WO2023090988A1 - Procédé de détection et/ou de comptage d'arbres - Google Patents

Procédé de détection et/ou de comptage d'arbres Download PDF

Info

Publication number
WO2023090988A1
WO2023090988A1 PCT/MY2022/050074 MY2022050074W WO2023090988A1 WO 2023090988 A1 WO2023090988 A1 WO 2023090988A1 MY 2022050074 W MY2022050074 W MY 2022050074W WO 2023090988 A1 WO2023090988 A1 WO 2023090988A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
detection points
trees
oil palm
images
Prior art date
Application number
PCT/MY2022/050074
Other languages
English (en)
Inventor
Mohammad Zafrullah SALIM
Original Assignee
Sime Darby Plantation Intellectual Property Sdn Bhd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sime Darby Plantation Intellectual Property Sdn Bhd filed Critical Sime Darby Plantation Intellectual Property Sdn Bhd
Publication of WO2023090988A1 publication Critical patent/WO2023090988A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10032—Satellite or aerial image; Remote sensing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20036—Morphological image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation
    • G06T2207/30188—Vegetation; Agriculture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30242—Counting objects in image

Definitions

  • the present invention relates generally to a method for detecting and/or counting trees (mature and immature trees) using high-resolution satellite images. More particularly to a method for detecting and/or counting immature oil palm trees from high-resolution satellite images using morphological filtering means.
  • Chinese Patent Application CN103353983A describes a morphological filtering method based on maximum tree structure, firstly constructing the maximum tree, based on the maximum tree established morphological filter, wherein the filtering process comprises maximum tree construction, filtering and reconstruction, a greyscale image is represented as maximum tree structure, using shear rule, the maximum tree nodes to kiss, maximum tree node does not meet Branch rule, so as to achieve the filtering effect, when executing the step of deleting the maximum tree node, i.e. not meeting the rules in deleting the connected region of the image, deleting the node containing the pixel point is a new grey value according to the filtering rule.
  • the method enables simplifying morphological image filtering operation, avoiding structural elements selecting difficulties and optimizing filtering effect of a morphological image. This prior art document does not describe the method of the present invention.
  • Chinese Patent Application CN104361330A describes fertilizing crops of corn precise medicine adding system identification method, comprising the following steps: an industrial camera and a lens collecting the RGB colour image, obtaining RGB color image using the improved green grey algorithm; using the improved method for obtaining a median of median filtering to remove image noise, the maximum between-cluster variance method to image binarization after removing noise, noise filtering binaryzation image by applying morphological algorithm based on Mahalanobis distance and corn leaf rule extracting crop row frame; fitting the main crop row is linear based on Hough conversion point of the main framework, the influence of noise
  • the invention maximally keeps crop row information, remove background interference and improve the operation speed, based on the Mahalanobis distance and corn leaf rule extracting accurate crop row frame, effectively prevent weeds, suitable for different crops and lighting conditions, crop row higher than 98.3% accuracy, is a precision spray nozzle automatically aligned in agriculture system provides an effective method.
  • This prior art document does not describe the method of the present invention.
  • a monitoring method of transmission line tree flash hidden trouble based on satellite optical image comprises the following steps: 1) using the best index method to analyze the 4 wave bands of the satellite image, wherein 1 wave band, 2 wave band, 3 wave band, 4 wave band are corresponding: blue light wave band, green light wave band, red light wave band and near infrared wave band; 4 wave bands are combined by four wave bands, respectively: 1, 2, 3 band, 1, 2, 4 band, 1, 3, 4 band, 2, 3, 4 wave band, respectively calculating the index of the wave band combination, selecting the maximum band combination as the experiment wave band, then using ENVI software to generate false color image of experiment wave band, performing normalization vegetation index processing and mask processing on the false color image, dividing the tree crown region image by morphological filtering and marking watershed algorithm; 2) using the maximum likelihood method for coarse classifying the transmission line image in the satellite image, the classified transmission line image in the ENVI class for manual
  • United States Patent Application US20200125823A1 describes a system for detecting objects from aerial imagery, the system comprising memory for storing instructions, at least one processor configured to execute the instructions to - 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.
  • the advantage of this prior art document is that it quickly and precisely detects target objects from aerial imagery of an area of interest. Possible to estimate the accuracy of the aforementioned object
  • United States Patent Application US 10586105B2 describes a computer-implemented method for crop type identification using satellite observation and weather data, comprising extracting current data and historical data from pixels of a plurality of satellite images of a target region, wherein the current data and the historical data includes a plurality of remote sensing measurements and weather data over a period of time, generating a set of temporal sequences of vegetation indices having corresponding timestamps from the plurality of remote sensing measurements, the vegetation indices being based on vegetation color associated with a crop growth cycle, wherein each temporal sequence is associated with a respective pixel location within a satellite image and a crop season, based on the weather data, converting each timestamp of the temporal sequences into a modified temporal variable correlating with actual crop growth, training a classifier using a set of historical temporal sequences of vegetation indices with respect to the modified temporal variable as training features and corresponding historically known crop types as training labels, identifying at least one crop type for each pixel location within the satellite images using
  • pixel-based classification was often used to classify feature classes from the images. However, it is very suitable to extract information based on the spectral signature of the object.
  • the object-based classification has been widely applied because of its ability to identify additional information such as the size, shape, texture, and also adjacent object occurrences related to the other.
  • OBIA object -based image analysis
  • OBIA developed by Hay and Castilla is a discipline in spatial science that focuses on subdividing remote sensing images to meaning full objects through utilization of spatial and spectral properties.
