Image segmentation method and device

Abstract

The invention provides an image segmentation method and device, and the method comprises the steps: obtaining the track data of an agricultural image, the track data comprises a plurality of track points, and each track point is provided with a plurality of sub-data; for each track point, carrying out spatial feature extraction on the multiple items of sub-data of the track point to obtain target sub-data of each track point so as to determine the motion trend of the track point, and carrying out time feature extraction on the multiple items of sub-data of the track point to obtain a first time difference threshold value so as to determine the track segment to which the track point belongs; according to the motion trend of each track point and the track segment to which each track point belongs, integrating the plurality of track segments into a track segmentation result, classifying each track point in the track segmentation result, and determining the attribute of the track point; and according to the attribute of each track point, segmenting the agricultural image so as to differentially display a field area and a road area in the agricultural image. According to the invention, the accuracy of image segmentation is improved.

Description

Technical Field

[0001] The present application relates to the technical field of image segmentation, and in particular to an image segmentation method and device. Background Art

[0002] At present, among the existing field-road separation technologies, the more typical methods include the field-road separation method based on the KMeans clustering algorithm. This method usually preprocesses the trajectory data (such as standardization and dimensionality reduction), and then divides the trajectory points into different clusters through the KMeans clustering algorithm, and then distinguishes fields and roads based on the clustering results. However, this method has several significant disadvantages:

[0003] (1) Dependence on the predefined number of clusters: The KMeans algorithm requires the number of clusters to be set in advance, which requires certain experience and is difficult for scenes with complex distribution and unknown number of fields and road trajectories. If the number of clusters set in advance is unreasonable, it is easy to lead to deviations in clustering results.

[0004] (2) Unable to handle complex temporal differences: The traditional KMeans method mainly performs clustering based on spatial features, but ignores the temporal difference information in trajectory data. It cannot effectively deal with the temporal feature differences between fields and roads, which may lead to classification errors.

[0005] (3) Lack of flexibility: The KMeans algorithm has certain requirements for the type of trajectory data. When faced with different types of trajectory data, it shows the limitation of insufficient flexibility and is difficult to adapt to the needs of field-road separation in various environments. Summary of the invention

[0006] In view of this, an object of the present application is to provide an image segmentation method and device to overcome at least one of the above-mentioned defects.

[0007] In a first aspect, an embodiment of the present application provides an image segmentation method, the method comprising: acquiring trajectory data of an agricultural image, the trajectory data comprising a plurality of trajectory points, each trajectory point carrying multiple sub-data; for each trajectory point, performing spatial feature extraction on the multiple sub-data of the trajectory point to obtain target sub-data of each trajectory point to determine the motion trend of the trajectory point, performing time feature extraction on the multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs; integrating the multiple trajectory segments into a trajectory segmentation result according to the motion trend of each trajectory point and the trajectory segment to which it belongs, and classifying each trajectory point in the trajectory segmentation result to determine the attribute of the trajectory point; and segmenting the agricultural image according to the attribute of each trajectory point to distinguish and display the field area and the road area in the agricultural image.

[0008] In an optional embodiment of the present application, the sub-data includes longitude data, latitude data, timestamp, speed and direction angle, wherein the movement trend of each trajectory point is determined in the following manner: for each trajectory point, the longitude data and latitude data in the sub-data corresponding to the trajectory point are extracted, and the time in the sub-data corresponding to the trajectory point and the adjacent trajectory point is extracted to determine the time difference between the two adjacent trajectory points; the geographical distance between the two adjacent trajectory points is determined according to the time difference, longitude data and latitude data between the two adjacent trajectory points, and the speed difference between the two adjacent trajectory points is determined according to the geographical distance and the time difference; the movement direction angle of each trajectory point is determined according to the speed difference between the two adjacent trajectory points by using the inverse tangent function to determine the movement trend of each trajectory point.

[0009] In an optional embodiment of the present application, the trajectory segment to which each trajectory point belongs is determined in the following manner: for each trajectory point, extract the timestamp in the sub-data corresponding to the trajectory point, and determine the time interval between two adjacent trajectory points; count the occurrence frequencies of all time intervals, and determine the time interval with the highest occurrence frequency as the first time difference threshold; for each time interval, determine the coefficient corresponding to the time interval, and calculate the product of the first time difference threshold and the coefficient corresponding to the time interval, and determine the product as the second time difference threshold of the time interval; for each time interval, determine whether the time interval is less than or equal to the second time difference threshold; for each time interval, if the time interval is less than or equal to the second time difference threshold, divide the two adjacent trajectory points corresponding to the time interval into the same trajectory segment; if the time interval is greater than the second time difference threshold, create a new trajectory segment, and divide the two adjacent trajectory points corresponding to the time interval into the new trajectory segment.

[0010] In an optional embodiment of the present application, the method integrates multiple trajectory segments into a trajectory segmentation result according to the movement trend of each trajectory point and the trajectory segment to which it belongs, and classifies each trajectory point in the trajectory segmentation result to determine the attribute of the trajectory point, including: clustering each trajectory segment and integrating it into a trajectory segmentation result, the trajectory segmentation result including multiple trajectory points arranged in chronological order and sub-data corresponding to each trajectory point; obtaining the time interval between two adjacent trajectory points in the trajectory segmentation result, and determining whether the time interval between two adjacent trajectory points is less than or equal to the first time difference threshold; if the time interval between two adjacent trajectory points is less than or equal to the first time difference threshold, it is considered that the attribute of the two adjacent trajectory points corresponding to the time interval is the first attribute; if the time interval between two adjacent trajectory points is greater than the first time difference threshold, it is considered that the attribute of the latter trajectory point in the two adjacent trajectory points corresponding to the time interval is the second attribute.

