TW202125324A - Methods and systems for automatic object detection from aerial imagery - Google Patents

Abstract

Methods and systems for detecting objects from aerial imagery are disclosed. The method includes obtaining an image of an area, obtaining a plurality of regional aerial images from the image of the area, classifying the plurality of regional aerial images as a first class or a second class by a classifier, wherein: the first class indicates a regional aerial image contains a target object, the second class indicates a regional aerial image does not contain a target object, and the classifier is trained by first and second training data, wherein the first training data include first training images containing target objects, and the second training data include second training images containing target objects obtained by adjusting at least one of brightness, contrast, color saturation, resolution, or a rotation angle of the first training images; and recognizing a target object in a regional aerial image in the first class.

Description

Method and system for automatic object detection in aerial images

(Cross reference to related applications)

The present invention hereby expressly incorporates the full text of US Patent No. 15/367,975 filed on December 2, 2016 by reference.

The present invention generally relates to a method and system for detecting objects in an aerial image, and more specifically, relates to an aerial image of an aerial image (an aerial image of a region of interest) that uses artificial intelligence technology to perform template matching (template matching). an area of interest) method and system for objects within.

Automatic object detection is very useful for finding and identifying target objects in the image. Humans can recognize one or several target objects in the image with a little effort, but for humans, it is challenging to find and recognize a large number of target objects in the image. When the target objects in the image are displayed in different sizes and proportions, or even in different rotation perspectives, they look different from different viewpoints. Certain computer-implemented methods can detect target objects based on their appearance or characteristics. However, for certain applications, such as cash crops or certain agricultural applications, the accuracy of these object detection methods may not be good enough.

Therefore, when the number of potential target objects in the region of interest increases and the resolution of aerial images is limited, object detection from aerial images becomes more challenging. When there are a considerable number of potential target objects in a large proportion of the area, it becomes impractical to rely on humans to find and identify the target objects. Increasing the resolution of aerial images may be helpful to improve the accuracy of object detection. However, at the same time, performing object recognition and detection on high-resolution images will increase computational complexity and limit the feasibility of specific applications. Sex and efficiency.

Therefore, there is a need for a method and system for quickly and accurately detecting target objects from aerial images of regions of interest. The disclosed methods and systems aim to overcome or improve one or more of the problems disclosed above and/or other problems in the prior art.

One aspect of the present invention relates to a system for detecting objects from aerial images. The system includes a memory for storing instructions and at least one processor configured to execute the instructions to: obtain an image of a region, obtain a plurality of partial aerial images from the image of the region, and a classifier to obtain the plurality of partial aerial images A partial aerial image is classified into a first type or a second type, where: the first type indicates that a partial aerial image includes a target object, the second type indicates that a partial aerial image does not include a target object, and the classification The device is trained through first and second training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the second training The image includes a target object obtained by adjusting at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the first training image, and identifying a target in a partial aerial image in the first category object.

Another aspect of the present invention relates to a method for detecting an object in an aerial image, the method comprising: obtaining an image of a region, obtaining a plurality of partial aerial images from the image of the region, and using a classifier to obtain the plurality of partial aerial images. A partial aerial image is classified into a first type or a second type, where: the first type indicates that a partial aerial image includes a target object, the second type indicates that a partial aerial image does not include a target object, and the classification The device is trained through first and second training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the second training The image includes a target object obtained by adjusting at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the first training image, and identifying a target in a partial aerial image in the first category object.

It is still another aspect of the present invention, which relates to a non-transitory computer readable medium storing instructions, which when executed will cause one or more processors to perform an operation of detecting objects from aerial images. The operations include: Obtain an image of a region, obtain a plurality of partial aerial images from the image of the region, and classify the plurality of partial aerial images into a first type or a second type by a classifier, wherein: the first type represents A partial aerial image includes a target object, the second type means that a partial aerial image does not include a target object, and the classifier is trained through first and second training data, where the first training data includes the first training Image, the first training image includes a target object, and the second training data includes a second training image, the second training image includes by adjusting the brightness, contrast, color saturation, resolution, or rotation of the first training image A target object obtained by at least one of the angles, and a target object is identified in a partial aerial image in the first category.

The foregoing generally describes only some example aspects of the present invention. It should be understood that the foregoing cover description and the following detailed description are both exemplary and illustrative, and they are not intended to limit the patent scope of the present invention.

The present invention generally relates to methods and systems for detecting objects in aerial images. The intended target object may be a plant, tree, oil palm tree, object, building, facility, land, landform feature, or any combination thereof. Generally speaking, the target object to be detected can include anything, such as objects, buildings, facilities, plants, trees, animals, and even humans. The target object may have many characteristics in color, shape, and/or appearance, and these characteristics of the target object can be used to detect the target object in the image in the region of interest.

The first figure is an exemplary aerial image of an area for automatic object detection according to the disclosed embodiment. For example, oil palm trees are the exemplified target objects to be detected in aerial images of the area, and these oil palm trees have a specific height above the ground. In some embodiments, the disclosed method and system may include detecting the target object based on the height information of the target object in the aerial image of the area. For example, the DSM of a region may include the ground surface and all objects, as well as related height information of the ground surface and all objects. It is expected that the disclosed method and system may include detecting the target object through the height information contained in the region of interest DSM. In some embodiments, the disclosed method and system may include detecting many area models containing height information and/or target objects in the image, such as a digital elevation model (DEM) of the area.

In some embodiments, the disclosed method and system may include the use of one or more light detection and ranging (Light Detection And Ranging, LiDAR) sensors, real-time DSM sensors, and post-production DSM sensors. The calculation of a plurality of aerial images of the area or any combination of these to obtain DSM, DEM and/or aerial images of a region. In some specific embodiments, the disclosed method and system may include the use of one of the aforementioned sensors and/or an unmanned aerial vehicle (Unmanned Aerial Vehicle, UAV) 100 (as shown in Figure 13), an unmanned target Cameras, aircraft, helicopters, balloons or satellites to collect DSM, DEM and/or aerial images of an area. In some embodiments, the disclosed method and system may further include wireless connection, such as Bluetooth, Wi-Fi, cellular (such as GPRS (General Packet Radio Service), WCDMA (Wideband Code) Division Multiple Access, Broadband Code Division Multiple Access), HSPA (High Speed Packet Access), LTE (Long Term Evolution, long-term evolution technology), or a newer generation of cellular communication system), and satellite connection Or a wired connection, such as a USB cable or an optical cable (Lighting line), to receive DSM, DEM and/or aerial image related data from an area of UAV 100, unmanned drone, aircraft, helicopter, balloon or satellite.

In some embodiments, the disclosed method and system may include obtaining the DSM, DEM and/or aerial image of the area used for target detection from a plurality of DSM, DEM and/or aerial images in each part of the area. . For example, the disclosed method and system may include combining or stitching a plurality of aerial images of each part of the area to obtain an aerial image of the area in the first image for object detection. The disclosed method and system include determining an appropriate mathematical model that associates pixel coordinates in one image with pixel coordinates in another image for image alignment. The disclosed method and system may further include a combination of direct pixel-to-pixel comparisons and gradient descent to estimate the correctness associated with many pairs of aerial images. alignment. The disclosed method and system may further include identifying and matching different features in the aerial image of a partial area to establish a correspondence between the aerial image pairing. The disclosed method and system may further include determining the final composite surface, warping or projective transformation and placing all aligned aerial images on it. The disclosed method and system may further include seamless mixing of overlapping aerial images, even in the presence of parallax, lens distortion, scene motion, and exposure differences.

The second figure is a flowchart illustrating an exemplary method 200 for automatic object detection in aerial images according to the disclosed embodiment. One aspect of the present invention relates to a non-transitory computer readable medium storing instructions, wherein when the instructions are executed, one or more processors will execute the method 200 illustrated in the second figure to detect aerial images Objects in. The computer-readable medium may include volatile or non-volatile, magnetic, semiconductor, magnetic tape, optical, removable, non-removable, or other kinds of computer-readable media or computer-readable storage devices. For example, the computer-readable medium can be a storage unit or a memory module storing the computer commands in it, as disclosed. In some embodiments, the computer-readable medium can be an optical disc or a flash drive in which the computer commands are stored. In some embodiments, the computer-readable medium can be a cloud or remote storage device storing the computer commands, and these commands can be downloaded to another device for execution.

The method 200 may include the following steps: obtaining a DSM image of a region (step 220), obtaining a DSM image of a target object (step 240), and detecting the DSM image based on the region and the target object in steps 220 and 240 The target object in the area (step 260). Please note that the DSM of a region contains the altitude information of the region. By using the height information of the area as the grayscale value of the grayscale image of the area, the DSM image of the area can be obtained, and vice versa. Therefore, if appropriate, "DSM" and "DSM image" can be used interchangeably throughout the present invention.