  • the idea of analyzing an image in object space rather than in pixel space is developed due to insufficiencies of pixel-based methods, especially on high-resolution imageries.
  • object space has been reinforced in computational capacities and availabilities for analysis of high-resolution images such as IKONOS, GeoEye, and WorldView.
  • IKONOS GeoEye
  • WorldView WorldView
  • Chinese Patent Application CN110570399A describes a method enables estimating age of oil palm trees in large areas, and obtaining the spatial distribution information of age of oil palm trees.
  • the invention claims an oil palm age measuring method based on time sequence remote sensing data, using remote sensing image in time sequence remote sensing satellite data, combination for detecting LandTrendr model of forest disturbance to obtain palm forest disturbance in year Franchise range, thereby calculating the age of oil palm.
  • the invention oil palm age can be used for large area estimation, obtaining the space distribution information of the oil palm, estimation can be applied to palm oil. This prior art document does not describe the method of the present invention.
  • a Progressive Morphological Filter for Removing Non-Ground Measurements from Airborne LIDAR Data [https://users.cis.fiu.edu/ ⁇ chens/PDF/TGRS.pdf] describes a progressive morphological filter built to remove non-ground LIDAR measurements in both urban and mountain area accurately by gradually increasing the sizes of the opening operation and using elevation difference threshold. It further states that two fundamental operations, dilation and erosion are commonly employed to enlarge (dilate_ or reduce (erode) the size of features in binary images. Dilation and erosion operations may be combined to produce opening and closing operations. The concept of erosion and dilation has been extended to multilevel (grayscale) images and corresponds to finding minimum or maximum of the combinations of pixel values.
  • the combination of erosion and dilation generates opening and closing operations that are employed to filter LIDAR data.
  • the opening operation is achieved by performing an erosion of dataset followed by a dilation while the closing operation is accomplished by carrying out a dilation first and then an erosion.
  • Chinese patent (CN 105551002B) describes an image morphological filtering method, wherein, comprises: obtaining the image information and the morphological processing information, wherein said image information includes the original image and the image size of the original image, said morphology processing information comprises filtering operation type and structural elements corresponding to the filter operation type according to the image size, the memory of the structural elements and image processor to carry out decomposition on the original image, obtaining multiple sub -image, using the image processor in turn with the filter operation corresponding to the type of the filter operation based on said structural element to said plurality of sub-images, all pixel point of the plurality of processing unit, wherein said image processor when processing any image, using the image processor of the image comprises filtering operation at the same time, obtaining the image information and the morphological processing information, further comprising: according to the image size, the memory of said structural elements and said image processor to the structure element to decompose, obtaining multiple substructure element, using the image processor in turn with the filter operation corresponding to the type of the plurality of
  • Chinese patent application CN 111582035A describes an identification method for age tree age, comprising: obtaining the resolution satellite image in the calendar year key growth period according to the object candidate characteristic library; and constructing the resolution NDVI time sequence image in the key growth period; wherein the object candidate characteristic library is constructed according to the ground object vector training set data and the multi-time phase high resolution NDVI time sequence image; cutting the resolution NDVI time sequence image in the key growth period according to the orchard area vector; obtaining the calendar year orchard area vector; using unsupervised classification method to extract the almanac region vector to obtain the binarized image of the almanac region according to the object candidate characteristic base; and performing inverse time sequence pixel-by-pixel accumulation algorithm to the binary image of the almanac region to obtain the or
  • the present invention relates generally to a method for detecting and/or counting trees (mature and immature trees) using high-resolution satellite images. More particularly to a method for detecting and/or counting immature oil palm trees from high-resolution satellite images using morphological filtering means.
  • the present invention provides a method for automatically detecting and/or counting trees comprising obtaining a plurality of satellite images (1) of an agriculture estate or plantation, marking boundaries of plurality of regions within the plurality of satellite images (1) of the agriculture estate or plantation and applying at least one morphological filter (2) to the boundaries of the plurality of regions within the plurality of satellite images (1) of the agriculture estate or plantation to produce plurality of morphological filtered images comprising multiple shadows of detection points.
  • FIG. 1 illustrates the method of the present invention.
  • Figure 2 illustrates a high-resolution satellite image of an area of immature oil palm trees.
  • Figure 3 illustrates the erosion on the high-resolution satellite imagery of an area of immature oil palm trees using erosion filter.
  • Figure 4 illustrates the result from the watershed segmentation creating boundary for each palm tree.
  • Figure 5 illustrates the result from seed points using minima pixels on the high- resolution satellite imagery of the area of immature oil palm trees.
  • Figure 6 illustrates cleaning up of detection points using point displacement on the high- resolution satellite imagery of the area of immature oil palm trees.
  • Figure 7 illustrates cleaning up of detection points using thinning on the high-resolution satellite imagery of the area of immature oil palm trees.
  • Figure 8 illustrates the results of the detection of the immature oil palm trees.
  • Figure 9 illustrates three study areas for the purposes of detection of the immature oil palm trees: (a) flat terrain, (b) flat and hilly terrains and (c) hilly terrain.
  • Figure 10 illustrates a high-resolution satellite image of an area of mature oil palm trees.
  • Figure 11 illustrates a high-resolution satellite image of immature and mature oil palm trees of an oil palm plantation.
  • Remote sensing is the process of detecting and monitoring the physical characteristics of an area by measuring its reflected and emitted radiation at a distance (typically from satellite or aircraft), whereby the cameras on satellites and airplanes take images of large areas on the Earth's surface, allowing us to see more details than what we can see from standing on the ground. [Source: https://www.usgs.gov/]