[0011] In an optional embodiment of the present application, the method of distinguishing and displaying the field area and the road area in the agricultural image according to the attribute of each trajectory point includes: identifying the attribute of each trajectory point, marking the trajectory point with the first attribute as a field trajectory point, and marking the trajectory point with the second attribute as a road trajectory point; screening out an adjacent field trajectory point and a road trajectory point from a plurality of field trajectory points and a plurality of road trajectory points, and calculating the Euclidean distance between the field trajectory point and the road trajectory point; determining a potential intersection point according to the Euclidean distance, and correcting the potential intersection point according to the density and distribution characteristics of the trajectory points around the potential intersection point to determine a final dividing point; integrating the field trajectory points, the road trajectory points and the final dividing point to distinguish and display the field area and the road area in the agricultural image.

[0012] In an optional embodiment of the present application, the potential intersection point is determined in the following manner: a distance difference threshold is determined based on the first time difference threshold; whether the Euclidean distance is less than or equal to the distance difference threshold is determined; if the Euclidean distance is less than or equal to the distance difference threshold, a field trajectory point and a road trajectory point corresponding to the Euclidean distance are determined as potential intersection points; if the Euclidean distance is greater than the distance difference threshold, a field trajectory point and a road trajectory point corresponding to the Euclidean distance are not determined as potential intersection points.

[0013] In an optional embodiment of the present application, the trajectory data of the agricultural image is obtained in the following manner: first data collected by a sensor is obtained, the first data includes multiple trajectory points, and each trajectory point carries multiple sub-data; the first data is formatted, and the multiple sub-data of each trajectory point in the first data are converted into units and standardized to determine the second data; for each trajectory point in the second data, the actual speed and actual direction angle of the trajectory point are calculated; for each trajectory point in the second data, it is determined whether the actual speed corresponding to the trajectory point is equal to the speed stored in the second data; for each trajectory point in the second data, if the trajectory point is equal to the speed stored in the second data, the actual speed and actual direction angle of the trajectory point are calculated; for each trajectory point in the second data, if the actual speed corresponding to the trajectory point is equal to the speed stored in the second data, the actual speed of the trajectory point is equal to the speed stored in the second data. If the actual speed corresponding to the trajectory point is not equal to the speed stored in the second data, the trajectory point is considered to be unreasonable and is corrected; for each trajectory point in the second data, if the actual speed corresponding to the trajectory point is equal to the speed stored in the second data, the trajectory point is considered to be reasonable and is saved, and it is judged whether the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data; for each trajectory point in the second data, if the actual direction angle corresponding to the trajectory point is not equal to the direction angle stored in the second data, the trajectory point is considered to be unreasonable and is corrected; if the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data, the trajectory point is considered to be reasonable and is saved.

[0014] In a second aspect, an embodiment of the present application further provides an image segmentation device, which includes: an acquisition module, used to acquire trajectory data of an agricultural image, wherein the trajectory data includes multiple trajectory points, and each trajectory point carries multiple sub-data; a first determination module, used to perform spatial feature extraction on the multiple sub-data of each trajectory point to obtain target sub-data of each trajectory point to determine the movement trend of the trajectory point, and perform time feature extraction on the multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs; a second determination module, used to integrate multiple trajectory segments into a trajectory segmentation result according to the movement trend of each trajectory point and the trajectory segment to which it belongs, and classify each trajectory point in the trajectory segmentation result to determine the attributes of the trajectory point; a display module, used to segment the agricultural image according to the attributes of each trajectory point to distinguish and display the field area and the road area in the agricultural image.

[0015] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method described above are performed.

[0016] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are executed.

[0017] The image segmentation method and device provided by the embodiment of the present application obtain the trajectory data of the agricultural image, the trajectory data includes multiple trajectory points, each trajectory point has multiple sub-data; for each trajectory point, the multiple sub-data of the trajectory point are subjected to spatial feature extraction to obtain the target sub-data of each trajectory point to determine the movement trend of the trajectory point, and the multiple sub-data of the trajectory point are subjected to time feature extraction to obtain the first time difference threshold to determine the trajectory segment to which the trajectory point belongs; according to the movement trend of each trajectory point and the trajectory segment to which it belongs, the multiple trajectory segments are integrated into a trajectory segmentation result, and each trajectory point in the trajectory segmentation result is classified to determine the attribute of the trajectory point; according to the attribute of each trajectory point, the agricultural image is segmented to distinguish and display the field area and the road area in the agricultural image. Through the present application, the accuracy of image segmentation is improved.

[0018] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flowchart of an image segmentation method provided in an embodiment of the present application;

[0021] Figure 2 A schematic diagram of the structure of an image segmentation device provided in an embodiment of the present application;

[0022] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work belongs to the scope of protection of the present application.

[0024] First, the application scenarios to which the present application is applicable are introduced. The present application can be applied to the field of image segmentation.

[0025] Research has found that distinguishing the spatial information of fields and roads is of great significance to the modernization and intelligent development of agriculture and transportation. In agricultural production, the realization of farm intelligence is inseparable from the precise operation of agricultural machinery. Accurately distinguishing roads and fields provides a reliable basis for the precise operation of intelligent agricultural machinery. In intelligent transportation systems, accurate identification of road information can effectively assist the navigation system in precise navigation, and at the same time, it helps to optimize the vehicle dispatching system.