Step 220 may include obtaining a DSM image of the region of interest. For example, the step 220 of obtaining a DSM image of a region may include accessing the DSM image of the region of interest from a computer-readable medium or a computer-readable storage device. For another example, the step 220 of obtaining a DSM image of a region may include an external input, such as the image input 120 (which will be described in the disclosed system), to receive the DSM image of the region of interest. The image input 120 can be communicatively connected to, for example, UAV 100, unmanned drone, aircraft, helicopter, balloon, or satellite. In other words, the step 220 of obtaining a DSM image of an area may include receiving the DSM image of the area of interest from the UAV 100, an unmanned target drone, an aircraft, a helicopter, a balloon, or a satellite. In some embodiments, the step 220 of obtaining a DSM image of a region may include obtaining a plurality of DSM images of each part of the region, and combining or stitching the plurality of DSM images of each part of the region to obtain a feeling. DSM image of the area of interest. For example, the step 220 of obtaining a DSM image of a region may include obtaining a plurality of DSM images of each part of the region, and identifying and matching different features of the plurality of DSM images of each part of the region to establish a correspondence between DSM image pairs relation. The step 220 of obtaining a DSM image of a region may further include mixing a plurality of DSM images of each part of the region according to the corresponding relationship between the established DSM image pairs to obtain a DSM image of the region of interest.

In some embodiments, the step 220 of obtaining a DSM image of an area may include obtaining a plurality of aerial images of an area, combining or stitching these aerial images of various parts of the area to obtain an aerial image of the area. And convert the stitching aerial image of the area into a DSM image of the area. For example, the step 220 of obtaining a DSM image of an area may include receiving a plurality of aerial images of each part of an area, and stitching the plurality of aerial images of each part of the area to obtain the image of the area shown in the first figure. Aerial image. These aerial images of each part of the area can be associated with a plurality of DSMs of each part of the area. In other words, a plurality of aerial images of each part of the area may correspond to a plurality of DSMs of each part of the area. Step 220 may include obtaining the DSM image of the area in the third image corresponding to the stitching aerial image of the area in the first image according to the correspondence between the aerial images and the DSM of each part of the area. The third diagram illustrates the exemplary aerial image corresponding to the area in the first diagram according to the disclosed specific embodiment, and an exemplary DSM image of the area used for automatic object detection.

In some embodiments, the step 220 of obtaining DSM images of the area may include using one or more LiDAR sensors, real-time DSM sensors, post-production DSM sensors, and calculation of a plurality of aerial images of the area. Or any combination of these to collect DSM and/or aerial images of the area or parts of the area. In some embodiments, the step 220 of obtaining DSM images of the area may include collecting by using one of the above-mentioned sensors and/or through UAV 100, unmanned target drone, aircraft, helicopter, balloon, or satellite camera. DSM and/or aerial images of an area or parts of the area. In some embodiments, the step 220 of obtaining DSM images of the area may further include wireless connection, such as Bluetooth, Wi-Fi, cellular (such as GPRS, WCDMA, HSPA, LTE or newer generation cellular communication). System) and satellite connection or wired connection, such as USB cable or fiber optic cable, to receive the collected data of DSM and/or aerial images in the area from UAV 100, unmanned drone, aircraft, helicopter, balloon or satellite.

In some embodiments, the step 220 of obtaining a DSM image of the area may further include obtaining a color aerial image of the area corresponding to the DSM image of the area, obtaining a color aerial image of a target object, and obtaining a color aerial image of the area according to the area. The image and the color of the target object are used to identify one or more sub-areas of the area as one or more target sub-areas.

For example, the step 220 of obtaining a DSM image of a region may further include obtaining the RGB aerial image of the region in the first image corresponding to the DSM image of the region of interest in the third image. In addition, the step 220 of obtaining a DSM image of a region may further include obtaining an RGB aerial image of an oil palm tree (the target object). Furthermore, the step 220 of obtaining a DSM image of a region may further include identifying green as the specific primary color of the oil palm tree. Furthermore, the step 220 of obtaining a DSM image of a region may further include when the individual G values of the pixels in the aerial image of the region are greater than the individual R and B values, identifying these pixels as possible pixels of the oil palm tree. For example, the following conditional operation can be used to check whether a pixel is recognized as a possible pixel of oil palm: "If (Pixel.G ＞ Pixel.R && Pixel.G ＞ Pixel.B) Get Pixel", where Pixel. R, Pixel . G and Pixel. B represents the individual R, G, and B levels of the pixel. Furthermore, the step 220 of obtaining a DSM image of a region may further include identifying a certain number of possible pixels adjacent to the oil palm tree as a target subregion.

In some embodiments, the step 220 of obtaining a DSM image of a region may further include identifying a specific primary color of the target object. For example, the step 220 of identifying a specific primary color of the target object may include comparing individual R, G, and B levels within the pixels of the aerial image of the target object, and determining the representative primary color of these pixels. In addition, the step 220 of identifying a specific primary color of the target object may further include calculating the number of representative primary colors of the pixels, and identifying the maximum number of representative primary colors of the pixels as the specific primary color of the target object. For example, the step 220 of identifying a specific primary color of the target object may include: when green is the representative primary color with the largest number of pixels in the aerial image of the oil palm tree, identifying the green color as the specific primary color of the oil palm tree .

In some embodiments, the step 220 of obtaining a DSM image of a region may further include enhancing the image contrast of one or more target sub-regions on the DSM image of the region. For example, the step 220 of enhancing the contrast of the target subregion may include using histogram equalization to enhance the contrast of the target subregion of the DSM image of the region corresponding to the identified target subregion in the aerial image of the region. For example, the step 220 of enhancing the contrast by using the histogram equalization may include calculating the probability quality function of the target sub-region pixels, calculating the cumulative distribution function (CDF, cumulative distributive function) value according to the gray scale, and multiplying by (gray scale -1) The CDF value is increased, and the new grayscale value is mapped to the pixels of the target sub-regions. The step 220 of enhancing the contrast may include: enhancing the contrast through other algorithms, such as global stretching, anisotropic diffusion, non-linear cone technology, multi-scale morphology technology, multi-resolution splines (multi-resolution splines) ), mountain clustering, retinex theory, wavelet transformations, curvelet transformations, k-sigma clipping, fuzzy logic, genetic algorithms, or greedy Algorithm.

Step 240 may include obtaining a DSM image of the target object. The fourth figure shows two exemplary DSM image patterns of an exemplary target object type used for automatic object detection according to the disclosed embodiment. For example, the step 240 of obtaining a DSM image of a target object may include accessing the DSM image of the oil palm from the computer-readable medium or computer-readable storage device in the fourth image. For another example, the step 240 of obtaining a DSM image of a target object may include an external input, such as image input 120 (which will be described in the disclosed system), and receiving the DSM image of the oil palm tree in Figure 4 (a) . For another example, the step 240 of obtaining a DSM image of a target object may include an internal input, such as the image input 120, to receive a selection signal. The selection signal may include a DSM image that recognizes a part of the DSM image of the area in step 220 as a target object. For example, the selection signal may include identifying an area surrounding the DSM image of the oil palm tree on the DSM image of the area as the target object, when the user uses a mouse cursor, his finger or a pen to select the area on the display screen.

In some embodiments, the step 240 of obtaining a DSM image of a target object may include accessing or receiving a plurality of DSM images of the target object, and selecting one of them as the DSM image of the target object. The step 240 of selecting the DSM image of the target object may include selecting the DSM image of the target object according to the shape of the target object. For example, the step 240 of selecting the DSM image of the target object may include selecting the DSM image of the target object whose shape is similar to most of the same target object. In some embodiments, the step 240 of selecting the DSM image of the target object may include selecting the DSM image of the target object based on the comparison of the DSM image of the target object. For example, the step 240 of selecting the DSM image of the target object may include selecting the DSM image of the target object with better contrast than other objects. In some embodiments, the step 240 of obtaining a DSM image of the target object may include obtaining more than one DSM image of the target object. For example, the step 240 of obtaining the DSM image of the target object may include obtaining two DSM images of the target object according to the shape of the target object and the comparison of the DSM image of the target object.

In some embodiments, the step 240 of obtaining DSM images of the target object may include using one or more LiDAR sensors, real-time DSM sensors, post-production DSM sensors, and multiple aerial images of the area. Calculate or any combination of these to collect one or more DSM and/or aerial images of the target object. In some embodiments, the step 240 of obtaining a DSM image of the target object may further include using one of the aforementioned sensors and/or a camera through UAV 100, unmanned drone, aircraft, helicopter, balloon or satellite, Collect one or more DSM and/or aerial images of the target object. In some embodiments, the step 240 of obtaining the DSM image of the target object may further include wireless connection, such as Bluetooth, Wi-Fi, cellular (such as GPRS, WCDMA, HSPA, LTE or newer generation cellular Communication system) and satellite connection or wired connection, such as USB cable or optical fiber cable, to receive DSM and/or aerial image of the target object from UAV 100, unmanned drone, aircraft, helicopter, balloon or satellite.

In some embodiments, the step 240 of obtaining the DSM image of the target object may include obtaining one or more aerial images of the target object corresponding to the one or more DSM images of the target object, and according to the shape of the target object And/or the comparison of aerial images of the target object, select one or more DSM images of the target object.