  • Orbital remote sensing with new generation satellites have been providing high- resolution images for visual identification of individual tree crowns from the images since 1999.
  • the increase in spatial resolution has also had a profound effect in image processing techniques and has motivated the development of new object-based procedures to extract information.
  • the spatial resolution and spectral resolution of remote sensing images have been greatly improved, and thus application fields of remote sensing technology are expanding.
  • High resolution satellite remote sensing imagery is a basic spatial data source to construct the digital Earth and can be widely applied in multiple subjects and areas, such as geology, vegetation, agriculture, forestry, and oceanography, etc., especially in disaster emergency monitoring, real-time monitoring of land cover, ocean monitoring, and Earth-crust displacement and ground settlement monitoring.
  • high-resolution satellite images possess the characteristics of timeliness and authenticity, rapidly acquiring information, relatively low cost, no geographic restrictions, and abundant spatial information and texture information, etc.
  • developing high-resolution satellite image processing techniques and conducting various application studies have continuously received attention.
  • this method is limited in the sense that it is not able to detect and/or count immature oil palm trees, specifically aged between 12 to 24 months after planting in the oil palm plantation.
  • This method is not suitable for detecting immature oil palm trees because the crown of an immature oil palm tree is not easily identifiable due to the trees’ size.
  • the tree crown is essentially the top of any tree with branches stemming out from the main trunk of the tree.
  • the immature oil palm trees are barely visible due to its canopy size as represented in the remote sensing images (see Figure 2) mainly due to limitations of the satellite image resolution.
  • image resolution means how much of information is portrayed or displayed in the image.
  • High-resolution contains more pixels on the image which creates high quality and clear images for processing. Images which contains fewer pixels are lower in resolution and hence if the images are stretched or expanded, it will be blurry for processing.
  • high-resolution satellite images are the preferred for the method of the present invention.
  • the present invention relates generally to a method for detecting and/or counting trees (mature and immature trees) using high-resolution satellite images. More particularly to a method for detecting and/or counting immature oil palm trees from high-resolution satellite images using morphological filtering means.
  • One object of the present invention is to provide a means for detecting and/or counting immature and mature trees.
  • a further object of the present invention is to provide an automated means for detecting and/or counting immature oil palm trees via remote sensing.
  • the present invention targets immature oil palm trees aged between 12 to 24 months.
  • Another object of the present invention is to provide a means for detecting and/or counting immature oil palm trees using high-resolution satellite images.
  • High-resolution satellites images are preferred for this present invention as the images cover less than Im per pixel to provide high quality and clear images for processing. Also, remote sensing with high- resolution images is cost-effective and a reliable means to provide information pertaining to the immature oil palm trees.
  • High-resolution satellite images for the purposes of this present invention is obtained I purchased from the company, Planet Labs Inc. (https:/ / www.planet.com/ company/ ).
  • the object of the present invention is to produce reliable detection points from the high quality and clear satellite images for detecting and/or counting immature oil palm trees using morphological filter.
  • Morphological filters were originally developed and used for erosion and land deformation purposes and to the best of the knowledge of the inventors of the present invention, it is not known for any third parties to use morphological filtering means (specifically erosion morphological filter) to detect and/or count oil palm trees (immature and mature oil palm trees, specifically immature trees) in estates with any types of terrains (flat terrains, hilly and flat terrains and hilly terrains).
  • this present invention focuses on immature oil palm trees, however, can also be used to detect and/or count mature oil palm trees depending on the preference of the user of the present invention. As there is already a wide range of high accuracy and established methods for detecting and/or counting mature oil palm trees, hence this might not be preferred means for detecting and/or counting mature oil palm trees.
  • the present invention can also be used to detect and/or count other immature trees apart from immature oil palm trees, preferably if the shape and planting design is similar to the shape and planting design of oil palm trees in an oil palm plantation. This is because the method of the present invention focuses on detecting and enhancing and/or augmenting the shadows of the immature oil palm trees in order to produce the detection points (immature oil palm trees).
  • Yet another object of the present invention is to ensure optimum estate management in order for the oil palm trees to have same access to nutrients (i.e. fertilisers) for optimal growth and also sustainable management of the immature oil palm trees with the correct and precise application of fertiliser. This would result in accurate stand per hectare for each estate based on accurate information obtained from the high-resolution satellite images for the estates in order to manage resources and maximise profits for the company.
  • nutrients i.e. fertilisers
  • a further object of the present invention is to provide a method for detecting and/or counting immature oil palm trees which is much efficient, higher in productivity and cost efficient as compared to using drone imaging for counting the immature oil palm trees.
  • the object of the present invention is to provide a method to be used for estates with any types of terrains.
  • the method of the present invention specifically focuses on the analysis of high-resolution satellite images using morphological filtering which allows for the detection and/or counting immature oil palm trees in the oil palm estates for any types of terrains - flat terrains, flat and hilly terrains and/or hilly terrains.
  • the present invention does not require reliance on human resource to identify areas planted with mature and/or immature oil palm trees (focusing on immature oil palm trees) which is unrealistic due to the large area of an estate.
  • the present invention provides high accuracy, feasible, automated and efficient means of detecting immature oil palm trees for sustainable and profitable management of oil palm estates.