[0026] Traditional field-road separation methods mainly rely on manual annotation or simple geometric features, such as manual division based on topographic maps or remote sensing images. However, this method is not only labor-intensive and inefficient, but also cannot adapt to complex terrain and environments, and has low accuracy and precision. For example, the recognition method that directly divides by position coordinates is a type of automatic recognition method that uses simple geometric features. When faced with complex scenes where fields and roads are mixed, the accuracy is often low.

[0027] Based on this, the embodiment of the present application provides an image segmentation method and device, which obtains the trajectory data of agricultural images; for each trajectory point, the spatial feature extraction of the multiple sub-data of the trajectory point is performed to obtain the target sub-data of each trajectory point to determine the movement trend of the trajectory point, and the time feature extraction of the multiple sub-data of the trajectory point is performed to obtain the first time difference threshold to determine the trajectory segment to which the trajectory point belongs; according to the movement trend of each trajectory point and the trajectory segment to which it belongs, multiple trajectory segments are integrated into a trajectory segmentation result, and each trajectory point in the trajectory segmentation result is classified to determine the attribute of the trajectory point; according to the attribute of each trajectory point, the agricultural image is segmented to distinguish and display the field area and the road area in the agricultural image. Through this application, the accuracy and precision of image segmentation are improved.

[0028] See also Figure 1 , Figure 1 Flow chart of the image segmentation method provided in the embodiment of the present application. Figure 1 As shown in , the image segmentation method provided by the embodiment of the present application includes:

[0029] S101. Acquire trajectory data of agricultural images.

[0030] The trajectory data includes multiple trajectory points, and each trajectory point carries multiple sub-data.

[0031] In this step, the trajectory data of agricultural images is obtained by:

[0032] Acquire first data collected by a sensor, the first data including a plurality of trajectory points, each trajectory point having multiple sub-data;

[0033] Here, the first data transmitted back from the vehicle is obtained in real time. The first data must undergo data preprocessing and data cleaning to avoid confusion of dirty data and ensure data quality.

[0034] Use data analysis technology to extract key indicators of vehicle operation: After data preprocessing is completed, use data analysis technology to extract key indicators of vehicle operation, such as speed, time interval and position change.

[0035] For example, GPS (Global Positioning System) devices, vehicle sensors, or drones and other devices can be used to obtain track data of fields and roads. The track data may include key information such as timestamp, longitude data, latitude data, speed, and direction angle.

[0036] Performing formatting processing on the data, performing unit conversion and standardization processing on multiple sub-data of each trajectory point in the first data, so as to determine the second data;

[0037] Preferably, the first data collected is formatted, and the time is uniformly formatted into the standard format of "yyyy-MM-ddHH:mm:ss". At the same time, other data fields (such as speed and direction angle) are converted and standardized, and the unit of speed is unified into km / h, and the unit of direction angle is unified into degree. After the data format is unified, subsequent processing is more convenient.

[0038] For each trajectory point in the second data, calculating the actual speed and the actual direction angle of the trajectory point;

[0039] For each trajectory point in the second data, determine whether the actual speed corresponding to the trajectory point is equal to the speed stored in the second data; if the actual speed corresponding to the trajectory point is not equal to the speed stored in the second data, the trajectory point is considered to be unreasonable and is corrected; if the actual speed corresponding to the trajectory point is equal to the speed stored in the second data, the trajectory point is considered to be reasonable and is saved, and determine whether the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data;

[0040] For each trajectory point in the second data, if the actual direction angle corresponding to the trajectory point is not equal to the direction angle stored in the second data, the trajectory point is considered unreasonable and is corrected; if the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data, the trajectory point is considered reasonable and is saved.

[0041] In an optional embodiment, unreasonable data in the collected data is deleted, including removing missing, duplicate or abnormal values, and the unreasonable data is corrected or eliminated by calculating the rationality of the speed and direction angle.

[0042] Here, the actual speed between the two trajectory points can be determined by calculating the distance and timestamp between the two trajectory points; a coordinate system is established through the longitude and latitude data of the two trajectory points to determine the actual direction angles of the two trajectory points. If the actual direction angle is greater than the direction angle in the second data, it is considered that the direction angle corresponding to the trajectory point is unreasonable, and the direction angle in the second data is modified to the actual direction angle. The same applies to the speed, which will not be elaborated here.

[0043] Here, all corrected second data are determined as trajectory data.

[0044] S102: for each trajectory point, extract spatial features from multiple sub-data of the trajectory point to obtain target sub-data of each trajectory point to determine the movement trend of the trajectory point; extract time features from multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs.

[0045] Preferably, the sub-data includes longitude data, latitude data, timestamp, speed and direction angle.

[0046] In an optional embodiment, spatial feature extraction is performed on multiple sub-data of each trajectory point. Specifically, the motion trend of each trajectory point is determined in the following manner:

[0047] For each trajectory point, extract the longitude data and latitude data in the sub-data corresponding to the trajectory point, extract the time in the sub-data corresponding to the trajectory point and the adjacent trajectory point, and determine the time difference between the two adjacent trajectory points;

[0048] The longitude and latitude data are read from the trajectory data to form a WGS84 geographic coordinate data set, which includes each trajectory point and the longitude and latitude data of each trajectory point.

[0049] Here, longitude is from 0° west longitude to 180° east longitude (or 0° east longitude to 180° west longitude), east longitude is represented by positive numbers, and west longitude is represented by negative numbers; latitude is from 90° south latitude to 90° north latitude, south latitude is represented by negative numbers, and north latitude is represented by positive numbers.