Step 260 may include detecting the target object in the area according to the DSM image of the area and the target object in steps 220 and 240. In some embodiments, the step 260 of detecting the target object may include calculating the matching rate between the DSM image of the target object and the plurality of DSM sub-images of the region, and determining one or the ratio of the region according to the matching rate. Multiple DSM sub-images are regarded as the target object. For example, the step 260 of detecting the target object may include calculating the pairing between the DSM image of an oil palm tree in the fourth (a) image and the plurality of DSM sub-images from the area of the DSM image of the area in the third image Rate. The plurality of DSM sub-images of the area may have the same or similar size as the DSM image of the oil palm tree. For example, the size of the plurality of DSM sub-images in the area may be 300×300 pixels, and the DSM image of the oil palm tree in the fourth (a) figure may be 300×300 pixels or the like. For example, the plurality of DSM sub-images in the area may include sub-images of 300×300 pixels for every 1, 2, 5, or 10 pixels of the DSM image in the area. In other words, the step 260 of detecting the target object may include comparing the template DSM image of the oil palm tree (T ) with the DSM image of the area ( I ) every time 1, 2, 5, or 10 pixels are slid. For example, for each position (x, y ) of the sliding on the DSM image of the region, the pairing rate R can be calculated as follows:

其中x’ 和y’ 代表該油棕櫚樹(T’ )的該範本DSM影像以及該區域(I’ )的該DSM子影像之內之像素位置。

Wherein x 'and y' representing the oil palm (T ') of the image and the template region DSM (I' DSM pixel positions within the sub-images) of the.

The fifth diagram is a diagram of an exemplified image of the pairing rate between the DSM image in the third diagram used for automatic object detection and the pairing ratio of the present image in the fourth diagram, according to the disclosed specific embodiment. Mode. In the fifth figure, the brighter a location is, the higher the possibility that the location is the target object. For example, the bright spot on the image of the matching rate in the fifth image may be the position of the oil palm tree in the region of interest.

In some specific embodiments, the step 260 of calculating the pairing rate may include the pairing method according to the normal specification, such as the square difference method, the normalized square difference method, the cross-correlation method, the normalized cross-correlation method, and the correlation coefficient method. , Normalized correlation coefficient method or any combination thereof to calculate the matching rate.

In some embodiments, the step 260 of determining the DSM sub-image of the region as the target object may include when the matching rate Rs of the template image with the target object is higher than a pairing threshold, such as the oil palm ( When the template DSM image of T) is 80%, 70%, or 60% of the matching rate, one or more DSM sub-images in the region are determined to be the oil palm trees.

In some embodiments, the step 260 of detecting the target object may include reducing the resolution of the DSM image of the region and the target object in steps 220 and 240, and reducing the resolution of the region and the target object. DSM image to detect the target object in the area. For example, the step 260 of detecting the target object may include reducing the resolution of the DSM image of the oil palm tree in the third image and the fourth (a) image to 0.1 times the original resolution. The step 260 of detecting the target object may further include calculating the matching rate between the DSM image with reduced resolution of the oil palm tree and the plurality of DSM sub-images with reduced resolution in the region, and determining the matching rate of the region according to the matching rate One or more DSM sub-images are regarded as the target object.

In some embodiments, the step 260 of detecting the target object may include detecting the target object in the area based on the DSM image of the area in step 220 and one or more images of the target object in step 240. For example, the step 260 of detecting the target object may include calculating the matching ratios between the two DSM images of the oil palm trees in the fourth (a) and fourth (b) images and the plurality of DSM sub-images in the region, respectively. , And based on the pairing rates of the two DSM images from the oil palm trees, one or more DSM sub-images in the area are determined as the oil palm trees. For example, the step 260 of detecting the target object may include calculating the matching ratio between the DSM image of an oil palm tree selected according to the shape of the target object in step 240 and the plurality of DSM sub-images of the region. The step 260 of detecting the target object may also include calculating the matching ratio between another DSM image of an oil palm tree selected according to the image comparison in step 240 and a plurality of DSM sub-images of the region. The step 260 of detecting the target object may further include when the pairing rate of the template DSM image from an oil palm tree selected according to the shape of the target object or the image comparison is higher than a pairing threshold, a portion of the area Or multiple DSM sub-images are determined to be the oil palm tree. For another example, the step 260 of determining the oil palm trees may include when the template DSM image from an oil palm tree selected according to the shape of the oil palm tree and the oil palm tree image is compared, the pairing rate of the template DSM image is higher than that of a pairing. When the threshold is set, one or more DSM sub-images in the area are determined to be the oil palm tree.

In some embodiments, the step 260 of determining the DSM sub-image of the region as the target object includes determining the target object according to one or both of the following two criteria. Within a distance (D1 ) on the aerial image of the area, the pairing rate of one or more DSM sub-images in the area is the largest. The height of the one or more DSM sub-images in the area is higher than the height of the lowest position within another distance (D 2 ) by a height threshold (H 1 ). For example, the step 260 of determining the oil palm trees may include when the pairing rate is within 2 meters (ie D 1 = 2 meters) of the other pairing rates, and the radius of the aerial image of an oil palm tree is One or more DSM sub-images are determined to be the oil palm tree. For another example, the step 260 of determining the oil palm trees may include using an exemplary height threshold of 2.5 meters (ie H 1 = 2.5 meters), when the height is within 3 meters above the lowest position height (ie D ₂ = 3 meters), which is the exemplified radius of a single area where an oil palm tree and the land coexist, then one or more DSM sub-images of the area are determined to be the oil palm tree. According to the aforementioned D ₁ , D ₂ and H ₁ parameters, oil palm trees 2.5 meters above the ground can be detected. According to its height and distribution, these factors can be adjusted for many target objects.

In some embodiments, step 260 may include: detecting the target object in the area according to the enhanced DSM image of the area in step 220 and the DSM image of the target object. For example, the step 260 of detecting the target object may include one or two DSM images of the oil palm tree in the fourth image, and the contrast-enhanced DSM image of the area whose target sub-region has been identified and contrast-enhanced in step 220 , To detect the oil palm tree in the area.

In some embodiments, the method 200 may further include obtaining one or more positions of the target objects detected in step 260. For example, obtaining the positions of the target objects may include obtaining the positions of the oil palm tree detected on the DSM image of the area in the third image. For another example, the step 290 of obtaining the positions of the target objects may include obtaining the oil palms detected on the aerial image of the region according to the correspondence between the DSM image of the region and the aerial image of the region. These positions of the tree. The sixth figure is based on the disclosed specific embodiment, marking the region exemplified aerial image pattern of the positions of the exemplified target objects detected according to the exemplified method for automatic object detection in the second figure. In the sixth figure, the detected oil palm trees are drawn in red circles in the aerial image of the area.

In some embodiments, step 290 may further include displaying the location of the detected target object on the aerial image or map of the area. For example, the step 290 of displaying the detected objects may include displaying the location of the one or more detected oil palm trees in the aerial image of the area, as shown by the red circle in the sixth figure. For another example, the step 290 of displaying the detected target objects may include based on the correlation or correspondence between the locations on the DSM image of the area and the locations on the map (not shown) of the area, in the The location of the one or more detected oil palm trees is displayed on the map of the area. For example, the previous position of the DSM image of the area can be associated with a set of longitude, latitude, and altitude. Step 290 may include obtaining a combination of longitude, latitude, and altitude of the detected oil palm trees, and displaying the detected oil palm trees on a map according to the combination of longitude, latitude, and/or altitude. For example, the step 290 of displaying the detected oil palm trees may include displaying the detected oil palm trees on a Geographic Information System (GIS) map based on a combination of longitude and latitude. For another example, the step 290 of displaying the detected oil palm trees may include displaying the detected oil palm trees on a map based on the combination of longitude, latitude and altitude, such as a 3D GIS map .

In some embodiments, step 290 may further include calculating the number of detected target objects. For example, the step 290 of calculating the detected target objects may include calculating the detected oil palm trees shown in the sixth figure.

The seventh figure is a flowchart illustrating another exemplary method 700 for automatic object detection in aerial images according to the disclosed embodiment. The method 700 may include steps 220, 240, and 260, and may further include: obtaining an aerial image of the area corresponding to the DSM image of the area (step 710); obtaining one of the detected target objects on the aerial image of the area Or multiple locations (step 720); obtain one or more local aerial images at or around one or more locations of the detected target objects (step 730); capture from the one or more local aerial images Take one or more texture features as one or more feature vectors (step 740); obtain a plurality of training data (step 750); train a classifier based on the plurality of training data (step 760); One or more feature vectors are used to classify the one or more partial aerial images by the trained classifier (step 770); and based on the classification results, the one or more partial aerial images are identified Wait for the target object (step 780). The training data may include a plurality of aerial images of the same object as the target objects.

Step 710 may include acquiring an aerial image of the area, which corresponds to the DSM image of the area in step 220. For example, step 710 may include acquiring the aerial image of the region of interest in the first image, which corresponds to the DSM image of the region of interest in the third image. For example, the step 710 of obtaining an aerial image of the area may include accessing the aerial image of the region of interest from a computer readable medium or computer readable storage device. For another example, the step 710 of obtaining an aerial image of the area may include an external input, such as an image input 120 (which will be described in the disclosed system), to receive the DSM image of the area. In some embodiments, the step 710 of obtaining aerial images of the area may include obtaining a plurality of aerial images of each part of the area, and combining or stitching the plurality of aerial images of each part of the area to obtain Obtain aerial images of the area. For example, the step 710 of obtaining an aerial image of the area may include obtaining a plurality of aerial images of each part of the area in the first image, and stitching the plurality of aerial images of each part of the area to obtain the interest Aerial imagery of the area.

In some embodiments, the step 710 of obtaining the aerial image of the area may include obtaining the aerial image of the area in many color spaces. For example, the step 710 of obtaining the aerial image of the area may include obtaining at least RGB (Red-Green-Blue, red-green-blue three primary colors), grayscale, and HIS (Hue-Saturation-Intensity, color direction-saturation- Intensity), L*a*b, multispectral space, or any combination of these areas in the color space of the aerial image.