  • morphological filtering means specifically erosion morphological filters
  • oil palm trees immature and mature oil palm trees
  • the present invention is aim to provide an automated, user-friendly, simple and inexpensive means to allow oil palm managers to count immature oil palm trees using remote sensing technique.
  • the present invention relies on high-resolution satellite images which can be used with any open software available for use, improved accuracy and precise in detecting and counting the immature oil palm trees and also feasible for use.
  • the present invention provides an automated method for detecting and/or counting trees, comprising a method for automatically detecting and/or counting trees, comprising: • obtaining a plurality of satellite images (1) of an agriculture estate or plantation;
  • the plurality of morphological filtered images comprising the multiple shadows of detection points are segmented to produce multiple detection points.
  • the segmentation is conducted by means of a watershed segmentation (3).
  • the multiple detection points are further assembled and/or displaced by means of a point displacement (4) and the overlapping multiple detection points are removed or thinned out by means of a point thinning (5).
  • the orthorectified high-resolution satellite images (1) have a ground sampling distance of less than Im per pixel. Orthorectification increases the accuracy of the high-resolution satellite images for the purposes of the present invention.
  • the morphological filter (2) used for the present invention is erosion, dilation, opening, closing or any combination thereof, preferably erosion morphological filter (2).
  • High-resolution satellite images (1) using erosion morphological filter (2) allows for the detection and/or counting immature oil palm trees in the oil palm estates for any types of terrains - flat terrains, flat and hilly terrains and/or hilly terrains.
  • the at least on morphological filter (2) of part c) is erosion, dilation, opening, closing or any combinations thereof, preferably an erosion morphological filter (2).
  • Detection points means mature and/or immature trees, specifically focusing on immature oil palm trees for this present invention.
  • This method uses minima pixels of the high-resolution satellite images as this method focuses on using erosion morphological filter (2) to enhance and/or augment shadow of the oil palm trees to produce detection points (immature oil palm trees). Pixel value for shadows are usually at the lower range, hence the focus on minima pixels for the method of this present invention. Maxima pixels are used to detect mature palm trees as the pixels of the tree crown are usually at the higher range. Maxima pixel is the maximum value given for the pixel of a particular high-resolution satellite image (1) and minima pixel is the minimum value given for a particular high-resolution satellite image (1).
  • the detection points can be further refined by adjusting at least one of brightness, colour, resolution or combination thereof based on preference or needs of the user of the present invention .
  • the present invention provides an automated means for detecting and/or counting immature oil palm trees, comprising a method for detecting and/or counting trees, comprising obtaining a plurality of satellite images (1) of an agriculture estate or plantation, marking boundaries of plurality of regions within the plurality of satellite images (1) of the agriculture estate or plantation and applying at least one morphological filter (2) to the boundaries of the plurality of regions within the plurality of satellite images (1) of the agriculture estate or plantation to produce plurality of morphological filtered images comprising multiple shadows of detection points (immature oil palm trees).
  • Morphology is a broad set of image processing operations that process images based on shapes. Morphological operations apply a structuring element to an input image, creating an output image of the same size. In a morphological operation, the value of each pixel in the output image is based on a comparison of the corresponding pixel in the input image with its neighbours.
  • the most basic morphological operations are dilation and erosion. Dilation adds pixels to the boundaries of objects in an image, while erosion removes pixels on object boundaries. The number of pixels added or removed from the objects in an image depends on the size and shape of the structuring element used to process the image.
  • the state of any given pixel in the output image is determined by applying a rule to the corresponding pixel and its neighbours in the input image.
  • Morphological filter (2) for this present invention means a nondinear processing methodology based on mathematical morphology.
  • Nondinear mathematical morphology works on preserving the edges and borders of the images, in a different manner as compared to linear methods.
  • Linear method operates by assigning the average value of all pixels in the neighbourhood of the input pixels to the corresponding pixel in the output image, also known as averaging filter. Linear method is suitable to use for smoothing an image but at the same time it blurs the image.
  • Morphological operators namely dilate, erode, open, and close — can be applied for image filtering purposes to grow or shrink image regions, as well as to remove or filldn image region boundary pixels.
  • Dilation and erosion are basic operators in the area of mathematical morphology.
  • the basic effect of dilation on an image is to gradually enlarge the boundaries of regions of foreground pixels, typically white pixels. As areas of foreground pixels grow in size, holes within those regions become smaller.
  • the basic effect of erosion on an image is to erode away the boundaries of regions of foreground (bright) pixels, typically white pixels. As areas of foreground pixels shrink in size, holes within those areas become larger.
  • Dilation can also be used for edge detection by taking the dilation of an image and then subtracting away the original image, thereby highlighting just those new pixels at the edges of objects that were added by the dilation. This will highlight just those pixels at the edges of objects that were removed by the erosion.
  • Opening is derived from the fundamental morphological operations of erosion and dilation. The basic effect of opening is like erosion whereby it tends to remove some of the foreground (bright) pixels from the edges of regions of foreground pixels. In general, it is less destructive than erosion. Closing is opening performed in reverse and can be defined simply as a dilation followed by an erosion using the same structuring element for both operations.
  • Closing is similar in some ways to dilation in that it tends to enlarge the boundaries of foreground (bright) regions in an image and fill-in small background holes known as pepper noise. However, it is often less destructive of the original boundary shape. [Source: http://www.theobjects.com/dragonfly/dfhelp/4-
  • Morphological filters (2) that can be used for this present invention are erosion, dilation, opening and/or closing morphological filters (2).