[0050] Determine the geographic distance between the two adjacent trajectory points according to the time difference, longitude data and latitude data between the two adjacent trajectory points, and determine the speed difference between the two adjacent trajectory points according to the geographic distance and the time difference;

[0051] Preferably, the geographical distance between two adjacent trajectory points can be calculated by the following Haversine formula:

[0052] The Haversine formula is a formula for calculating the great circle distance between two points on the surface of the Earth. The formula takes into account that the Earth is approximately an ellipsoid rather than a perfect sphere, and therefore provides a more accurate distance calculation result. The basic form of the formula involves calculating the spherical center angle (that is, the angle between the longitude and latitude data between two adjacent track points). This angle is calculated by the difference between the longitude and latitude data and the radius of the Earth (usually the average value in the WGS84 standard). The specific form of the formula is as follows:

[0053]

[0054] c=2×atan2(a,1-a)

[0055] d=R×c

[0056] Among them, Δφ represents the difference in latitude data between two adjacent trajectory points, φ1 and φ2 represent the latitude data of two adjacent trajectory points, Δλ represents the adjusted difference in longitude data between two adjacent trajectory points on the longitude circle, R represents the radius of the earth, a represents the measured value of the latitude difference between two adjacent trajectory points, c represents the intermediate variable, and d represents the geographical distance between two adjacent trajectory points.

[0057] The velocity difference between two adjacent trajectory points is calculated using the following formula:

[0058]

[0059] Here, Δv represents the speed difference between two adjacent track points, d represents the geographical distance between two adjacent track points, Δt represents the time difference between two adjacent track points, and 3600 represents the conversion of seconds into hours (1 hour = 3600 seconds).

[0060] If the distance is in kilometers and the time difference is in seconds, the speed will be in kilometers per hour (km / h).

[0061] According to the speed difference between two adjacent trajectory points, the motion direction angle of each trajectory point is determined by the inverse tangent function to determine the motion trend of each trajectory point.

[0062] Specifically, the motion direction angle of each trajectory point is calculated in the following way:

[0063] The motion direction angle can be obtained by calculating the azimuth angle between two adjacent trajectory points, which usually involves the use of the inverse tangent function (arctan). Specifically, the direction angle calculation can be performed by the following steps:

[0064] First, obtain the longitude and latitude coordinates of two adjacent trajectory points (assuming they are point A (lat1, lon1) and point B (lat2, lon2)).

[0065] Then, the longitude difference (Δlon=lon2-lon1) and the latitude difference (Δlat=lat2-lat1) between the two points are calculated.

[0066] Next, use the Haversine formula or its simplified form to calculate the actual distance between the two points.

[0067] The calculation of the direction angle θ is based on the difference in latitude and longitude, taking into account the curvature of the earth. A common approximate calculation method is to use the atan2 function, which returns the angle between the vector from the x-axis to the point (y, x) and the x-axis in radians. In the calculation of the direction angle of the trajectory point, it can be expressed as:

[0068] θ=atan2(sin(Δlon)×cos(lat2), cos(lat1)×sin(lat2)-sin(lat1)×cos(lat2)×

[0069] cos(Δlon)

[0070] Among them, θ represents the moving direction angle of the trajectory point, Δlon represents the difference in longitude data between two adjacent trajectory points, lat2 represents the latitude data corresponding to the subsequent trajectory point between two adjacent trajectory points, and lat2 represents the latitude data corresponding to the preceding trajectory point between two adjacent trajectory points.

[0071] This formula takes into account the effect of the earth's curvature on the calculation of longitude difference, and gives the angle from the north direction to the trajectory segment clockwise. The direction angle describes the relative movement direction between trajectory points, which is very important for understanding and analyzing the movement trend of trajectory data.

[0072] In another optional embodiment, time feature extraction is performed on multiple sub-data of each trajectory point. Specifically, the trajectory segment to which each trajectory point belongs is determined in the following manner:

[0073] For each trajectory point, extract the timestamp in the sub-data corresponding to the trajectory point to determine the time interval between two adjacent trajectory points;

[0074] Here, according to the timestamp in the sub-data corresponding to the trajectory point, the time intervals between all adjacent trajectory points are calculated, and the occurrence frequencies of all calculated time intervals are counted (the frequency is represented by the size of the number), and stored in the data structure for subsequent analysis and use.

[0075] Counting the occurrence frequencies of all time intervals, and determining the time interval with the highest occurrence frequency as the first time difference threshold;

[0076] In this step, based on the statistical frequency data, the time interval with the highest occurrence frequency is selected as the first time difference threshold, which will be used as a preliminary standard to determine whether adjacent trajectory points belong to the same trajectory segment.

[0077] For each time interval, determine the coefficient corresponding to the time interval, and calculate the product of the first time difference threshold and the coefficient corresponding to the time interval, determine the product as the second time difference threshold of the time interval, and judge whether the time interval is less than or equal to the second time difference threshold. If the time interval is less than or equal to the second time difference threshold, the two adjacent trajectory points corresponding to the time interval are divided into the same trajectory segment. If the time interval is greater than the second time difference threshold, a new trajectory segment is created, and the two adjacent trajectory points corresponding to the time interval are divided into the new trajectory segment.

[0078] Specifically, the corresponding relationship between the time interval and the coefficient is shown in Table 1 below:

[0079] coefficient Seconds interval 1.25 2-3 1.5 4-5 2 5-15 2.5 16-30 3 31-60

[0080] Table 1

[0081] According to the interval of seconds in which the time interval is located, the target coefficient corresponding to the time interval is determined; according to the product of the first time difference threshold and the target coefficient, the second time difference threshold corresponding to the time interval is determined; by comparing the time interval with the second time difference threshold, the trajectory segment corresponding to the two trajectory points corresponding to the time interval is determined.

