[GIS] How to change the class of one pixel in a classified image

arcmaperdas-imaginerasterraster-calculator

I have a classified image (I did it in ERDAS) with six categories. There are some misclassified pixels (especially high-density urban category). Is there any way that I can choose one pixel and then change its class (for instance change bareland to high-density urban) in ERDAS or Arcmap?
I do appreciate your help. Thanks.

Best Answer

The simplest way is to create a new point layer for identifying individual cells which have been misclassified: for each incorrect cell, create a point at that location, and update the attribute table with the corrected value.

Then, rasterize the points to a layer with the same extent and cell size as your input layer, then merge the two with a raster calculator, doing something along the lines of:

if(rasterized_points > 0, rasterized_points, classified_input_image)

Related Solutions

[GIS] ArcObjects: modifying feature field value

Below are a couple of extension methods for IFeature/IRow that show the actual setting of a value. You'll want to check out cursors and row/feature buffers if you need better performance: http://help.arcgis.com/en/sdk/10.0/arcobjects_net/conceptualhelp/index.html#/d/0001000002rs000000.htm

the extension methods would need to be called when you are already in an edit operation (editor.StartOperation()/editor.StopOperation()): http://resources.esri.com/help/9.3/ArcGISDesktop/dotnet/abbd6e92-04f8-4957-a205-a53773fbf023.htm

    /// <summary>
    /// Sets the value on IFeature
    /// </summary>
    /// <param name="feature">The feature with the field to be set</param>
    /// <param name="fieldName">Name of the field.</param>
    /// <param name="value">The new field value</param>
    /// <param name="ignoreMissingFields">if set to <c>true</c> [ignore missing fields].</param>
    public static void SetValue(this IFeature feature, string fieldName, object value, bool ignoreMissingFields)
    {
        (feature as IRow).SetValue(fieldName, value, ignoreMissingFields);
    }

    /// <summary>
    /// Sets the value on IRow
    /// </summary>
    /// <param name="row">The row with the field to update</param>
    /// <param name="fieldName">Name of the field.</param>
    /// <param name="value">The new field value</param>
    /// <param name="ignoreMissingFields">if set to <c>true</c> [ignore missing fields].</param>
    public static void SetValue(this IRow row, string fieldName, object value, bool ignoreMissingFields)
    {
        int index = row.Fields.FindField(fieldName);

        if (index != -1)
        {
            if (row.Fields.get_Field(index).Editable)
            {
                row.set_Value(index, value);
            }
        }
        else
        {
            if (!ignoreMissingFields)
            {
                throw new System.ArgumentException("Field '{0}' Not Found.".FormatString(fieldName));
            }
        }
    }

[GIS] What are Bands, Dimensions, and Histograms in QGIS

BANDS

When working with e.g. scanned maps raster the three (four) bands are exactly the same as RGB(A) channels in e.g. GIMP. But that does not apply to other kinds of raster data.

For example, remote sensing data such as multispectral images acquired from satellites can contain (almost) any number of bands with each pixel value for a specific band corresponding to the intensity of electromagnetic radiation at a specific wavelength. With multispectral images thinking simply in terms of "color mixing and masking" is not going to be helpful: the interpretation of the bands depends on how the phenomema being researched reflect electromagnetic radiation at different wavelengts.

Or you could have a single band raster, where each pixel value corresponds to some measured or interpolated variable, such as rainfall.

DIMENSIONS

At least in QGIS the dimension of a raster is about the size of the image in pixels, i.e. its Y * X size independent of the size/and unit of one pixel. So the dimension does not change when you change the CRS, thought the measured size of the area covered by the raster might change depending on the projection.

If you change the resolution of the raster (i.e. resample it) the dimension changes, but the measured size of the raster will not change as long as the new pizel size does not leave a remainder (i.e. dimensions divided by pixel size gives a whole number).

Best Answer

Related Solutions

[GIS] ArcObjects: modifying feature field value

[GIS] What are Bands, Dimensions, and Histograms in QGIS

Related Question