  • erosion morphological filter (2) is preferred as the aim here is to detect, highlight and amplify shadows created by the immature oil palm trees in order to create detection points for the detecting and/or counting immature oil palm trees.
  • Detection points’ for this present invention means the mature or immature trees, specifically focusing on immature oil palm trees.
  • the erosion filter detects and highlights the shadows created from the immature oil palm trees, hence, the shadows are amplified while the overall brightness of the rest of the area of the satellite images are reduced.
  • the amplification of the shadows by the erosion filter is important to successfully detect the immature oil palm trees.
  • the user of the present invention would firstly need to obtain high-resolution satellite images of oil palm estates of interest.
  • this present invention focuses on detecting and counting immature oil palm trees, visual interpretation on the high-resolution satellite images is firstly required to identify whether the particular oil palm estate contains immature or mature trees (per Figure 11). Mature oil palm trees can be easily identified with its large crown size and canopy per Figure 10, while immature oil palm trees can be identified per Figure 2.
  • the erosion filter per method of the present invention would highlight and amplify shadows of the oil palm trees to create the necessary detection points for the detecting and/or counting immature oil palm trees per method of the present invention.
  • this method of the present invention is possible to be used for counting mature oil palm trees but common means of detecting and counting oil palm trees are done using vegetative index, hence this wouldn’t be the preferred mode for detecting and counting mature oil palm trees.
  • Vegetation indices is a remote sensing phenology studies use data gathered by satellite sensors that measure wavelengths of light absorbed and reflected by green plants. [Source: https://www.usgs.gov/core-science-systems/eros/phenology/science/vegetation-indices]
  • Dilation filters operate by making objects of interest (such as roads, water bodies, buildings etc.) more visible by filling in pixels to the objects in order to make the object as complete as possible and the satellite images are enhanced (made bigger). Due to the size of the immature oil palm trees and current best resolution of the high-resolution satellite images, the dilation filters are not preferred for detecting immature oil palm trees as the immature oil palm trees or the shadows of the immature oil palm trees could be missed out using this filtering means, hence the detection accuracy is greatly reduced. This is because dilation filters enhance other objects around the oil palm trees hence missing the immature oil palm trees due to its size.
  • objects of interest such as roads, water bodies, buildings etc.
  • the dilation filter also does not allow for the detection of shadows of the trees to generate the necessary detection points to filter out the immature oil palm trees from the high- resolution satellite images.
  • the opening filter is useful for removing small objects (such as immature trees) from an image while preserving the shape and size of larger objects (such as soil, roads, ground vegetation) in the image. This would result in the opening filter removing the trees form the image since it would be considered as small objects and preserve other objects in the image which are bigger than the trees such as hills, lake, roads, buildings and water bodies.
  • the closing filter operates similar as a dilation filter and opposite of an opening filter whereby it fills up the small holes in an image with pixels while preserving the shape and size of the objects in the said image.
  • This filter again does not allow the user of the present invention to detect shadows of the trees to generate the necessary detection points to filter out the immature oil palm trees.
  • erosion filter is preferred as the filter is able to reduce the brightness of the surrounding areas of the trees while amplifying the shadows created by the trees.
  • the amplification of the shadows is important in order to create detection points. Due to the size of the immature oil palm trees, amplification of the shadows or pixels containing the shadows would be the focus. This is because it appears in close proximity to the immature trees and is considered as the neighbouring pixels which can be identified using erosion filter to produce the detection points.
  • Image segmentation is the process of partitioning an image into multiple segments. Image segmentation is typically used to locate objects and boundaries in images. [Source: Ying Tan, in Gpu-Based Parallel Implementation of Swarm Intelligence Algorithms, 2016]
  • Watershed segmentation (3) is preferred for the purpose of the present invention because the principle of the watershed segmentation is to treat a grey image as a topographic surface whereby the bright pixels represent mountaintops and dark pixels as valleys and then to gradually floods the said image. With the water level rising, the enclosed watershed lines can form boundaries of the segments, i.e., the immature trees (per Figure 4). Seed points were then generated which represents their respective boundaries that will eventually be called detection points.
  • Morphologically filtered image was segmented by using watershed segmentation (3) to produce the detection points.
  • the detection points were the cleaned up or filtered further using point displacement (4) and point thinning (5).
  • Point displacement (4) or displacement of the detection points for the purposes of the present invention means moving and congregating detection points at small distances so that they overlap with each other to obtain congregated points within the area of interest of the immature oil palm trees.
  • Point thinning (5) or thinning of the displaced detection points for the purposes of the present invention means removing all the overlapping points from point displacement output to obtain only one point that, represents a particular tree.
  • Thinning is a morphological operation that is used to remove selected foreground pixels from binary images (black & white), similar to the operations of an erosion or opening filter.
  • foreground pixel is the black section of the image and background pixel is the white section of the image.
  • Thinning of the pixels of the satellite images is commonly applied to tidy up the output of edge detectors by reducing the lines to a single pixel thickness.
  • the thinning operation essentially is used to translate the origin of the structuring element to each possible pixel position in the image and at each such position, to compare with the underlying image pixel. If the foreground and background pixels in the structuring element exactly match foreground and background pixels in the image, then the image pixel underneath the origin of the structuring element is set to background (zero). Otherwise it is left unchanged.
