API Reference

rsimg_io

readTiff (class):
writeTiff: write out the .tiff formated remote sensing image.

readTiff(path_in)

Description: Write out the .tiff formated remote sensing image

Param:: path_in
Returns:: img(np.array), extent(tuple), projection(str), dimentions(tuple)

writeTiff(im_data, im_geotrans, im_geosrs, path_out)

Write out the .tiff formated remote sensing image

Param:: im_data, im_geotrans, im_geosrs, path_out
Retrun:: None

geo_imgxy

get_utm_zone(lon)

Description: Get utm zone from the given wgs84 coordinates.

Param:: lon: the given longitute, should be in the range of [-180, 180]
Retrun:: utm_zone_number

coor2coor(srs_from, srs_to, x, y)

Description: Transform coordinates from srs_from to srs_to

Param:

srs_from: EPSG number (e.g., 4326, 3031)

srs_to: EPSG number (e.g., 4326, 3031)

x: x-coord corresponding to srs_from

y: y-coord corresponding to srs_from

Retrun:

(x, y): x-coord and y-coord corresponding to srs_to

geo2imagexy(x, y, gdal_trans, integer=True)

Description: from georeferenced location (i.e., lon, lat) to image location(col,row).

Note: the coordinate system should be same between x/y and gdal_trans.

Param:

gdal_proj: obtained by gdal.Open() and .GetGeoTransform(), or by geotif_io.readTiff()[‘geotrans’]

x: project or georeferenced x, i.e.,lon

y: project or georeferenced y, i.e., lat

Retrun:

utm_zone_number

imagexy2geo(row, col, gdal_trans)

Description: from image location(col,row) to georeferenced location (i.e., lon, lat).

Param:

img_gdal: GDAL data (read by gdal.Open()

row: row of the input image (dataset)

col: col of the input image (dataset)

Retrun:

x: geographical coordinates at x-axis (left up of pixel)

y: geographical coordinates at y-axis (left up of pixel)

img_normalize

img_normalize(img, max_bands, min_bands)

Description: image normalization.

Param:

max_bands: list, max value of each the band of image.

min_bands: list, min values of each the band of image.

image: np.array.

Retrun:

image_nor: the normalized image.

img2patch

class img2patch(img, patch_size, edge_overlay)

Description: Conversion between the remote sensing image and patches

Param:

img: np.array()

patch_size: size of the patch

edge_overlay: an even number, single-side overlay of the neighboring images.

Method:

.toPatch():

Description: convert img to patches.

param:

None

retrun:

patch_list: contains all generated patches.

start_list: contains all start positions(row, col) of the generated patches.

Method:

.higher_patch_crop(higher_patch_size)

Description: crop the higher-scale patch

param:: higher_patch_size: int, the lager patch size compared the low-scale patch size.
retrun:: higher_patch_list: list, contains higher-scale patches corresponding to the lower-scale patches.

Method:

.toImage(patch_list)

Description: merge patches into one image.

param:: patch_list: list of the all patches.
retrun:: img_array: the merged image by patches

class crop2size(img, channel_first=False)

Description: randomly crop corresponding to specific size.

Param:

img: np.array().

channel_first: True or False.

Method:

.toSize(size=(256, 256))

Descrition: randomly crop corresponding to specific size

param:: size: tuble/list, (height, width)
retrun:: patch, the cropped patch from the image.

Method:

.toScales(scales=(2048, 512, 256))

Description: randomly crop multiple-scale patches (from high to low) from remote sensing image.

param:: scales: tuple or list (high scale -> low scale)
return:: patches_group_down: list of multiscale patches.

class crop2extent(extent, size_target=None)

Description: Crop image with specific geographical extent.

Method:

.img2extent(path_img, path_save=None)

param:

path_img: string, the image path to be croped.

size_target: size to which image should be croped list/tuple, (row, col)

path_save: string, the path for output saving.

retrun:

img_croped: the croped image, np.array()

imgShow

imgShow(img, ax=None, extent=None, color_bands=(2, 1, 0), clip_percent=2, per_band_clip=False, focus_per=None, focus_pix=None)

Description: show the single image.