In some embodiments, the step 710 of obtaining aerial images of the area may include using one or more LiDAR sensors, real-time DSM sensors, post-production DSM sensors, and calculation of multiple aerial images of the area Or any combination of these to collect aerial images of the area or parts of the area. In some embodiments, the step 710 of obtaining aerial images of the area may include using one of the above-mentioned sensors and/or through the camera of the UAV 100, unmanned target drone, aircraft, helicopter, balloon, or satellite, to collect the An aerial image of an area or parts of the area. In some embodiments, the step 710 of obtaining aerial images of the area may further include wireless connection, such as Bluetooth, Wi-Fi, cellular (such as GPRS, WCDMA, HSPA, LTE or newer generation cellular communication) System), and satellite or wired connection, such as USB cable or fiber optic cable, to receive the collected data of aerial images in the area from UAV 100, unmanned drone, aircraft, helicopter, balloon or satellite.

Step 720 may include obtaining one or more positions of the detected target objects in step 260 on the aerial image of the area. For example, the step 720 of obtaining the positions of the detected target objects may include obtaining the detected DSM images on the DSM image of the area in the third image according to the correspondence between the DSM image of the area and the aerial image of the area. The locations of oil palm trees, and the locations of the detected oil palm trees on the aerial image of the area in the first image. In other words, the step 720 of obtaining the position of the detected target object may include obtaining the red circle in the sixth figure, that is, the position of the detected oil palm trees.

Step 730 may include acquiring one or more local aerial images at or around one or more locations of the detected target objects. The eighth figure is based on the disclosed specific embodiment, in accordance with the exemplified method used for automatic object detection in the second figure, marking the area where the positions of the exemplified target objects have been detected and exemplifying a partially enlarged image of the aerial image . For example, the step 730 of obtaining the partial aerial images may include obtaining the 300x300 on the detected oil palm trees 801, 802, 803 in the eighth image from the aerial image of the area in step 710 according to the position obtained in step 720. Partial aerial image. For example, the step 730 of obtaining the partial aerial images may include taking the detected position in step 720 as the center of the 300x300 partial aerial images, and obtaining the 300x300 partial aerial images of the detected oil palm trees. For another example, the step 730 of obtaining the partial aerial images may include taking the detected position in step 720 as the center of the circle, and obtaining the circular partial aerial image on the detected oil palm tree. The radius of the circular aerial image of the detected oil palm tree may include, for example, 150 pixels. The shapes of the partial aerial images of the detected target objects may include other shapes, such as rectangles, triangles, or other shapes similar to the shapes of the target objects.

In some embodiments, the step 730 of obtaining the local aerial images may include using the positions of the detected target objects as the origin to establish one or more coordinates, and obtaining the coordinates around the origin. One or more 300x300 partial aerial images. These coordinates can be used to represent or refer to the acquired partial aerial images.

In some embodiments, the step 730 of obtaining the partial aerial images may include obtaining one or more partial aerial images at one or more positions of the detected target objects in many color spaces. For example, the step 730 of acquiring the partial aerial images may include acquiring the detected oil palm trees in a color space such as RGB, grayscale, HSI, L*a*b, multispectral space, or any combination of these. One or more 300x300 partial aerial images. For example, the step 730 of obtaining the partial aerial images may include obtaining the partial aerial images of the detected oil palm trees in the aforementioned color spaces from the aerial images of the region in the aforementioned color spaces in step 710. In some embodiments, the step 730 of obtaining the partial aerial images may include: obtaining one or more partial aerial images of the detected target objects in a color space, and obtaining the partial aerial images in the color space. One or more partial aerial images of the detected target object are converted into relative parts in another color space. For example, the step 730 of obtaining the partial aerial images may include obtaining one or more RGB partial aerial images of the detected oil palm trees, and converting these images into gray-scale relative parts. For another example, the step 730 of obtaining the partial aerial images may include obtaining one or more RGB partial aerial images of the detected oil palm trees, and converting these images into HSI relative parts.

Step 740 may include extracting one or more texture features from the one or more local aerial images, and using them as one or more feature vectors of the detected target objects in step 260. For example, the step 740 of retrieving the texture features may include according to Gabor filter, gray-level co-occurrence matrix (Gray-Level Co-occurrence Matrix, GLCM), local binary pattern (Local Binary Pattern, LBP), directional gradient histogram (Histograms of Oriented Gradients, HOG), first-level feature description, second-level feature description, or any combination thereof, to capture the one or more texture features. The step 740 of extracting features may include using the aforementioned method to extract the information and non-redundant features of the partial aerial images to help the subsequent classification in step 770.

In some embodiments, the step 740 of extracting one or more texture features may include at least one of the one or more partial aerial images in one color space, and/or the steps in another color space. One or more partial aerial images are captured, and the one or more texture features are captured. For example, the step 740 of retrieving one or more texture features may include the step of extracting the detected oil palm tree in gray scale according to the Multi-block Local Binary Pattern (MB-LBP) The one or more texture features are captured from one or more partial aerial images. For another example, the step 740 of extracting one or more texture features may include extracting the one or more local aerial images of the detected oil palm tree presented in RGB according to the Gabor filter. Multiple texture features. For another example, the step 740 of retrieving one or more texture features may include using multi-block local binary mode (MB-LBP) from the detected oil palm tree in both grayscale and RGB The one or more texture features are captured from one or more partial aerial images. For another example, the step 740 of capturing one or more texture features may include capturing the one or more partial aerial images of the detected oil palm tree in grayscale according to GLCM. Texture features, and according to HOG, the one or more texture features are extracted from one or more partial aerial images of the detected oil palm tree presented in L*a*b.

Step 750 may include obtaining a plurality of training materials. The training data may include a plurality of aerial images of similar objects equivalent to the target objects. The ninth diagram is a diagram of a plurality of exemplified training data that can be used to train the exemplified classifier for automatic object detection according to the disclosed specific embodiment. For example, step 750 may include obtaining a plurality of aerial images of oil palm trees, as shown in figure 9(a), as the training data. In some embodiments, step 750 may further include obtaining a plurality of aerial images of non-target objects, as shown in Figure 9(b), as part of the training data. For example, the step 750 of obtaining the training data may include accessing the training data from a computer-readable medium or a computer-readable storage device. For another example, the step 750 of obtaining the training data may include receiving the training data from an external input, such as the image input 120 (which will be described in the disclosed system).

Step 760 may include training a classifier based on the plurality of training data in step 750. The classifier uses pattern pairing to determine the closest pairing function, which can be adjusted according to training data. The training data may include observations or patterns. For example, in supervised learning, each pattern belongs to a specific predetermined level. Level can be regarded as a decision to be made. All observations combined with the grade label of the observations are called data sets. When a new observation has been received, the observation is classified based on previous experience. For example, the step 760 of training the classifier may include training at least one Support Vector Machine (SVM) classifier and artificial neural network (Artificial Neural Network) using the training data of the oil palm and non-target objects in the ninth image. Network, ANN) classifier, decision tree classifier, Bayes classifier, or any combination thereof. For another example, the step 760 of training the classifier may include training at least one support vector machine (SVM) classifier, artificial Neural network (ANN) classifier, decision tree classifier, Bayes classifier, or any combination thereof.

Step 770 may include using the trained classifier in step 760 to classify the one or more partial aerial images in step 730 according to the one or more feature vectors in step 740. For example, the step 770 of classifying the partial aerial images may include using the trained SVM classifier in step 760 to classify according to the one or more feature vectors extracted by the Gabor filter and GLCM in step 740 In step 730, the one or more partial aerial images of the detected oil palm tree. For another example, the step 770 of classifying the partial aerial images may include using the trained ANN classifier in step 760 to classify according to the one or more feature vectors extracted by LBP and HOG in step 740 In step 730, the one or more partial aerial images of the detected oil palm tree. For another example, the step 770 of classifying the local aerial images may include using the trained one or more feature vectors from the Gabor filter, GLCM, LBP, and HOG in step 740. The ANN classifier is used to classify the one or more partial aerial images of the detected oil palm tree in step 730. The method 700 may include any combination of the aforementioned texture extraction algorithm and the classifiers.

The classification results of the classifier may include two types of results or multiple types of results. For example, when the detected oil palm tree is classified into the same type of object in the ninth (a) image according to the feature vector of a local aerial image, an SVM classifier can output "0". When the detected oil palm tree is classified into the same type of object in the ninth (b) image according to the feature vector of a local aerial image, an SVM classifier can output "1". The tenth figure is based on the disclosed specific embodiment, in accordance with the exemplified method for automatic object detection in the seventh figure, the area where the classification result is marked on the positions of the detected target objects exemplifies the aerial image Partially enlarged diagram. The partial aerial images marked with pink circles at positions 1001, 1002, and 1003 are classified as the target oil palm trees. The partial aerial images marked with blue circles at positions 1016, 1017, and 1018 are classified as the non-target objects.