  • High-resolution satellite images (1) mean high-accuracy images. High-resolution also means the pixel size of the images are small, hence, providing more detail. High resolution satellite images for this present invention means the ground sampling distance of the image is less than Im per pixel. Less than Im imagery is able to capture greater details. Therefore, high- resolution satellite images (1) are preferred for this present invention as the immature oil palm trees are much more visible for detecting and/or counting.
  • Morphological filtering method has been used for imaging in non- agriculture purposes. Hence, to the best of the knowledge of the inventors of the present invention, it is not known for any third parties to use morphological filtering means to detect and/or count immature oil palm trees in estates with any types of terrains (i.e. flat terrains, flat and hilly terrains, hilly terrains).
  • Ground sampling distance of the present invention means the distance between two consecutive pixel centres measured on the ground. The lower the value of the ground sampling distance, the higher the spatial resolution of the image.
  • a satellite image is made up of tiny squares, each of a different grey shade or colour which are called pixels (short for picture elements) and represent the relative reflected light energy recorded for that part of the image.
  • the resolution of an image is dependent upon size of pixel used to create the image. A total image is acquired when all the pixels are combined. The larger the pixel size, the lower the resolution. The smaller the pixel size, the higher the resolution.
  • the high-resolution satellite images (1) used for this present invention were obtained from the company, Planet and are orthorectified for use in the method of the present invention.
  • Orthorectification is the process of removing the effects of image perspective (tilt) and relief (terrain) effects for the purpose of creating a planimetrically correct image.
  • the resultant orthorectified image has a constant scale wherein features are represented in their 'true' positions. This allows for the accurate direct measurement of distances, angles, and areas (i.e. mensuration).
  • Orthorectified images are commonly used as in visualization tools such as Google Earth, OSSIM Planet, ArcMap, WMS, etc.
  • Each image has a ground sampling distance of 80 cm per pixel and 4 bands comprising of the colours blue, green, red and near-infrared.
  • the study area was Derawan Estate, Sarawak, Malaysia whereby the oil palm estate was planted with oil palm in the year 2018. An area of 2 ha was manually selected for the purposes and demonstration of the method of the present invention.
  • QGIS software is used for the purposes of the present invention.
  • QGIS is the leading Free and Open Source Geographic Information System (GIS).
  • Erosion morphological filter (2) erodes away the boundaries of regions of foreground pixels such as the roads and drains. This would result in a decrease in size of the pixel area while the holes within those areas becomes bigger. Referring to Figure 3, it can be seen that the erosion morphological filter (2) reduced the brightness of the image while the shadow areas created by each oil palm tree were increased. Each square shape represents an immature oil palm tree and will be further segmented to produce detection points.
  • Watershed segmentation (3) is a region-based method that looks for similarities between neighbourhood pixels and groups them into unique regions where the image is regarded as a topographic landscape with ridges and valley.
  • the bright pixels on an image surface represents mountain tops while the dark pixels represent valleys.
  • the square shapes as seen in Figure 4 has been identified as valleys via the watershed segmentation (3).
  • this method uses minima pixels to enhance and/or augment shadow of the oil palm trees to produce detection points (immature oil palm trees) instead of conventional means of using maxima pixels (tree crown) which is mainly used to detect mature oil palm trees.
  • the detection points produced using minima pixels are located within its own individual boundary which represents a tree. These points were then further processed via point displacement (4) and point thinning (5).
  • Minima pixels works best for detection of immature oil palm trees because erosion morphological filter (2) amplifies shadows which are created by the immature oil palm trees. All points produced in Figure 5 were moved and adjusted to be in the centre of its own boundary per Figure 6. After adjusting or displacing all the points to the same location, the thinning process was applied. Point thinning (4) is applied to remove overlapping detection points to ensure there will only be one point for each boundary that represents a tree per Figure 7.
  • Figure 8 illustrates the results of this method per Figure 7 placed over with the satellite image of Figure 1.
  • Figure 9 shows three study areas: (a) flat terrains, (b) flat and hilly terrains and (c) hilly terrains for accuracy assessment, with results illustrated in Table 1 below.
  • the erosion filter is able to successfully detect immature palm trees with more than 90% accuracy (recall).
  • the accuracy (recall) and the overall performance (F-Measure) decreases as the topography changes from flat to hilly terrains. This was contributed by the number of false detection points within the total number of detection points produced for hilly terrains.
  • the erosion filter still able to accurately detect immature palms between 92% to 100% for flat terrains, flat and hilly terrains and hilly terrains.
  • the overall performance can be improved by masking out the non-palm area to reduce the number of false detection points.
  • F-Measure or F-Score provides a way to combine both precision and recall into a single measure that captures both properties, giving each the same weighting. This is the harmonic mean of the two fractions - precision and recall.
  • the F-measure balances the precision and recall.
  • Precision is a metric that calculates the percentage of correct predictions for the positive class. Recall calculates the percentage of correct predictions for the positive class out of all positive predictions that could be made.
  • Palm counting for both mature and immature oil palm trees are required for optimum estate management. At current, there are established methods for detecting and/or counting mature oil palm trees using high-resolution satellite images. However, palm counting for immature oil palm trees have only been successful using drone imaging which is low in productivity and costly.
  • the inventors of the present invention have been exploring a number of means to detect and/or count immature oil palm trees not focusing on drone imaging but utilising high- resolution satellite imaging (1). There is lack of initiative in the industry in utilising high- resolution satellite images (1) to detect and/or count immature oil palm trees. Also, there is gap I void in utilising morphological filters (2) in this area as it is not widely used for high-resolution satellite imagery purposes, especially in the agriculture field and more specifically not been used or explored in detecting and/or counting immature oil palm trees.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