[0082] In the clustering process, the threshold is adjusted dynamically according to the ratio. For cases where the trajectory points are dense or driving continuously, the first time difference threshold is tightened; for cases where the trajectory points are sparse or have obvious stops, the first time difference threshold is relaxed. When the time difference between trajectory points exceeds the threshold, these points are divided into different trajectory segments, which are stored in segmented trajectory data sets respectively and clustered in segments.

[0083] Here, denser trajectory points indicate a shorter time interval, and sparser trajectory points indicate a longer time interval. Tightening and loosening the first time difference threshold indicates an increase or decrease in a nearly fixed value of the first time difference threshold. The fixed value is set manually and will not be described in detail here.

[0084] S103 , according to the movement trend of each trajectory point and the trajectory segment to which it belongs, multiple trajectory segments are integrated into a trajectory segmentation result, and each trajectory point in the trajectory segmentation result is classified to determine the attribute of the trajectory point.

[0085] Each trajectory segment is clustered and integrated into a trajectory segmentation result.

[0086] In this step, the trajectory segmentation result includes multiple trajectory points arranged in time order and sub-data corresponding to each trajectory point;

[0087] All optimized trajectory segments are integrated into the final trajectory segmentation result list. Each trajectory segment is checked to ensure that there is no missing or repeated data, and then the integrated trajectory segment is output as the output result.

[0088] Obtaining the time interval between two adjacent trajectory points in the trajectory segmentation result, and determining whether the time interval between the two adjacent trajectory points is less than or equal to a first time difference threshold;

[0089] If the time interval between two adjacent trajectory points is less than or equal to the first time difference threshold, the attribute of the two adjacent trajectory points corresponding to the time interval is considered to be the first attribute; if the time interval between two adjacent trajectory points is greater than the first time difference threshold, the attribute of the latter trajectory point in the two adjacent trajectory points corresponding to the time interval is considered to be the second attribute.

[0090] Here, the trajectory points are only divided into two categories: two adjacent trajectory points whose time interval is less than or equal to the first time difference threshold are classified into the same category, and the class label field of both trajectory points is set to 0; and two adjacent trajectory points whose time interval is greater than the first time difference threshold are classified into different categories, and the class label field of the trajectory point behind the two adjacent trajectory points is set to 1, indicating that they belong to different categories. For example, the first attribute is that the class label field corresponding to the trajectory point is 0, and the second attribute is that the class label field corresponding to the trajectory point is 1.

[0091] At the same time, in response to possible misclassifications in the preliminary clustering, the results are corrected by combining information such as speed and direction angle, dynamically adjusting the time difference threshold, and using the dbscan (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm to reprocess possible misclassified trajectory points to ensure the correct classification of trajectory data.

[0092] Preferably, a trajectory clustering model is constructed based on time intervals: the time intervals between adjacent trajectory points are calculated in seconds. Different coefficients are defined according to the second intervals, and the second with the highest frequency in the time interval is taken. The second is multiplied by different coefficients to divide the trajectory segments, and finally different trajectory segments are divided to construct a clustering model based on time intervals.

[0093] That is, the time interval with the highest frequency of occurrence is selected as the first time difference threshold, and the threshold is dynamically adjusted according to the coefficient to divide different trajectory segments, and then the segmented trajectory data set is preliminarily clustered. The clustering process in this step only divides the trajectory points into two categories: two trajectory points whose time interval between two adjacent trajectory points is less than or equal to the first time difference threshold are classified into the same category, and the class label field is set to 0 (0 represents field trajectory points); while two trajectory points whose time interval between two adjacent trajectory points is greater than the first time difference threshold are marked as 1 (1 represents road trajectory points), indicating that they belong to different classes.

[0094] Next, we need to further refine the points with class labels of 0, and filter the trajectory points from the data with class label fields of 0 according to information such as step length, speed, and point distribution density. By calculating the average interval distance of trajectory points and combining the trajectory sampling rate of a specific frequency, we dynamically generate a spatial distance threshold. By analyzing the time interval between trajectory points, we set the second time difference threshold, calculate the spatial distance and time difference between two adjacent trajectory points, and construct a distance matrix and a time matrix to evaluate the similarity between points.

[0095] At the same time, the set distance threshold and the second time difference threshold can be used to apply the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm to perform refined clustering processing on the trajectory data based on the previous preliminary clustering.

[0096] Specifically, for each trajectory point, find the points whose distance is less than the distance threshold and whose time difference is less than the time difference threshold from the distance matrix and the time matrix, mark them as the same cluster, set the class label field to 0, and set the rest to 1.

[0097] S104. Segment the agricultural image according to the attributes of each trajectory point to distinguish and display the field area and the road area in the agricultural image.

[0098] Identify the attribute of each track point, mark the track point with the first attribute as a field track point, and mark the track point with the second attribute as a road track point;

[0099] Here, the trajectory points are divided into "field trajectory points" (cluster label is 0) and "road trajectory points" (cluster label is 1), and possible misclassification phenomena are corrected, and finally accurate field trajectory points and road trajectory points are output.

[0100] Selecting an adjacent field trajectory point and a road trajectory point from among a plurality of field trajectory points and a plurality of road trajectory points, and calculating the Euclidean distance between the field trajectory point and the road trajectory point;

[0101] The Euclidean distance between field track points and road track points is calculated as follows:

[0102] First, we need to determine the coordinates of the field track point and the road track point. Assuming the coordinates of the field track point are (x1, y1) and the coordinates of the road track point are (x2, y2), the Euclidean distance f between the field track point and the road track point can be calculated using the following formula:

[0103]

[0104] Here, f represents the Euclidean distance between the field trajectory point and the road trajectory point, (x1, y1) represents the coordinates of the field trajectory point in the two-dimensional space, and (x2, y2) represents the coordinates of the road trajectory point in the two-dimensional space.