Step 780 may include identifying the target objects between the one or more partial aerial images based on the classification results. For example, the step 780 of identifying the target objects may include identifying the oil palm among the one or more partial aerial images of the detected oil palm trees in step 730 according to the classification result in step 770 Tree. For example, the local aerial images of the detected oil palm trees 1001, 1002, 1003 in the tenth image can be classified as the same objects as those in the ninth (a) image, and the output from the SVM classifier for these are all " 0". Therefore, the step 780 of recognizing the target objects may include recognizing the partial aerial images of the detected oil palm trees 1001, 1002, 1003 as the target oil palm trees according to the classification result "0". For example, the local aerial images of oil palm trees 1016, 1017, and 1018 that have been detected in the tenth image can be classified as the same objects as those in the ninth (b) image, and the output from the SVM classifier for these are all " 1". Therefore, the step 780 of recognizing the target objects may include recognizing the local aerial images of the detected oil palm trees 1016, 1017, and 1018 as the non-target objects according to the classification result "1".

In some embodiments, the method 700 may further include a step 790 of obtaining one or more positions of the identified target objects in step 780. For example, the step 790 of obtaining the positions of the recognized target objects may include obtaining the positions of the recognized oil palm trees 1001, 1002, 1003 on the aerial image of the area. In the tenth image, the identified oil palm trees are marked with pink circles in the aerial image of the area, and the identified non-target objects are marked with blue circles in the aerial image of the area in the figure. The step 790 of obtaining the positions of the recognized target objects may include obtaining the positions of the recognized oil palm trees marked with pink circles in the aerial image of the area.

In some embodiments, step 790 may further include displaying one or more positions of the identified target object on an aerial image or map of the area. For example, the step 790 of displaying the identified target objects may include displaying the location of one or more identified oil palm trees 1001, 1002, 1003 on the aerial image of the area. For another example, the step 790 of displaying the identified target objects may include the correlation or correspondence between the locations on the aerial image of the area and the locations on the map (not shown) of the area, in accordance with The map of the area shows the location of the one or more identified oil palm trees. For example, the previous position of the aerial image of the area may be associated with a set of longitude, latitude, and altitude. In some embodiments, the step 790 of displaying the identified target objects may include obtaining a combination of the longitude, latitude, and altitude of the identified oil palm tree, and according to the longitude, latitude, and/or altitude. Combine, display the identified oil palm trees on the map. For example, the step 790 of displaying the identified oil palm trees may include displaying the identified oil palm trees on a geographic information system (GIS) map based on a combination of longitude and latitude. For another example, the step 790 of displaying the identified oil palm trees may include displaying the identified oil palm trees on a map, such as a 3D GIS map, according to the combination of longitude, latitude, and altitude.

In some embodiments, step 790 may include counting the number of the identified target objects. For example, step 790 may include calculating the identified oil palm trees.

The eleventh figure is based on the disclosed specific embodiment, in accordance with the exemplified method used for automatic object detection in the seventh figure, marking the regions of the exemplified target object positions that have been correctly detected and recognized, exemplifying the aerial image Partially enlarged diagram. When the ground truth information of the target objects is available, the accuracy of the above object detection method can be evaluated. The white circles 1101, 1102, 1103 in the eleventh figure are examples of oil palm trees that have been correctly detected and identified.

The twelfth figure is based on the disclosed specific embodiment, in accordance with the instantiation method for automatic object detection in the seventh figure, marking the region of the classification result on the positions of the detected and classified instantiation target objects. Aerial image schema. In the diagram, the partial aerial images marked with a pink circle are identified as the target oil palm trees, and the partial aerial images marked with a blue circle are classified as the non-target objects. In a specific embodiment, when MB-LBP is used for feature extraction and the ground sampling distance of the image is 3 cm, the accuracy and recall rates of object detection from aerial images can reach 90.6%, respectively. And 83.4%.

Another aspect of the present invention involves the use of one or more integrated circuits, one or more field programmable gate arrays, one or more processors or controllers that execute instructions to implement the method, or any combination of the above, to execute aerial imagery The method of object detection. The method may include, but is not limited to, all the above methods and specific embodiments. In some specific embodiments, part of the steps in the foregoing methods or specific embodiments can be performed remotely or separately. In some embodiments, the method can be executed by one or more distributed systems.

Still another aspect of the present invention relates to a system for detecting objects in aerial images. Figure 13 is a block diagram illustrating an exemplary system 400 for automatic object detection in aerial images according to the disclosed specific embodiment. The automatic object detection system 400 may include an aerial image unit 410 configured to obtain a DSM image of a region, a target image unit 420 configured to obtain a DSM image of a target object, and a DSM image configured according to the region and the target object To detect a detection unit 430 of the target object in the area.

The aerial image unit 410 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (that is, a Function program) to perform the operations in step 220 above. The aerial image unit 410 may be configured to obtain a DSM image of a region. In some embodiments, the aerial image unit 410 can be communicatively connected to an image input 120. The image input 120 can provide many of the above-mentioned image inputs to the aerial image unit 410. For example, the image input 120 may receive aerial images of the area, DSM and/or DEM of the area from UAV 100, unmanned target drones, aircraft, helicopters, balloons or satellites, and these images, DSMs and DSMs of the area /Or the DEM is transmitted to the aerial image unit 410. In some embodiments, the aerial image unit 410 can also be communicatively connected to a detection unit 430. The aerial image unit 410 may be configured to provide the DSM image of the area and the aerial image of the area or parts of the area to the detection unit 430. In some embodiments, the aerial image unit 410 can also be communicatively connected to a target image unit 420. The aerial image unit 410 may be configured to transmit the received target images from the image input 120 to the target image unit 420.

The target image unit 420 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (ie, a Function program) to perform the operations in step 240 above. The target image unit 420 may be configured to obtain a DSM image of a target object. In some embodiments, the target image unit 420 can also be communicatively connected to a user interface 140. The target image unit 420 may be configured to receive the target image from the user interface 140. In some embodiments, the target image unit 420 may be configured to receive the selection of the target images from the user interface 140. In some embodiments, the target image unit 420 can also be communicatively connected to the detection unit 430. The target image unit 420 may be configured to send the target image to the detection unit 430 for object detection.

The detection unit 430 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (ie, a Function program) to perform the above operations in step 260. The detecting unit 430 may be configured to detect the target object in the area according to the DSM image of the area and the target object from the aerial image unit 410 and the target image unit 420. In some embodiments, the detection unit 430 may be configured to obtain one or more positions of the detected target objects, as in the operation in step 290 described above. In some embodiments, the detection unit 430 can also be communicatively connected to a display 160. The detection unit 430 may be configured to display one or more positions of the detected target objects on the aerial image or map of the area on the display 160, as in the operation in step 290 described above. In some specific embodiments, the detection unit 430 may be configured to calculate the number of detected target objects, as in the operation in step 290 described above. In some embodiments, the detection unit 430 can also be communicatively connected to an output 180. The detection unit 430 may be configured to transmit the calculated number of detected target objects to the output 180.

In some embodiments, the automatic object detection system 400 may include an aerial image unit 410, a target image unit 420, a detection unit 430, a positioning unit 440, a local aerial image unit 450, a capture unit 460, and a classification and recognition unit 470.

The aerial image unit 410 may be further configured to obtain the aerial image corresponding to the area of the DSM image in the area, as the operation in step 710 described above.

The positioning unit 440 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (ie, a function Program) to perform the operations in step 720 above. The positioning unit 440 may be further configured to obtain one or more positions of the detected target objects on the aerial image of the area. In some embodiments, the positioning unit 440 may be communicatively connected to the detection unit 430. The positioning unit 440 may be configured to receive the detected target objects from the detection unit 430, and obtain one or more positions of the detected target objects on the aerial image of the area. In some specific embodiments, the positioning unit 440 may also be communicatively connected to the local aerial image unit 450. The positioning unit 440 may be configured to transmit the positions of the detected target objects to the local aerial image unit 450. In some specific embodiments, the positioning unit 440 may also be communicatively connected to the classification and identification unit 470. The positioning unit 440 may be configured to transmit the acquired positions of the detected target objects to the classification and identification unit 470.

The local aerial image unit 450 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (ie A function program) to perform the above operations in step 730. The local aerial image acquisition unit may be configured to acquire one or more local aerial images at one or more locations of the detected target objects. In some embodiments, the local aerial image unit 450 can also be communicatively connected to the detection unit 430. The local aerial image unit 450 may be configured to receive the detected target objects and/or aerial images of the area from the detection unit 430. In some embodiments, the local aerial image unit 450 can also be communicatively connected to the capturing unit 460. The local aerial image unit 450 may be configured to transmit the acquired local aerial images at one or more positions of the detected target object to the capturing unit 460. In some embodiments, the local aerial image unit 450 can also be communicatively connected to the classification and identification unit 470. The local aerial image unit 450 may be configured to transmit the acquired local aerial images at one or more positions of the detected target object to the classification and identification unit 470.

The capture unit 460 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as a set of instructions, subroutines, or functions that can be executed on a processor or controller (that is, a Function program) to perform the above operations in step 740. The capturing unit 460 may be configured to capture one or more texture features from the one or more partial aerial images as one or more feature vectors. In some embodiments, the capturing unit 460 can also be communicatively connected to the local aerial image unit 450. The capturing unit 460 may be configured to receive the captured partial aerial images at one or more positions of the detected target objects from the partial aerial imaging unit 450. In some embodiments, the capturing unit 460 can also be communicatively connected to the user interface 140. The capturing unit 460 may be configured to accept user input or selection of the capturing algorithm from the user interface 140. In some embodiments, the capturing unit 460 can also be communicatively connected to the classification and identification unit 470. The extraction unit 460 may be configured to transmit the extracted one or more feature vectors to the classification and identification unit 470.

