distance of all intermediate points from the straight line, and finds the maximum distance value D_max. Compare D_max with the decimation threshold T_dp: If D_max is less than T_dp, all the intermediate points on the curve are discarded; If D_max is greater than or equal to T_dp, the curve is divided into two parts by the point; The above process is repeated for the two parts of the curve until all the points are processed; (4) Project all the points of the contour in the real working surface MP_B onto the horizontal coordinate, and calculate the trapezoidal area formed by sequentially connecting the points and the points projected onto the horizontal coordinate, wherein the trapezoidal area from left to right connecting points is positive, and the trapezoidal area from right to left connecting points is negative, and the sum of the trapezoidal areas is the total area of the entire figure.

Images