[0105] Determine potential intersection points based on Euclidean distance;

[0106] Specifically, potential junction points can be identified by:

[0107] Determine a distance difference threshold according to the first time difference threshold;

[0108] Determine whether the Euclidean distance is less than or equal to the distance difference threshold;

[0109] If the Euclidean distance is less than or equal to the distance difference threshold, a field track point and a road track point corresponding to the Euclidean distance are determined as potential intersection points; if the Euclidean distance is greater than the distance difference threshold, a field track point and a road track point corresponding to the Euclidean distance are not determined as potential intersection points.

[0110] According to the density and distribution characteristics of the trajectory points around the potential intersection point, the potential intersection point is modified to determine the final demarcation point;

[0111] The trajectory points marked as potential junctions were corrected, and the road trajectory points close to the fields were re-marked as field trajectory points according to the density and distribution characteristics of the trajectory points around the junctions to avoid misclassification.

[0112] Integrate field track points, road track points and final demarcation points to distinguish field areas and road areas in agricultural images.

[0113] All corrected track points are integrated into the final classification result, and the complete field track point dataset, road track point dataset and final junction point data are saved to ensure that the marking information of each track point is accurate.

[0114] Construct a field-road separation model to identify and distinguish field trajectory points from road trajectory points: After completing the preliminary clustering of trajectory data, some trajectory points are located on the boundary between fields and roads, which can easily lead to misclassification. In order to distinguish "field trajectory points" from "road trajectory points", it is necessary to further construct a field-road separation model to correctly classify trajectory points as field trajectory points or road trajectory points, and correct and optimize possible misclassifications to ensure the accuracy of the results.

[0115] For the trajectory points with cluster label 0, they are marked as "field trajectory points"; and the trajectory points with cluster label 1 are marked as "road trajectory points", and then find the points close to the road trajectory points or the field trajectory points. If they are close to the road trajectory points, the Euclidean distance between them and the road trajectory points is calculated; similarly, if they are close to the field trajectory points, the Euclidean distance between them and the field trajectory points is calculated.

[0116] According to the previously counted time difference frequency multiplied by different coefficients, a threshold is set to identify potential misclassified points. If a trajectory point is close to a road trajectory point or a field trajectory point and the calculated Euclidean distance is less than this threshold, it is necessary to analyze the trajectory points around it. If the labels of most of the surrounding trajectory points are the same and the distance to this trajectory point is within the threshold, it will be reclassified as the label of most of the surrounding trajectory points to avoid misclassification.

[0117] This application can visualize the separation results of field tracks and road tracks to help users intuitively understand the track distribution. The map is used to display the separated field track points, road track points and final dividing points. The field track points and road track points are distinguished by color. For example, the field track points are represented by red, the road track points are represented by blue, and the overall direction of the track is shown by a dotted line.

[0118] In the trajectory display, the final intersection point between the field and the road is marked with a specific symbol (such as a green dot), and a zoom function is provided so that the user can view the location and distribution of the final intersection point in detail.

[0119] At the same time, a report containing the separation results is automatically generated. The report includes information such as the number of field track points, the number of road track points and the number of final boundary points, the ratio of field track points to road track points, the number of final intersection points and their specific locations. The report can be exported to PDF or other formats for user analysis and reference.

[0120] Display and report field and road separation results to visualize the processed field and road trajectories and generate a detailed report summarizing the separation results.

[0121] The image segmentation method and device provided by the embodiment of the present application obtain the trajectory data of the agricultural image, the trajectory data includes multiple trajectory points, each trajectory point has multiple sub-data; for each trajectory point, the multiple sub-data of the trajectory point are subjected to spatial feature extraction to obtain the target sub-data of each trajectory point to determine the movement trend of the trajectory point, and the multiple sub-data of the trajectory point are subjected to time feature extraction to obtain the first time difference threshold to determine the trajectory segment to which the trajectory point belongs; according to the movement trend of each trajectory point and the trajectory segment to which it belongs, the multiple trajectory segments are integrated into a trajectory segmentation result, and each trajectory point in the trajectory segmentation result is classified to determine the attribute of the trajectory point; according to the attribute of each trajectory point, the agricultural image is segmented to distinguish and display the field area and the road area in the agricultural image. Through the present application, the efficiency and accuracy of image segmentation are improved.

[0122] This application effectively solves the problem of mixing field trajectories and road trajectories by comprehensively considering the clustering algorithm of time intervals and spatial characteristics, and provides a high-precision and highly adaptable field-road separation method, which ensures that the division of labor of each function is clear and can operate efficiently in complex agricultural scenarios. It also provides users with intuitive visualization results and detailed reports, providing strong support for agricultural automation and intelligent transportation applications.

[0123] The trajectory point clustering algorithm based on time interval in the present application clusters trajectory points based on time intervals, and automatically identifies and separates field and road trajectories by analyzing the time intervals between trajectory points and combining spatial features. The trajectory point clustering algorithm based on time intervals, especially the core and application part of the algorithm, namely the dynamic adjustment mechanism of the time interval threshold and its application in field-road separation, can not only adapt to complex agricultural operation scenarios, but also avoid the shortcomings of traditional clustering methods in continuity and time correlation.

[0124] This application can accurately separate the trajectories of fields and roads by correctly identifying the final intersection point, and correcting the misclassified trajectory points in combination with the clustering results, thereby achieving accurate field-road separation.