Param:

img: (row, col, band) or (row, col), DN range should be in [0,1]

extent: list, the coordinates of the extent.

num_bands: a list/tuple, [red_band,green_band,blue_band]

clip_percent: for linear strech, value within the range of 0-100.

per_band_clip: if True, the band values will be clipped by each band respectively.

focus_per: list, [up_start_percent,down_end_percent,left_start_percent, right_end_percent]; 0 < value < 1

focus_pix: list, [up_start_pixel,down_end_pixel,left_start_pixel, right_end_pixel]; 0 < value < (width or height) of the image

Retrun:

None

imsShow(img_list, img_name_list, clip_list=None, color_bands_list=None, axis=True, row=None, col=None)

Description: show the multiple images.

Param:

img_list: containes all images

img_names_list: image names corresponding to the images

clip_list: percent clips (histogram) corresponding to the images

color_bands_list: color bands combination corresponding to the images

row, col: the row and col of the figure

axis: True or False

Retrun:

None

lay_stack

lay_stack(path_imgs, path_out, union=True, res=None)

Description: layer stacking of the multiple bands of image.

Param:

path_imgs: list, contains the paths of bands/imgs to be stacked

path_out: str, the output path of the layer stacked image

union: bool, if true, the output extent is the extents union of input images. Otherwise, the output extent is the extents intersection of input images.

res: resolution of the layer stacked image.

Retrun:

imgs_stacked: np.array(), the stacked image.

metric_proc

smooth(y, window=31, num_sam=None)

Description: smooth the sequential data.

Param:

y: the sequential data.

window: the smooth window used for averating y.

num_sam: the number of sampled data.

Retrun:

x: the smoothed sequential data of x-axis.

y: the smoothed sequential data of y-axis.

metric

acc_matrix(cla_map, truth_map=None, sam_pixel=None, id_label=None)

Description: calculate the accuracy matrix.

Param:

cla_map: classification result of the full image.

truth_map: truth image (either truth_map or sam_pixel should be given).

sam_pixel: np.array, (num_samples,3), col 1,2,3 are the row,col and label.

id_label: 0/1/2/…, Calculating producer’s or user’s accuracy for target class.

Retrun:

acc_oa: overall accuracy

confus_mat: confusion matrix

acc_miou(cla_map, truth_map, labels=None)

Description: calculate the miou metric.

Param:

cla_map: classification result of the full image

truth_map: truth image (either truth_map or sam_pixel should be given)

labels: a list, the class id for calculating. e.g., [0,1,2]

Retrun:

miou: MIoU metric

oa_binary(pred, truth)

Description: calculate overall accuracy (2-class classification) for each batch.

Param:

pred: 4D tensor

truth: 4D tensor

Retrun:

oa: overall accuracy

miou_binary(pred, truth)

Description: calculate miou (2-class classification) for each batch.

Param:

pred: 4D tensor

truth: 4D tensor

Retrun:

miou: miou metric

raster_vec

raster2vec(raster_path, output_path, dn_values)

Description: Read input band with Rasterio.

Param:

raster_path: raster path

output_path: ouput vector path

dn_values: list, consist of the raster value to be vectorization

Retrun:

miou: miou metric

vec2mask(path_vec, path_raster, path_save=None)

Description: generate/save mask file using the vector file(e.g.,.shp,.gpkg).

Param:

path_vec: str, path of the vector data.

path_raster: str, path of the raster data.

path_save: str, path to save.

Retrun:

mask: np.array().

transform_time

date2doy(year, month, day, hour=0, minute=0)

Description: convert year-month-day-hour-minute to doy (day-of-year)

Param:: year month: 0~12 day: 0~31 hour minute
Retrun:: doy: day of the year.

doy2date(year, doy)

Description: Convert doy(day-of-year) to year-month-day-hour-minute formate the function returns the month and the day of the month.

Param:

year

doy

Retrun:

month

day: the day of the month

dt64_to_dyr(dt64)

Description: Convert datetime64 to decimal year format. e.g., ‘2020-05-23T03:25:22.959373696’ -> 2020.3907103825136.

Param:: dt64: np.datetime64 format time
Retrun:: dt_float: float, the decimal date.

second_to_dyr(time_second, time_start='2000-01-01 00:00:00.0')

Description: convert time (second format) to decimal year. This function suitable for the jason data, sentinel-3 data, and the cryosat2 data for time conversion.

Param:: time_second: seconds from the time start.
Retrun:: time_second_dyr: decimal date.