The classification and identification unit 470 may include appropriate types of hardware, such as integrated circuits and field programmable gate arrays, or software, such as instruction sets, subroutines, or functions that can be executed on a processor or controller (ie A function program) to perform the above operations in steps 750, 760, 770, and 780. In some embodiments, the classification and identification unit 470 can also be communicatively connected to the user interface 140. The classification and identification unit 470 can be configured to obtain a plurality of training data from the user interface 140. In some specific embodiments, the classification and identification unit 470 may also be communicatively connected to the positioning unit 440. The classification and identification unit 470 may be configured to receive the acquired positions of the detected target objects from the positioning unit 440. In some embodiments, the classification and identification unit 470 may also be communicatively connected to the local aerial image unit 450. The classification and identification unit 470 may be configured to receive the acquired partial aerial images at one or more positions of the detected target objects from the partial aerial image unit 450. In some embodiments, the classification and identification unit 470 may also be communicatively connected to the capturing unit 460. The classification and identification unit 470 may be configured to receive the extracted one or more feature vectors from the extraction unit 460.

The classification and identification unit 470 may be configured to obtain a plurality of training data, and the training data includes a plurality of aerial images of the same type of object as the target object. The classification and identification unit 470 may be further configured to train a classifier based on the plurality of training data. The classification and identification unit 470 may be further configured to classify the one or more partial aerial images by the trained classifier according to the one or more feature vectors. The classification and identification unit 470 may be further configured to identify the target objects between the one or more partial aerial images based on the classification results.

In some embodiments, the classification and identification unit 470 may be further configured to obtain one or more positions of the identified target objects, as in the operation in step 790 described above. In some embodiments, the classification and identification unit 470 can also be communicatively connected to the display 160. The classification and recognition unit 470 may be configured to display one or more positions of the recognized target objects on the aerial image or map of the area on the display 160, as in the operation in step 790 described above. In some embodiments, the classification and identification unit 470 may be configured to calculate the number of detected target objects, as in the operation in step 790 described above. In some embodiments, the classification and identification unit 470 can also be communicatively connected to an output 180. The classification and identification unit 470 may be configured to transmit the calculated number of the identified target objects to the output 180.

Figure 14 is a flowchart illustrating an exemplary method 1400 for automatic object detection in aerial images according to the disclosed specific embodiment. The method 1400 may include the following steps: obtaining an image of a region (step 1410), obtaining a plurality of partial aerial images from the image of the region (step 1420), and classifying the plurality of partial aerial images into a first by a classifier One type or one second type, wherein: the first type indicates that a partial aerial image includes a target object, the second type indicates that a partial aerial image does not include a target object, and the classifier passes through the first and second Training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the second training image includes adjusting the first A target object obtained from at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the training image (step 1430), and identify a target object in a partial aerial image in the first category (step 1440) And when an optimized soil adjustment plant growth index on the identified target object is lower than a plant disease critical value, it is determined that the identified target object has a plant disease (step 1450)

Step 1410 may include obtaining an image of a region. For example, the step 1410 of obtaining an image of a region may include accessing the image of the region of interest displayed in the first image from a computer-readable medium or a computer-readable storage device. For another example, the step 1410 of obtaining an image of a region may include an external input, such as the image input 120 in Figure 13, to receive an image of the region of interest. The image input 120 can be communicatively connected to, for example, UAV 100, unmanned drone, aircraft, helicopter, balloon, or satellite. In other words, the step 1410 of obtaining an image of an area may include receiving an image of the area of interest from the UAV 100, an unmanned target drone, an aircraft, a helicopter, a balloon, or a satellite.

In some embodiments, the step 1410 of obtaining an image of a region may include obtaining a plurality of images of each part of the region, and combining or stitching the plurality of images of each part of the region to obtain the region of interest Of the image. For example, the step 1410 of obtaining an image of a region may include obtaining a plurality of RGB images of each part of the region, and identifying and matching different features of the plurality of RGB images of each part of the region to establish a correspondence between RGB image pairs relation. The step 1410 of obtaining an image of a region may further include mixing a plurality of RGB images of each part of the region according to the corresponding relationship between the established RGB image pairs to obtain the RGB image of the region of interest.

Step 1420 may include obtaining a plurality of partial aerial images from the image of the area. For example, the step 1420 of obtaining a plurality of partial aerial images from the image of the area may include obtaining a plurality of 300x300 partial aerial images from the aerial image of the area in the first image. In some embodiments, the step 1420 of obtaining a plurality of partial aerial images from the image of the area may include the step 730 of obtaining partial aerial images as described in the method 700.

In some embodiments, the step 1420 of obtaining a plurality of local aerial images may include detecting on the image of the area based on a numerical surface model (DSM) image of the area and a DSM image of a target object. To obtain a plurality of local aerial images, as described in method 200 and method 700. Optionally, the step 1420 of obtaining a plurality of partial aerial images may include obtaining the plurality of partial aerial images according to a plurality of candidate positions on the image of the region. The plurality of candidate positions may be, for example, candidate positions of every pixel, every ten pixels, or every fifty pixels on the image of the region.

In some embodiments, when a ground sampling distance (GSD) of the image of the region is less than or equal to a GSD threshold, the step 1420 of obtaining a plurality of local aerial images may include the step 1420 of obtaining a plurality of local aerial images based on the region and the target object. The plurality of positions detected on the image of the region are detected on the DSM image of the region to obtain the plurality of partial aerial images, as described in method 200 and method 700. Optionally, when the GSD of the image in the region is greater than the GSD critical value, the step 1420 of obtaining a plurality of partial aerial images may include obtaining the plurality of partial aerial images according to the plurality of candidate positions of the image in the region image.

Step 1430 may include classifying the plurality of partial aerial images into a first type or a second type by a classifier. The first category indicates that a partial aerial image contains a target object. The second category indicates that a partial aerial image does not contain a target object. The classifier is trained through first and second training data, where the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the first The second training image includes a target object obtained by adjusting at least one of brightness, contrast, color saturation, resolution, or rotation angle of the first training image.

For example, the step 1430 of classifying the plurality of partial aerial images into a first type or a second type may include classifying the plurality of partial aerial images obtained in step 1420 as indicating that a partial aerial image contains a target object The first category or the second category indicates that a partial aerial image does not contain a target object. When the target object is the oil palm tree in the first image, the step 1430 of classifying the plurality of partial aerial images may include classifying one or more of the plurality of partial aerial images into the first category, where the partial aerial images include Oil palm tree. Optionally, the step 1430 of classifying the plurality of partial aerial images may include classifying one or more of the plurality of partial aerial images into a second category, wherein the partial aerial images do not include oil palm trees.

In step 1430, classifying the plurality of partial aerial images may include using a classifier to classify the plurality of partial aerial images. The classifier is trained through the first and second training data. The first training data may include a first training image that includes a target object, such as the oil palm tree in the ninth (a) image. In some specific embodiments, the classifier in step 1430 may be trained as in step 760 in method 700.

The second training data may include a second training image, which includes a target object obtained by adjusting at least one of brightness, contrast, color saturation, resolution, or rotation angle of the first training image. For example, the second training image may include one or more images which include oil palm trees with different brightness shown in the fifteenth image, oil palm trees with different contrasts in the sixteenth image, and the seventeenth image. The oil palm trees with different color saturations in the figure, the oil palm trees with different resolutions in the eighteenth figure, and the oil palm trees with different rotation angles in the nineteenth figure are adjusted by adjusting the ninth ( a) Obtained from the brightness, contrast, color saturation, resolution and/or rotation angle of the oil palm tree in the picture.

In some embodiments, when the second training data includes a second training image obtained by adjusting each of the brightness, contrast, color saturation, resolution, and a rotation angle of the first training image, the The classifier can improve the recognition rate of oil palm trees, for example, from 70% to 90%.

In some embodiments, the step 1430 of classifying the plurality of partial aerial images includes determining whether the appearance of a target object in a training image is rotationally symmetric, and in response to determining that the appearance of the target object is not rotationally symmetric, by The rotation angle of the first training image is adjusted to obtain the second training image. For example, aerial images of oil palm trees may not be rotationally symmetrical. In other words, after the angle of rotation, the oil palm tree looks different in the aerial image. The step 1430 of classifying the plurality of partial aerial images may include determining that the appearance of the oil palm tree is not rotationally symmetric. In response to this determination, step 1430 may also include adjusting the rotation angle of the oil palm tree to obtain the second training image. Therefore, the trainer will be trained by oil palm images with different rotation angles.

Optionally, when the target object is a basketball, the aerial image thereof may be rotationally symmetric, and step 1430 may include determining that the appearance of the basketball is rotationally symmetric. In response to this determination, step 1430 may not include adjusting the rotation angle of the oil palm tree to obtain the second training image. The trainer is not trained by oil palm images with different rotation angles.

In some embodiments, the rotation angle includes an angle greater than 0 degrees and less than 360 degrees. For example, after rotating an angle greater than 0 degrees and less than 360 degrees, the aerial images of people or animals may look different.

Step 1440 may include identifying a target object in a partial aerial image in the first category. For example, the step 1440 of identifying a target object in the first category may include identifying an oil palm tree in one or more images classified as the first category in step 1430. In some embodiments, the step 1440 of recognizing a target object in the first category may include the step 780 of recognizing the target object in the method 700.

