Target detection method and device for high-resolution remote sensing image

12014,536

18/464249

September 10, 2023

The present application discloses a target detection method and device for a high-resolution remote sensing image, which comprises: acquiring an original high-resolution remote sensing image from a sensor; acquiring target information of an area and an expanded area of the area expanding around by a predetermined distance; adaptively partitioning the original high-resolution remote sensing image to obtain different cluster areas to be detected, and obtaining an area required to be detected and an area not required to be detected; selecting a model of the area required to be detected and generating a target detection scheme; executing the target detection scheme to obtain a detection result; determining whether a computing platform has extra computing resources to detect the area not required to be detected, if so, performing dynamic partition detection for the area to obtain a detection result, and merging the detection results into a target detection result.

ZHEJIANG LAB (Zhejiang, CN)

ZHEJIANG LAB (Hangzhou, CN)

1. A target detection method for a high-resolution remote sensing image, wherein the method is applied to a computing platform, and comprises: acquiring an original high-resolution remote sensing image from a sensor; acquiring, from a historical detection result, target information of an area corresponding to the original high-resolution remote sensing image and an expanded area of the area expanding around by a predetermined distance; adaptively partitioning the original high-resolution remote sensing image to obtain different cluster areas to be detected according to the target information, and obtaining an area required to be detected and an area not required to be detected; selecting a model of the area required to be detected, and generating a target detection scheme of the area required to be detected; executing the target detection scheme to obtain a detection result of the area required to be detected; and determining whether the computing platform has extra computing resources to detect the area not required to be detected, when the computing platform has extra computing resources, performing dynamic partition detection for the area not required to be detected to obtain a detection result of the area not required to be detected, and merging the detection result of the area required to be detected and the detection result of the area not required to be detected into a target detection result, when the computing platform does not have extra computing resources, taking the detection result of the area required to be detected as the target detection result.

2. The method according to claim 1 , wherein before the step of acquiring the original high-resolution remote sensing image from the sensor, the method further comprises: off-line training target detection models suitable for different densities and sizes; measuring delay and detection accuracies of the off-line detection models on the computing platform; and loading the off-line detection models and the corresponding delay and detection accuracies on the computing platform.

3. The method according to claim 1 , wherein after the step of obtaining the target detection result, the method further comprises: updating the historical detection result stored on the computing platform according to the target detection result.

4. The method according to claim 1 , wherein the step of adaptively partitioning the original high-resolution remote sensing image to obtain different cluster areas to be detected according to the target information, and obtaining the area required to be detected and the area not required to be detected to be detected comprises: S 21 : performing clustering according to targets and positions of the targets, and calculating an initial value of a number of clustered categories, wherein the target information comprises the positions, sizes and categories of the targets; S 22 : executing a clustering algorithm according to the initial value of the number of the clustered categories to obtain a corresponding number of clusters; S 23 constructing a bounding rectangle for each cluster, and calculating a side length of the rectangle and a density of the targets in the rectangle; S 24 : updating the initial value of the number of the clustered categories and returning to step S 22 for re-clustering when the side length of the rectangle exceeds a side length threshold and the density of the targets in the rectangle exceeds a density threshold; otherwise, selecting a model set capable of processing the density and the size according to the density and the size of the targets; and S 25 : areas where the targets may appear in the clusters by combining motion information of the targets, and setting the areas as cluster areas to be detected, wherein the area required to be detected is composed of all the cluster areas to be detected, an area which is not the area required to be detected in the original high-resolution remote sensing image is the area not required to be detected, and the motion information of the targets is stored on the computing platform.

5. The method according to claim 4 , wherein the step of selecting the model of the area required to be detected and generating the target detection scheme of the area required to be detected comprises: S 31 : selecting the model set capable of processing the density and the size according to the density and the size of the targets; S 32 : evenly dividing one cluster area to be detected according to different input sizes of the models in the model set to generate at least one model combination feasible for the cluster area to be detected; S 33 : executing S 32 for each cluster area to be detected in the area required to be detected, and forming at least one model scheme after combining the models of all cluster areas to be detected; S 34 : calculating an accuracy and a total delay of each model scheme for detecting the area required to be detected according to the accuracy and delay information of each model scheme; and S 35 : selecting a scheme the total delay of the scheme meets a delay limitation of the computing platform among all the model schemes as a target detection scheme of the area required to be detected in a manner of overall accuracy priority.

6. A target detection device for a high-resolution remote sensing image, wherein the device is applied to a computing platform, and comprises: a memory; one or more processors coupled to the memory, the one or more processors being configured to: acquire an original high-resolution remote sensing image from a sensor; acquire, from a historical detection result, target information of an area corresponding to the original high-resolution remote sensing image and an expanded area of the area expanding around by a predetermined distance; adaptively partition the original high-resolution remote sensing image to obtain different cluster areas to be detected according to the target information, and then obtaining an area required to be detected and an area not required to be detected; select a model of the area required to be detected and generating a target detection scheme of the area required to be detected; execute the target detection scheme to obtain a detection result of the area required to be detected; and determine whether a computing platform has extra computing resources to detect the area not required to be detected, when the computing platform has extra computing resources, perform dynamic partition detection for the area not required to be detected to obtain a detection result of the area not required to be detected, and merging the detection result of the area required to be detected and the detection result of the area not required to be detected into a target detection result, when the computing platform does not have extra computing resources, taking the detection result of the area required to be detected as the target detection result.

7. Electronic equipment, comprising: one or more processors; a memory for storing one or more programs; wherein when the one or more programs are executed by the one or more processors, the one or more processors implements the target detection method for a high-resolution remote sensing image according to claim 1 .

8. A non-transitory computer-readable storage medium on which computer instructions are stored, wherein the instructions, when executed by a processor, implements the steps of the target detection method for a high-resolution remote sensing image according to claim 1 .

20130182002 July 2013 Macciola
20190273910 September 2019 Malaika
20220122341 April 2022 Liu
102043958 May 2011
108337399 July 2018
108445480 August 2018
109416728 March 2019
109902627 June 2019
109948415 June 2019
110390292 October 2019
112132093 December 2020
112199984 January 2021
112381053 February 2021
112668390 April 2021
113239815 August 2021
113705532 November 2021
113971653 January 2022
114244904 March 2022
114581781 June 2022

International Search Report (PCT/CN2023/088349); Date of Mailing: Jun. 21, 2023. cited by applicant
First Office Action(CN202210480814.4); Date of Mailing: Jun. 17, 2022. cited by applicant
Notice Of Allowance(CN202210480814.4); Date of Mailing: Jun. 27, 2022. cited by applicant
Moving-object-detection-and-tracking Wei et al. ; Journal of Bohai University, vol. 38, No. 4; Dec. 2017; pp. 370-377. cited by applicant
Aircraft-target-detection-based-on-feature-pyramid-in-high-resolution-remote-sensing-image Zhang et al.; China Academic Journal Electronic Publishing House, Jul. 31, 2018, pp. 1-8. cited by applicant
Wavelet-differencel-reduction-region-of-interest-priority-in-multispectral-video -small target-detection Law et al. ; IEEE International Conference on Image Processing, 2004, pp. 1903-1906. cited by applicant
An-experiment-based-quantitative-and-comparative-analysis-of-target-detection-and-image-classification-algorithms-for-hyperspectral-imagery Chang et al. ; IEEE Transactions on Geoscience and Remote Sensing, vol. 38, No. 2; Mar. 2000; pp. 1044-1063. cited by applicant

edspgr.12014536