代码：

  def ncdensityrdd(): Unit ={

    val conf = new SparkConf().setMaster("local").setAppName("Kernel Density")
    val sc = new SparkContext(conf)
    val tl = TileLayout(7, 4, 100, 100)
    val ld = LayoutDefinition(extent, tl)

    val kernelWidth=9
    def ptfToExtent[D](p: PointFeature[D]) = pointFeatureToExtent(kernelWidth, ld, p)
    def pointFeatureToExtent[D](kwidth: Double, ld: LayoutDefinition, ptf: PointFeature[D]): Extent = {
      val p = ptf.geom

      Extent(p.x - kwidth * ld.cellwidth / 2,
        p.y - kwidth * ld.cellheight / 2,
        p.x + kwidth * ld.cellwidth / 2,
        p.y + kwidth * ld.cellheight / 2)
    }

    def ptfToSpatialKey[D](ptf: PointFeature[D]): Seq[(SpatialKey,PointFeature[D])] = {
      val ptextent = ptfToExtent(ptf)
      val gridBounds = ld.mapTransform(ptextent)

      for {
        tup <- gridBounds.coordsIter.toSeq
        if tup._2 < tl.totalRows
        if tup._1 < tl.totalCols
      } yield (SpatialKey(tup._1,tup._2), ptf)
    }

    val pointRdd = sc.parallelize(pts, 10)

    def stampPointFeature(
                           tile: MutableArrayTile,
                           tup: (SpatialKey, PointFeature[Double])
                         ): MutableArrayTile = {
      val (spatialKey, pointFeature) = tup
      val tileExtent = ld.mapTransform(spatialKey)
      val re = RasterExtent(tileExtent, tile)
      val result = tile.copy.asInstanceOf[MutableArrayTile]
      val kernelWidth: Int = 9

      /* Gaussian kernel with std. deviation 1.5, amplitude 25 */
      val kern: Kernel = Kernel.gaussian(kernelWidth, 1.5, 25)
      KernelStamper(result, kern)
        .stampKernelDouble(re.mapToGrid(pointFeature.geom), pointFeature.data)

      result
    }

    def sumTiles(t1: MutableArrayTile, t2: MutableArrayTile): MutableArrayTile = {
      Adder(t1, t2).asInstanceOf[MutableArrayTile]
    }

    val tileRdd: RDD[(SpatialKey, Tile)] =
      pointRdd
        .flatMap(ptfToSpatialKey)
        .mapPartitions({ partition =>
          partition.map { case (spatialKey, pointFeature) =>
            (spatialKey, (spatialKey, pointFeature))
          }
        }, preservesPartitioning = true)
        .aggregateByKey(ArrayTile.empty(DoubleCellType, ld.tileCols, ld.tileRows))(stampPointFeature, sumTiles)
      .mapValues{ tile: MutableArrayTile => tile.asInstanceOf[Tile] }

//    tileRdd.foreach(t=>{
//      GeoTiff(t._2, extent, LatLng).write("/cephfs/test/density/%s%s.tif".format(t._1.row,t._1.col))
//    })

    val metadata = TileLayerMetadata(DoubleCellType,
      ld,
      ld.extent,
      LatLng,
      KeyBounds(SpatialKey(0,0),
        SpatialKey(ld.layoutCols-1,
          ld.layoutRows-1)))

    val resultRdd = ContextRDD(tileRdd, metadata)

    resultRdd.foreach(t=>{
            GeoTiff(t._2, extent, LatLng).write("/cephfs/test/density/%s%s.tif".format(t._1.row,t._1.col))
          })

    val stitchtile=resultRdd.stitch.tile
    GeoTiff(stitchtile, extent, LatLng).write("/cephfs/test/density/all.tif")
    sc.stop()
  }

  object Adder extends LocalTileBinaryOp {
    def combine(z1: Int, z2: Int) = {
      if (isNoData(z1)) {
        z2
      } else if (isNoData(z2)) {
        z1
      } else {
        z1 + z2
      }
    }

    def combine(r1: Double, r2:Double) = {
      if (isNoData(r1)) {
        r2
      } else if (isNoData(r2)) {
        r1
      } else {
        r1 + r2
      }
    }
  }