Step 1450 may include determining that the identified target object has a plant disease when an optimized soil adjustment plant growth index on the identified target object is lower than a plant disease critical value. For example, the step 1450 of determining that the identified target object has a plant disease may include obtaining a plurality of optimized soil-adjusted vegetation indices (Optimized Soil-Adjusted Vegetation Indices, OSAVIs) on the identified oil palm in the first category, and the plural An optimized soil adjustment plant growth index is compared with a plant disease critical value, and when the OSAVIs of one or more oil palm trees are lower than the plant disease critical value, it is determined that the one or more oil palm trees have plant diseases. The plant disease cut-off value for oil palm trees may be 0.85, for example, OSAVI=0.85. When the plurality of OSAVIs on one or more identified oil palm trees is less than 0.85, it is determined that the one or more identified target objects have plant diseases.

OSAVI is a vegetation index, which explains the varying degrees of extinction of red and near-infrared through the vegetation canopy. OSAVI minimizes the influence of soil brightness from the spectral vegetation index involving red and near-infrared (Near-Infrared, NIR) wavelengths. OSAVI can be obtained by the following method: OSAVI = (1 + L) * (NIR-RED)/(NIR + RED + L), where L = 0.16, NIR is the near-infrared band index on the identified target object, and RED It is the red band index on the identified target object. NIR and RED can be obtained from the multispectral image of the region of interest captured by the multispectral camera.

The fifteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of target objects of different brightness. Fifteenth (a), fifteen (b), fifteen (c), fifteen (d), fifteen (e), fifteen (f), fifteen (g), fifteen (h) and ten Figure 5 (i) is an image containing oil palm trees, which have -60%, -45%, -30%, -15%, 0%, +15%, +30%, +45%, and +60. % Of different brightness levels. One or more of these images with different brightness can be used to train the classifier in step 1430. Since the brightness of the aerial images of a region can be changed, the classifier trained by these training images can improve the recognition rate of oil palm trees, for example, from 70% to 80%.

The sixteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of different contrasted target objects. Sixteen (a), sixteen (b), sixteen (c), sixteen (d), sixteen (e), sixteen (f), sixteen (g), sixteen (h), and ten Picture 6 (i) is an image containing oil palm trees, which have -60%, -45%, -30%, -15%, 0%, +15%, +30%, +45%, and +60. % Of different contrast levels. One or more of these images with different contrasts can be used to train the classifier in step 1430. As the aerial images of a region can be changed in contrast, the classifier trained by these training images can increase the recognition rate of oil palm trees, for example, from 75% to 85%.

The seventeenth diagram is a diagram of a plurality of exemplified training data that can be used to train an exemplified classifier for automatic object detection according to the disclosed specific embodiment, and includes a plurality of exemplified training data of target objects with different color saturations. Seventeen (a), seventeen (b), seventeen (c), seventeen (d), seventeen (e), seventeen (f), seventeen (g), seventeen (h), and ten Picture 7(i) is an image containing oil palm trees, which have different color saturation levels of 0%, 25%, 50%, 75%, 100%, 175%, 250%, 325%, and 400%. One or more of these images with different color saturations can be used to train the classifier in step 1430. As the color saturation of the aerial images of a region can be changed, the classifier trained by these training images can improve the recognition rate of oil palm trees, for example, from 72% to 80%.

The eighteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of target objects with different resolutions. The eighteenth (a), eighteen (b), eighteen (c), and eighteen (d) images are images containing oil palm trees, which have different blur levels of 0, 15, 30, and 45, respectively. One or more of these images with different color saturations can be used to train the classifier in step 1430. Since the resolution of aerial images of a region can be changed, the classifier trained by these training images can improve the recognition rate of oil palm trees, for example, from 72% to 86%.

The nineteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of target objects with different rotation angles. The nineteenth (a), nineteen (b), nineteen (c), nineteen (d), nineteen (e), and nineteen (f) are images containing oil palm trees, each with 0 Different rotation angle levels of degree, 30 degree, 45 degree, 60 degree, 90 degree and 180 degree. One or more of these images with different rotation angles can be used to train the classifier in step 1430. Since the aerial images of a region can be changed in angle, the classifier trained by these training images can improve the recognition rate of oil palm trees, for example, from 75% to 86%.

The twentieth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data for target objects of different sizes. As shown in Figure 20(a), a training image contains small young palm trees as the target object. As shown in Figure 20(b), a training image contains a large mature palm tree as another target object. The target object in the first training image in step 1430 may include the two palm trees of different sizes. The classifier trained by these training images can improve the recognition rate of oil palm trees, for example from 82% to 86%, because it can identify those young oil palm trees and mature oil palm trees in the region of interest .

The twenty-first figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data for more than two target objects. As shown in Figure 21(a), a training image contains two palm trees as target objects. As shown in Figure 21(b), a training image contains three palm trees as target objects. The target object in the first training image in step 1430 may include the two or more palm trees. The classifier trained by these training images can increase the recognition rate of oil palm trees, for example, from 81% to 85%, because it can recognize two or more oil palm trees in local aerial images. The partial aerial image obtained in step 1420 may inevitably contain two or more oil palm trees. This is because a fixed image size, such as 300x300, can be used for all partial aerial images, while oil palm The trees can be randomly scattered in the area.

The twenty-second image is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and includes a plurality of examples of training data of non-target objects in the image in the region. The twenty-second (a), twenty-two (b), twenty-two (c), and twenty-two (d) diagrams contain a house, another house, a car, and two cars, respectively. The classifier in step 1430 can be trained from a first number of training images including the ninth (a) picture and the first and second pluralities of training images in the fifteenth to twenty-first pictures, and the ninth (b) A second number of training images of non-target images in the image and/or image 22. The first number of training images may be substantially equal to the second number of training images. For example, the first and second quantities can both be 10,000. The classifier trained by these training images can increase the recognition rate of oil palm trees to, for example, 90%, because these training images provide different features for the classifier.

The twenty-third image is based on the disclosed specific embodiment, which can be used to train an example classifier for automatic object detection, and includes a plurality of examples of training data for non-target objects that are not in the image in the region. The twenty-third (a), twenty-three (b), twenty-three (c), twenty-three (d), twenty-three (e), and twenty-three (f) pictures include beach scenes. Not in the image of the area in the first image. The classifier of step 1430 can be trained from a first number of training images including the ninth (a) picture and the first and second training images in the fifteenth to twenty-first pictures, including the ninth (b) A second number of training images of non-target images in the images in the image and/or the 22nd image, and a second number of training images containing the non-target objects in the image that are not in the region in the 23rd image Three numbers of training images. In some embodiments, the first number, the second number, and the third number of training images are substantially equal. For example, the first number, the second number, and the third number may all be eight thousand.

The twenty-fourth diagram is a diagram of the target object with plant diseases in the area according to the disclosed specific embodiment. As shown in Figure 24, the oil palm trees 2401 and 2402 have dead leaves. These dead leaves may cause different reflections in different wavebands (such as blue, green, red and near-infrared wavebands). UAV 100 can also be set to capture multi-spectral images of the area with a multi-spectral camera. The OSAVIs of oil palm 2401 and 2402 were 0.79 and 0.81, respectively.

After the oil palm trees 2401 and 2402 are identified in step 1440, the step 1450 of determining that the identified target object has a plant disease may include comparing the OSAVIs of the oil palm trees 2401 and 2402 with a plant disease threshold (for example, 0.85) , And determined that both oil palm trees 2401 and 2402 have plant diseases. Farmers can obtain the location of these oil palm trees with plant diseases by using the method in this article, and take measures to rescue these oil palm trees.

Another aspect of the present invention relates to a system for detecting objects from aerial images. Figure 13 is a block diagram illustrating an exemplary system for automatic object detection in aerial images according to the disclosed specific embodiment. The automatic object detection system 400 can be configured to perform the method 1400 described above and shown in FIGS. 14-24.

It is still another aspect of the present invention, which relates to a non-transitory computer readable medium storing instructions, which when executed will cause one or more processors to perform the operation of detecting objects from aerial images. This operation may include, but is not limited to, all the aforementioned methods and embodiments. In some specific embodiments, a part of the foregoing operations or steps in the embodiments may be performed remotely or separately. In some embodiments, this operation may be performed by one or more distributed systems.

Those skilled in the art will understand that many modifications and changes can be made to the disclosed methods and systems for detecting objects in aerial images. From the specifications and practical considerations of the disclosed method and system for detecting objects in aerial images, those skilled in the art can also understand other specific embodiments. The descriptions and examples considered here are only examples, and the exact scope of the present invention is listed in the following patent application scopes and their equivalents.

100: Flooding prediction system 120: video input 140: User Interface 160: display 180: output 200: method 220, 240, 260, 290: steps 400: Automatic object detection system 410: Aerial image unit 420: target image unit 430: Detection Unit 440: positioning unit 450: Partial aerial image unit 460: Capture Unit 470: Classification and Identification Unit 700: method 710~790: steps 801, 802, 803: oil palm 1001, 1002, 1003: pink circle/oil palm tree/location 1016, 1017, 1018: blue circle/oil palm tree/location 1101, 1102, 1103: white circles 1400: method 1410~1450: steps 2401, 2402: oil palm

The first figure is a schematic diagram of an example aerial image used for automatic object detection in an area according to the disclosed embodiment.

