Search

CN-121982572-A - Cultivated land feature extraction method, medium and system for satellite monitoring image spots

CN121982572ACN 121982572 ACN121982572 ACN 121982572ACN-121982572-A

Abstract

The invention provides a cultivated land feature extraction method, medium and system of a satellite monitoring pattern spot, which belongs to the technical field of electric digital data processing, the cultivated land feature extraction method of the satellite monitoring pattern spot firstly acquires multi-time phase satellite remote sensing data and carries out pretreatment, and then calculating a vegetation index to construct a time sequence feature matrix, obtaining the change features of the map spots through space-time scale decomposition, establishing a multidimensional feature extraction model, realizing the accurate identification of the map spots, and finally generating a feature distribution matrix through an evaluation model to finish the accurate extraction and characterization of the features of the map spots. According to the invention, the boundary and the internal structure of the cultivated map spot are accurately identified by analyzing the time sequence change characteristics of the vegetation index and combining with the space autocorrelation analysis. The invention improves the accuracy of feature extraction through nuclear density estimation and spatial cluster analysis, and solves the technical problems that the multi-time-phase remote sensing image intertillage map spot feature is difficult to accurately identify and extract in the prior art.

Inventors

  • LIU HUIJIE
  • WANG YU
  • CHEN YULIN
  • LIU QI
  • ZHANG ZHONGYANG
  • LI WEILIN
  • FAN ZILING

Assignees

  • 北京国测星绘信息技术有限公司

Dates

Publication Date
20260505
Application Date
20260130

Claims (10)

  1. 1. A method for extracting cultivated land features of a satellite monitoring pattern spot is characterized by comprising the steps of obtaining multi-temporal satellite remote sensing image data of an area to be analyzed, carrying out atmospheric correction and geometric correction, calculating a vegetation index for first wave band data and second wave band data in the multi-temporal satellite remote sensing image data to generate time sequence feature data, carrying out time scale decomposition and space scale decomposition for the time sequence feature data to obtain a time sequence change feature of the cultivated land spot, establishing a cultivated land vitality calculation model, calculating a cultivated land spot vitality score, establishing a cultivated land intensity calculation model, calculating a planting density score and a growing period score, calculating a cultivated land comprehensive feature score by using the cultivated land vitality score, the planting density score and the growing period score, establishing a cultivated land contribution calculation model and a cultivated land continuous sheet degree calculation model, and establishing a cultivated land protection value scoring model by using the cultivated land contribution score and the cultivated land continuous sheet degree score to generate a cultivated land feature distribution matrix.
  2. 2. The method for extracting the cultivated land features of the satellite monitoring pattern spot according to claim 1, wherein the step of acquiring the multi-temporal satellite remote sensing image data of the area to be analyzed is characterized in that the time interval for acquiring the data is 15 to 30 days, the first wave band data is visible light wave band data, the wavelength range is 0.62 to 0.69 micrometer, the second wave band data is near infrared wave band data, the wavelength range is 0.76 to 0.90 micrometer, the data spatial resolution is better than 30 meters, the radiation resolution is not lower than 12 bits, and the data cloud size is less than 10%.
  3. 3. The method for extracting the cultivated land features of the satellite monitoring pattern spots according to claim 1, wherein the steps of atmosphere correction and geometric correction are carried out, specifically, the steps of atmosphere correction are carried out by adopting a 6S atmosphere correction model, geometric correction is carried out by adopting a quadratic polynomial, the number of control points is not less than 5 times of the four corner points plus a center point of an image, the positioning precision of the control points is better than 0.5 pixel, the root mean square error of the corrected position precision is less than 0.5 pixel, the multi-time phase data registration is carried out by adopting an automatic registration algorithm, and the matching precision is controlled within 0.3 pixel.
  4. 4. The method for extracting cultivated land features of the satellite monitoring pattern spot according to claim 1, wherein the steps of time scale decomposition and space scale decomposition are specifically that a dobesiex 5 wavelet basis function is adopted for time scale decomposition, the decomposition scale is set to be 3 layers, a gaussian pyramid decomposition method is adopted for space scale decomposition, a 3-layer pyramid structure is constructed, and a 7×7 pixel window is adopted for calculating texture features.
  5. 5. The method for extracting cultivated land features of the satellite monitoring pattern spot according to claim 1, wherein the cultivated land activity calculation model comprises a growth potential index, a growth stability index and a growth period index, wherein the weight of the growth potential index is 0.4, the weight of the growth stability index is 0.3, the weight of the growth period index is 0.3, and 4 evaluation grades are set by adopting a trapezoid membership function.
  6. 6. The method for extracting cultivated land features of the satellite monitoring pattern spot according to claim 1, wherein the weight of the planting density score and the growing period score in the cultivated land utilization intensity calculation model is 0.5, the vegetation coverage is extracted by adopting a pixel bipartite model, the vegetation coverage threshold is set to be 0.15, the planting density score and the growing period score are equally divided into 5 grades, and the segmentation points are respectively set to be 0.2, 0.4, 0.6 and 0.8.
  7. 7. The method for extracting cultivated land features of satellite monitoring pattern spots according to claim 1, wherein in the calculation of the cultivated land integrated feature score, the cultivated land spot vitality score weight is 0.4, the planting density score weight is 0.3, the growing period score weight is 0.3, the score is divided into 5 grades, and the thresholds are respectively set to 0.8, 0.6, 0.4 and 0.2.
  8. 8. The method for extracting cultivated land features of the satellite monitoring pattern spots according to claim 1, wherein in the cultivated land protection value scoring model, the score weight of the cultivated land contribution degree is 0.6, the score weight of the cultivated land connection degree is 0.4, the score is unified to a range from 0 to 1 by adopting a range normalization method, at least 10% of typical sample areas are selected for field verification, and the model precision requirement is not lower than 85%.
  9. 9. A computer readable storage medium, wherein program instructions are stored in the computer readable storage medium, which program instructions, when run in a computer, are used to perform a method for extracting cultivated land features for satellite monitoring pattern according to any one of claims 1-8.
  10. 10. The cultivated land feature extraction system for the satellite monitoring pattern spots is characterized by comprising the computer-readable storage medium according to claim 9, wherein the system is any one of a computer, a server and a single chip microcomputer, the computer-readable storage medium is arranged in the system, and a microprocessor for executing program instructions stored in the computer-readable storage medium is arranged in the system.