La présente invention concerne un procédé de détection et/ou de comptage automatique d'arbres, comprenant l'obtention d'une pluralité d'images satellites (1) d'un domaine agricole ou d'une plantation agricole, le marquage des limites d'une pluralité de zones dans la pluralité d'images satellites (1) du domaine agricole ou de la plantation agricole et l'application d'au moins un filtre morphologique (2) aux limites de la pluralité de zones à l'intérieur de la pluralité d'images satellites (1) du domaine agricole ou de la plantation agricole pour produire une pluralité d'images filtrées morphologiques comprenant de multiples ombres de points de détection.
PCT/MY2022/050074 2021-11-16 2022-08-19 Procédé de détection et/ou de comptage d'arbres WO2023090988A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
MYPI2021006820 2021-11-16
MYPI2021006820 2021-11-16

Publications (1)

Publication Number Publication Date
WO2023090988A1 true WO2023090988A1 (fr) 2023-05-25

Family

ID=83508399

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/MY2022/050074 WO2023090988A1 (fr) 2021-11-16 2022-08-19 Procédé de détection et/ou de comptage d'arbres

Country Status (1)

Country Link
WO (1) WO2023090988A1 (fr)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103353983A (zh) 2013-07-30 2013-10-16 中南大学 一种基于最大树结构的形态学滤波方法
CN104361330A (zh) 2014-12-01 2015-02-18 郑州轻工业学院 一种玉米精准施药系统的作物行识别方法
CN105551002A (zh) 2015-12-17 2016-05-04 重庆大学 一种图像形态学滤波方法
EP3125151A2 (fr) * 2015-07-31 2017-02-01 Accenture Global Services Limited Inventaire, croissance et prédiction de risque au moyen d'un traitement d'image
CN110570399A (zh) 2019-08-16 2019-12-13 中国科学院地理科学与资源研究所 一种基于时间序列遥感数据的油棕树龄测算方法
US10586105B2 (en) 2016-12-30 2020-03-10 International Business Machines Corporation Method and system for crop type identification using satellite observation and weather data
US20200125823A1 (en) 2016-12-02 2020-04-23 GEOSAT Aerospace & Technology Methods and systems for automatic object detection from aerial imagery
CN111461918A (zh) 2020-02-19 2020-07-28 北京天和本安电力科技有限公司 基于卫星光学影像的输电线路树闪隐患监测方法
CN111582035A (zh) 2020-04-07 2020-08-25 北京农业信息技术研究中心 一种果树树龄识别方法、装置、设备及存储介质

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103353983A (zh) 2013-07-30 2013-10-16 中南大学 一种基于最大树结构的形态学滤波方法
CN104361330A (zh) 2014-12-01 2015-02-18 郑州轻工业学院 一种玉米精准施药系统的作物行识别方法
EP3125151A2 (fr) * 2015-07-31 2017-02-01 Accenture Global Services Limited Inventaire, croissance et prédiction de risque au moyen d'un traitement d'image
CN105551002A (zh) 2015-12-17 2016-05-04 重庆大学 一种图像形态学滤波方法
US20200125823A1 (en) 2016-12-02 2020-04-23 GEOSAT Aerospace & Technology Methods and systems for automatic object detection from aerial imagery
US10586105B2 (en) 2016-12-30 2020-03-10 International Business Machines Corporation Method and system for crop type identification using satellite observation and weather data
CN110570399A (zh) 2019-08-16 2019-12-13 中国科学院地理科学与资源研究所 一种基于时间序列遥感数据的油棕树龄测算方法
CN111461918A (zh) 2020-02-19 2020-07-28 北京天和本安电力科技有限公司 基于卫星光学影像的输电线路树闪隐患监测方法
CN111582035A (zh) 2020-04-07 2020-08-25 北京农业信息技术研究中心 一种果树树龄识别方法、装置、设备及存储介质

Non-Patent Citations (10)