[0125] The present invention cleans the trajectory data and extracts the time features (such as time interval) and spatial features (such as speed, direction angle, etc.) of the trajectory data, providing data for subsequent cluster analysis and laying a good foundation. Moreover, this feature extraction and classification mechanism ensures the accuracy of trajectory data processing and improves processing efficiency.

[0126] This application displays the results of field and road track separation through a visual map, and provides junction point identification and detailed report generation functions. The visual display of field and road tracks uses colors to distinguish the track points of roads and fields, so that the results can be displayed intuitively and vividly, which not only makes it easier for users to understand and analyze the separation results, but also provides an important reference for agricultural management.

[0127] The present application automatically identifies and separates the trajectories of agricultural fields and roads through the analysis of comprehensive temporal and spatial features. The method utilizes GPS trajectory data, time interval clustering algorithm, and spatial distance calculation to accurately separate the trajectories of fields and roads. The present application also adopts a modular design, which may mainly include a data preprocessing module, a feature extraction module, a time difference clustering module, a field-road separation module, a trajectory display module, and a report generation module. The division of each module meets the requirements of high cohesion and low coupling. The modules cooperate with each other and can efficiently realize the calculation and display of field and road separation in complex agricultural operation scenarios, which is not shown here.

[0128] The existing KMeans clustering method only relies on spatial features for classification and ignores the importance of temporal features, which easily leads to classification errors in complex environments. However, the present application introduces time interval analysis, utilizes the temporal features in the trajectory data, and combines the spatial position features to achieve higher-precision field-road separation and reduce classification errors. At the same time, the present application combines multidimensional feature analysis to significantly reduce the misclassification rate and improve the accuracy and reliability of the separation results, especially in the area where fields and roads meet.

[0129] Traditional methods rely on a predefined number of clusters, which makes it difficult to adapt to the needs of different scenarios. Moreover, if the number of clusters is not set properly, the clustering results may not match the actual situation. This application uses a dynamic clustering method based on time intervals to flexibly adapt to different types of trajectory data and diverse environments, thereby improving the applicability of field-road separation.

[0130] This application analyzes the temporal and spatial characteristics of trajectory points and dynamically adjusts the clustering process, effectively solving the separation problem of mixed fields and roads and improving the adaptability of the system in complex environments.

[0131] The method used in this application can process large amounts of data, adopt a segmented processing method, calculate separately, and finally comprehensively process the results, thereby improving the ability to cope with big data and meeting the needs of actual conditions.

[0132] In summary, this application aims to solve the technical defects of existing field road separation methods in terms of accuracy, flexibility and adaptability, so as to achieve more accurate and flexible field road separation applications, which are suitable for multiple fields such as modern agriculture and intelligent transportation.

[0133] Based on the same inventive concept, an image segmentation device corresponding to the image segmentation method is also provided in the embodiment of the present application. Since the principle of solving the problem by the device in the embodiment of the present application is similar to the above-mentioned image segmentation method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0134] See also Figure 2 , Figure 2 This is a schematic diagram of the structure of the image segmentation device provided in the embodiment of the present application. Figure 2 As shown in , the image segmentation device 200 includes:

[0135] An acquisition module 201 is used to acquire trajectory data of an agricultural image, wherein the trajectory data includes a plurality of trajectory points, each of which carries multiple sub-data;

[0136] The first determination module 202 is used to extract spatial features from multiple sub-data of each trajectory point to obtain target sub-data of each trajectory point to determine the movement trend of the trajectory point, and to extract time features from multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs;

[0137] The second determination module 203 is used to integrate multiple trajectory segments into a trajectory segmentation result according to the movement trend of each trajectory point and the trajectory segment to which it belongs, and classify each trajectory point in the trajectory segmentation result to determine the attribute of the trajectory point;

[0138] The display module 204 is used to segment the agricultural image according to the attribute of each track point, so as to distinguish and display the field area and the road area in the agricultural image.

[0139] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present application. Figure 3 As shown in , the electronic device 300 includes a processor 310 , a memory 320 and a bus 330 .

[0140] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 communicates with the memory 320 via the bus 330. When the machine-readable instructions are executed by the processor 310, the above-mentioned Figure 1 The specific implementation of the steps of the image segmentation method in the method embodiment shown can be found in the method embodiment, and will not be repeated here.

[0141] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The specific implementation of the steps of the image segmentation method in the method embodiment shown can be found in the method embodiment, and will not be repeated here.

[0142] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0143] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0144] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0146] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application can essentially be embodied in the form of a software product, or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0147] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An image segmentation method, characterized in that: include: Acquire trajectory data of an agricultural image, wherein the trajectory data includes a plurality of trajectory points, each of which carries multiple sub-data; For each trajectory point, perform spatial feature extraction on multiple sub-data of the trajectory point to obtain target sub-data of each trajectory point to determine the movement trend of the trajectory point, perform temporal feature extraction on multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs; According to the movement trend of each trajectory point and the trajectory segment to which it belongs, multiple trajectory segments are integrated into a trajectory segmentation result, and each trajectory point in the trajectory segmentation result is classified to determine the attribute of the trajectory point; The agricultural image is segmented according to the attribute of each track point to distinguish and display the field area and the road area in the agricultural image.

2. The method according to claim 1, characterized in that The sub-data includes longitude data, latitude data, timestamp, speed and direction angle, The movement trend of each trajectory point is determined in the following way: For each trajectory point, extract the longitude data and latitude data in the sub-data corresponding to the trajectory point, extract the time in the sub-data corresponding to the trajectory point and the adjacent trajectory point, and determine the time difference between the two adjacent trajectory points; Determine the geographical distance between the two adjacent trajectory points according to the time difference, longitude data and latitude data between the two adjacent trajectory points, and determine the speed difference between the two adjacent trajectory points according to the geographical distance and the time difference; According to the speed difference between two adjacent trajectory points, the motion direction angle of each trajectory point is determined by the inverse tangent function to determine the motion trend of each trajectory point.