Description

Cultivated land feature extraction method, medium and system for satellite monitoring image spots Technical Field The invention belongs to the technical field of electric digital data processing, and particularly relates to a cultivated land feature extraction method, medium and system for satellite monitoring image spots. Background Satellite remote sensing technology is widely applied in the field of farmland monitoring, and extraction of the characteristics of the farmland spots by analyzing remote sensing image data is the key point of current research. The traditional pattern recognition technology mainly comprises methods of supervision classification based on spectral features, unsupervised classification based on texture features, target recognition combined with ground object space relations and the like. These methods typically employ single-phase images or simple multi-temporal additive analysis, combined with spectral feature libraries of the earth's surface coverage type, to identify and classify patches. In practical application, common technical means include algorithms such as maximum likelihood classification, support vector machine classification, neural network classification and the like, and the methods can realize identification of the tilling spots to a certain extent. However, conventional patch recognition methods have significant limitations. Firstly, the remote sensing image of a single time phase is difficult to effectively distinguish the ground objects with similar spectral characteristics, and confusion and errors of pattern spot recognition are easy to cause. Secondly, the time sequence change characteristics of crop growth cannot be fully utilized by simple multi-time superposition analysis, so that the pattern spot recognition accuracy is not high. Again, the existing recognition method often ignores the spatial structural features and time sequence change rules of the pattern spots, so that the reliability of the recognition result is affected. In addition, when the traditional method processes multi-temporal remote sensing data with large range and high resolution, the calculation efficiency is low, and the requirement of real-time monitoring is difficult to meet. At present, although research is attempted to improve the pattern recognition accuracy by deep learning and other methods, the methods still have the problems of incomplete feature extraction, insufficient space-time correlation analysis, insufficient recognition process and the like. Particularly in complex surface environments, how to accurately identify and extract the characteristics of the cultivated land spots is still a technical problem to be solved, wherein the characteristics of the cultivated land spots are similar to those of other types of ground features. This severely affects the accuracy and reliability of the monitoring of the cultivated land. In summary, the technical problem in the prior art is that it is difficult to accurately identify and extract the characteristics of the map spots in the multi-phase remote sensing image. Disclosure of Invention In view of the above, the invention provides a method, medium and system for extracting cultivated land features of satellite monitoring image spots, which can solve the technical problem that the characteristics of the cultivated land spots of multi-time-phase remote sensing images are difficult to accurately identify and extract in the prior art. The invention provides a cultivated land feature extraction method of a satellite monitoring pattern spot, which comprises the following steps of obtaining multi-time-phase satellite remote sensing image data of a region to be analyzed, carrying out atmospheric correction and geometric correction, calculating a vegetation index for first wave band data and second wave band data in the multi-time-phase satellite remote sensing image data to generate time sequence feature data, carrying out time scale decomposition and space scale decomposition for the time sequence feature data to obtain a cultivated land spot time sequence change feature, establishing a cultivated land vitality calculation model, calculating cultivated land spot vitality score, establishing a cultivated land strength calculation model, calculating planting density score and growing period score, calculating cultivated land comprehensive feature score by using the cultivated land vitality score, the planting density score and the growing period score, establishing a cultivated land contribution calculation model and a cultivated land continuous quality calculation model, and establishing a cultivated land protection value score model by using the cultivated land contribution score and the cultivated land continuous quality score to generate a cultivated land feature distribution matrix. The method comprises the steps of acquiring multi-temporal satellite remote sensing image data of an area to be analyzed, wherein the time interval for ac