The second figure is a flowchart illustrating an exemplary method for automatic object detection in aerial images according to the disclosed specific embodiment.

The third figure is a diagram illustrating an example DSM image of the region for automatic object detection corresponding to the example aerial image of the region in the first figure according to the disclosed specific embodiment.

The fourth figure shows two exemplary DSM image patterns of an exemplary target object type used for automatic object detection according to the disclosed embodiment.

The fifth diagram is a diagram of an exemplified image of the pairing rate between the DSM image in the third diagram used for automatic object detection and the pairing ratio of the present image in the fourth diagram, according to the disclosed specific embodiment. Mode.

The sixth figure is an example aerial image pattern for marking the area where the positions of the detected and exemplified target objects are marked according to the instantiation method used for automatic object detection in the second figure according to the disclosed specific embodiment.

The seventh figure is a flowchart illustrating another exemplary method for automatic object detection in aerial images according to the disclosed specific embodiment.

The eighth figure is based on the disclosed specific embodiment, in accordance with the exemplified method used for automatic object detection in the second figure, marking the area where the positions of the exemplified target objects have been detected and exemplifying a partially enlarged image of the aerial image .

The ninth diagram is a diagram of a plurality of exemplified training data that can be used to train the exemplified classifier for automatic object detection according to the disclosed specific embodiment.

The tenth figure is based on the disclosed specific embodiment, in accordance with the exemplified method for automatic object detection in the seventh figure, the area where the classification result is marked on the positions of the detected target objects exemplifies the aerial image Partially enlarged diagram.

The eleventh figure is based on the disclosed specific embodiment, in accordance with the exemplified method used for automatic object detection in the seventh figure, marking the regions of the exemplified target object positions that have been correctly detected and recognized, exemplifying the aerial image Partially enlarged diagram.

The twelfth figure is based on the disclosed specific embodiment, in accordance with the instantiation method for automatic object detection in the seventh figure, marking the region of the classification result on the positions of the detected and classified instantiation target objects. Aerial image schema.

Figure 13 is a block diagram illustrating an exemplary system for automatic object detection in aerial images according to the disclosed specific embodiment.

Figure 14 is a flowchart illustrating an exemplary method for automatic object detection in aerial images according to the disclosed specific embodiment.

The fifteenth diagram is a diagram of a plurality of exemplified training data that can be used to train an exemplified classifier for automatic object detection according to the disclosed specific embodiment, and includes a plurality of exemplified training data of target objects of different brightness.

The sixteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of different contrasted target objects.

The seventeenth diagram is a diagram of a plurality of exemplified training data that can be used to train an exemplified classifier for automatic object detection according to the disclosed specific embodiment, and includes a plurality of exemplified training data of target objects with different color saturations.

The eighteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of target objects with different resolutions.

The nineteenth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data of target objects with different rotation angles.

The twentieth figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data for target objects of different sizes.

The twenty-first figure is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and contains a plurality of examples of training data for more than two target objects.

The twenty-second image is a pattern that can be used to train an example classifier for automatic object detection according to the disclosed specific embodiment, and includes a plurality of examples of training data of non-target objects in the image in the region.

The twenty-third image is based on the disclosed specific embodiment, which can be used to train an example classifier for automatic object detection, and includes a plurality of examples of training data for non-target objects that are not in the image in the region.

The twenty-fourth diagram is a diagram of the target object with plant diseases in the area according to the disclosed specific embodiment.

none

200: method

220, 240, 260, 290: steps

Claims (23)

A system for detecting objects from aerial images, the system includes: Memory used to store commands; At least one processor is configured to execute the instruction to: Obtain an image of a region; Obtain a plurality of partial aerial images from the image in the area; A classifier classifies the plurality of local aerial images into a first category or a second category, where: The first category indicates that a partial aerial image contains a target object, The second category means that a partial aerial image does not contain a target object, and The classifier is trained through first and second training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the first The second training image includes a target object obtained by adjusting at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the first training image; and Identify a target object in a local aerial image in the first category. The system according to claim 1, wherein the classification of the plurality of partial aerial images includes: Determine whether the appearance of a target object in a training image is rotationally symmetric; and In response to determining that the appearance of the target object is not rotationally symmetric, the second training image is obtained by adjusting the rotation angle of the first training image. The system according to claim 1, wherein the rotation angle includes an angle greater than 0 degrees and less than 360 degrees. The system according to claim 1, wherein said obtaining the plurality of partial aerial images includes: According to a plurality of positions detected on the image of the area based on a numerical surface model (DSM) image of the area and a DSM image of a target object, to obtain the plurality of local aerial images; or Obtain the plurality of partial aerial images according to the plurality of candidate positions on the image of the region. The system described in claim 4, wherein: When a ground sampling distance (GSD) of the image of the region is less than or equal to a GSD critical value, the obtaining the plurality of partial aerial images includes obtaining data in the region based on the DSM image of the region and the target object. The plurality of positions detected on the image to obtain the plurality of partial aerial images; and When the GSD of the image of the region is greater than the GSD threshold, the obtaining the plurality of partial aerial images includes obtaining the plurality of partial images according to the plurality of candidate positions of the image of the region. The system according to claim 1, wherein the target object of the first training image includes two different sizes. The system according to claim 1, wherein one of the first training images includes two target objects. The system according to claim 1, wherein the classifier is trained from: A first number of training images, including the first and second pluralities of training images; and A second number of training images, which contain non-target objects, The first number of training images is substantially equal to the second number of training images. The system according to claim 1, wherein the classifier is trained from: A first number of training images, including the first and second training images; A second number of training images, including non-target objects in the image in the region; and A third number of training images, which include non-target objects that are not in the image in the region. The system according to claim 9, wherein the first, second, and third numbers of training images are approximately equal. The system of claim 1, wherein the at least one processor is configured to execute the instruction to: When an optimized soil adjusted plant growth index on the identified target object is lower than a plant disease critical value, it is determined that the identified target object has a plant disease. A method for detecting objects in aerial images, the method includes: Obtain an image of a region; Obtain a plurality of partial aerial images from the image in the area; A classifier classifies the plurality of local aerial images into a first category or a second category, where: The first category indicates that a partial aerial image contains a target object, The second category means that a partial aerial image does not contain a target object, and The classifier is trained through first and second training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the first The second training image includes a target object obtained by adjusting at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the first training image; and Identify a target object in a local aerial image in the first category. The method according to claim 12, wherein the classification of the plurality of partial aerial images includes: Determine whether the appearance of a target object in a training image is rotationally symmetric; and In response to determining that the appearance of the target object is not rotationally symmetric, the second training image is obtained by adjusting the rotation angle of the first training image. The method according to claim 12, wherein the obtaining the plurality of partial aerial images includes: According to a plurality of positions detected on the image of the area based on a numerical surface model (DSM) image of the area and a DSM image of a target object, to obtain the plurality of local aerial images; or Obtain the plurality of partial aerial images according to the plurality of candidate positions on the image of the region. The method according to claim 14, wherein: When a ground sampling distance (GSD) of the image of the region is less than or equal to a GSD critical value, the obtaining the plurality of partial aerial images includes obtaining data in the region based on the DSM image of the region and the target object. The plurality of positions detected on the image to obtain the plurality of partial aerial images; and When the GSD of the image of the region is greater than the GSD threshold, the obtaining the plurality of partial aerial images includes obtaining the plurality of partial images according to the plurality of candidate positions of the image of the region. The method according to claim 12, wherein the target object of the first training image includes two different sizes. The method according to claim 12, wherein the classifier is trained from: A first number of training images, including the first and second pluralities of training images; and A second number of training images, which contain non-target objects, The first number of training images is substantially equal to the second number of training images. The method according to claim 12, wherein the classifier is trained from: A first number of training images, including the first and second training images; A second number of training images, including non-target objects in the image in the region; and A third number of training images, which include non-target objects that are not in the image in the region. The method according to claim 18, wherein the first, second, and third numbers of training images are approximately equal. The method described in claim 12 includes: When an optimized soil adjusted plant growth index on the identified target object is lower than a plant disease critical value, it is determined that the identified target object has a plant disease. A non-transitory computer-readable medium that stores instructions. When executed, it will cause one or more processors to perform the operation of detecting objects from aerial images. The operations include: Obtain an image of a region; Obtain a plurality of partial aerial images from the image in the area; A classifier classifies the plurality of local aerial images into a first category or a second category, where: The first category indicates that a partial aerial image contains a target object, The second category means that a partial aerial image does not contain a target object, and The classifier is trained through first and second training data, wherein the first training data includes a first training image, the first training image includes a target object, and the second training data includes a second training image, the first The second training image includes a target object obtained by adjusting at least one of the brightness, contrast, color saturation, resolution, or rotation angle of the first training image; and Identify a target object in a local aerial image in the first category. The non-transitory computer-readable medium according to claim 21, wherein the plurality of partial aerial images of the classification includes: Determine whether the appearance of a target object in a training image is rotationally symmetric; and In response to determining that the appearance of the target object is not rotationally symmetric, the second training image is obtained by adjusting the rotation angle of the first training image. The non-transitory computer-readable medium described in claim 21, wherein the operation includes: When an optimized soil adjusted plant growth index on the identified target object is lower than a plant disease critical value, it is determined that the identified target object has a plant disease.

Images