3. The method according to claim 2, characterized in that The track segment to which each track point belongs is determined by: For each trajectory point, extract the timestamp in the sub-data corresponding to the trajectory point to determine the time interval between two adjacent trajectory points; Counting the occurrence frequencies of all time intervals, and determining the time interval with the highest occurrence frequency as the first time difference threshold; For each time interval, determine a coefficient corresponding to the time interval, calculate a product of the first time difference threshold and the coefficient corresponding to the time interval, and determine the product as a second time difference threshold for the time interval; For each time interval, determining whether the time interval is less than or equal to the second time difference threshold; For each time interval, if the time interval is less than or equal to the second time difference threshold, the two adjacent trajectory points corresponding to the time interval are divided into the same trajectory segment; if the time interval is greater than the second time difference threshold, a new trajectory segment is created and the two adjacent trajectory points corresponding to the time interval are divided into the new trajectory segment.

4. The method according to claim 3, characterized in that The step of integrating multiple trajectory segments into a trajectory segmentation result according to the motion trend of each trajectory point and the trajectory segment to which it belongs, and classifying each trajectory point in the trajectory segmentation result to determine the attribute of the trajectory point includes: Clustering each trajectory segment and integrating it into a trajectory segmentation result, wherein the trajectory segmentation result includes a plurality of trajectory points arranged in time sequence and sub-data corresponding to each trajectory point; Obtaining a time interval between two adjacent trajectory points in the trajectory segmentation result, and determining whether the time interval between the two adjacent trajectory points is less than or equal to the first time difference threshold; If the time interval between two adjacent trajectory points is less than or equal to the first time difference threshold, it is considered that the attributes of the two adjacent trajectory points corresponding to the time interval are the first attributes; If the time interval between two adjacent trajectory points is greater than the first time difference threshold, it is considered that the attribute of the latter trajectory point of the two adjacent trajectory points corresponding to the time interval is the second attribute.

5. The method according to claim 1, characterized in that The method of displaying the field area and the road area in the agricultural image according to the attribute of each track point comprises: Identify the attribute of each track point, mark the track point with the first attribute as a field track point, and mark the track point with the second attribute as a road track point; Selecting one adjacent field trajectory point and one adjacent road trajectory point from among a plurality of field trajectory points and a plurality of road trajectory points, and calculating the Euclidean distance between the field trajectory point and the road trajectory point; Determine a potential intersection point according to the Euclidean distance, and modify the potential intersection point according to the density and distribution characteristics of the trajectory points around the potential intersection point to determine a final demarcation point; The field track points, the road track points and the final demarcation points are integrated to distinguish and display the field area and the road area in the agricultural image.

6. The method according to claim 5, characterized in that The potential junction points are determined by: Determining a distance difference threshold according to the first time difference threshold; Determine whether the Euclidean distance is less than or equal to a distance difference threshold; If the Euclidean distance is less than or equal to the distance difference threshold, a field track point and a road track point corresponding to the Euclidean distance are determined as potential intersection points; If the Euclidean distance is greater than the distance difference threshold, a field track point and a road track point corresponding to the Euclidean distance are not determined as potential intersection points.

7. The method according to claim 1, characterized in that Obtain trajectory data for agricultural images by: Acquire first data collected by a sensor, wherein the first data includes a plurality of trajectory points, and each trajectory point carries multiple sub-data; Performing formatting processing on the first data, performing unit conversion and standardization processing on multiple sub-data of each trajectory point in the first data, so as to determine second data; For each trajectory point in the second data, calculating the actual speed and the actual direction angle of the trajectory point, and determining whether the actual speed corresponding to the trajectory point is equal to the speed stored in the second data; For each trajectory point in the second data, if the actual speed corresponding to the trajectory point is not equal to the speed stored in the second data, the trajectory point is considered unreasonable and is corrected; For each trajectory point in the second data, if the actual speed corresponding to the trajectory point is equal to the speed stored in the second data, the trajectory point is considered reasonable and saved, and it is determined whether the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data; For each trajectory point in the second data, if the actual direction angle corresponding to the trajectory point is not equal to the direction angle stored in the second data, the trajectory point is considered unreasonable and is corrected; if the actual direction angle corresponding to the trajectory point is equal to the direction angle stored in the second data, the trajectory point is considered reasonable and is saved.

8. An image segmentation device, characterized in that: include: An acquisition module, used for acquiring trajectory data of an agricultural image, wherein the trajectory data includes a plurality of trajectory points, each of which carries a plurality of sub-data; A first determination module is used to extract spatial features from multiple sub-data of each trajectory point to obtain target sub-data of each trajectory point to determine the movement trend of the trajectory point, and to extract time features from multiple sub-data of the trajectory point to obtain a first time difference threshold to determine the trajectory segment to which the trajectory point belongs; A second determination module is used to integrate multiple trajectory segments into a trajectory segmentation result according to the movement trend of each trajectory point and the trajectory segment to which it belongs, and classify each trajectory point in the trajectory segmentation result to determine the attribute of the trajectory point; The display module is used to segment the agricultural image according to the attribute of each track point so as to distinguish and display the field area and the road area in the agricultural image.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and the processor executes the machine-readable instructions to perform the steps of any method as claimed in claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are executed.

Images