* Cited by examiner, † Cited by third party
Title
"Source: A simple method for detection and counting of oil palm trees using high-resolution multispectral satellite imagery", INTERNATIONAL JOURNAL OF REMOTE SENSING, vol. 37, no. 21, November 2016 (2016-11-01), pages 5122 - 5134
"Source: Saliency Based Segmentation of Satellite Images", JOINT ISPRS CONFERENCE, vol. 2015, March 2015 (2015-03-01), pages 25 - 27
MANCINI ADRIANO ET AL: "Soil / crop segmentation from remotely sensed data acquired by Unmanned Aerial System", 2017 INTERNATIONAL CONFERENCE ON UNMANNED AIRCRAFT SYSTEMS (ICUAS), IEEE, 13 June 2017 (2017-06-13), pages 1410 - 1417, XP033131934, DOI: 10.1109/ICUAS.2017.7991526 *
OURCE: DETECTION OF TREE CROWNS IN VERY HIGH SPATIAL RESOLUTION IMAGES, 8 June 2016 (2016-06-08)
SHAFRI HELMI Z. M. ET AL: "Semi-automatic detection and counting of oil palm trees from high spatial resolution airborne imagery", INTERNATIONAL JOURNAL OF REMOTE SENSING, vol. 32, no. 8, 29 March 2011 (2011-03-29), GB, pages 2095 - 2115, XP093002338, ISSN: 0143-1161, DOI: 10.1080/01431161003662928 *
VIEIRA GABRIEL DA SILVA ET AL: "Extending the Aerial Image Analysis from the Detection of Tree Crowns", 2019 IEEE 31ST INTERNATIONAL CONFERENCE ON TOOLS WITH ARTIFICIAL INTELLIGENCE (ICTAI), IEEE, 4 November 2019 (2019-11-04), pages 1681 - 1685, XP033713791, DOI: 10.1109/ICTAI.2019.00247 *
YADAV DEEPIKA ET AL: "Supervised Learning based Greenery region detection using Unnamed Aerial Vehicle for Smart City Application", 2019 10TH INTERNATIONAL CONFERENCE ON COMPUTING, COMMUNICATION AND NETWORKING TECHNOLOGIES (ICCCNT), IEEE, 6 July 2019 (2019-07-06), pages 1 - 7, XP033681059 *
YING TAN, GPU-BASED PARALLEL IMPLEMENTATION OF SWARM INTELLIGENCE ALGORITHMS, 2016
ZHENGRONG LI ET AL: "Towards automatic tree crown detection and delineation in spectral feature space using PCNN and morphological reconstruction", IMAGE PROCESSING (ICIP), 2009 16TH IEEE INTERNATIONAL CONFERENCE ON, IEEE, PISCATAWAY, NJ, USA, 7 November 2009 (2009-11-07), pages 1705 - 1708, XP031628341, ISBN: 978-1-4244-5653-6 *
ZHENGRONG LI: "Aerial image analysis using spiking neural networks with application to power line corridor monitoring", 1 May 2011 (2011-05-01), XP055261320, Retrieved from the Internet <URL:http://eprints.qut.edu.au/46161/1/Zhengrong_Li_Thesis.pdf> [retrieved on 20160329] *

Similar Documents

Publication Publication Date Title
Graesser et al. Detection of cropland field parcels from Landsat imagery
Hamylton et al. Evaluating techniques for mapping island vegetation from unmanned aerial vehicle (UAV) images: Pixel classification, visual interpretation and machine learning approaches
Conrad et al. Measuring rural settlement expansion in Uzbekistan using remote sensing to support spatial planning
Zhang et al. Object-oriented method for urban vegetation mapping using IKONOS imagery
Ruiz et al. A feature extraction software tool for agricultural object-based image analysis
Ke et al. A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing
Chen et al. A GEOBIA framework to estimate forest parameters from lidar transects, Quickbird imagery and machine learning: A case study in Quebec, Canada
Ke et al. A comparison of three methods for automatic tree crown detection and delineation from high spatial resolution imagery
Shafri et al. Semi-automatic detection and counting of oil palm trees from high spatial resolution airborne imagery
Khan et al. Remote sensing: an automated methodology for olive tree detection and counting in satellite images
Eckert et al. Identification and classification of structural soil conservation measures based on very high resolution stereo satellite data
S Bhagat Use of remote sensing techniques for robust digital change detection of land: A review
Clark et al. Landscape analysis using multi-scale segmentation and object-oriented classification
Vermeulen et al. Evaluation of a WorldView-2 image for soil salinity monitoring in a moderately affected irrigated area
Lu et al. Integration of optical, SAR and DEM data for automated detection of debris-covered glaciers over the western Nyainqentanglha using a random forest classifier
Ghofrani et al. Evaluating coverage changes in national parks using a hybrid change detection algorithm and remote sensing
Okubo et al. Land use/cover classification of a complex agricultural landscape using single-dated very high spatial resolution satellite-sensed imagery
Uehara et al. Time-series metrics applied to land use and land cover mapping with focus on landslide detection
Fisette et al. Methodology for a Canadian agricultural land cover classification
Nasiri et al. Integration of radar and optical sentinel images for land use mapping in a complex landscape (case study: Arasbaran Protected Area)
WO2023090988A1 (fr) Procédé de détection et/ou de comptage d&#39;arbres
Ajalkar et al. An Smart Intelligence Performance Analysis Using ANN Classifiers For Soil Color Texture Identification
Wang et al. Mapping Robinia pseudoacacia forest health in the Yellow River delta by using high-resolution IKONOS imagery and object-based image analysis
Jayasekare et al. Hybrid method for building extraction in vegetation-rich urban areas from very high-resolution satellite imagery
Huang et al. Comparing the effects of temporal features derived from synthetic time-series NDVI on fine land cover classification

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22783059

Country of ref document: EP

Kind code of ref document: A1