diff --git a/MEMORY.md b/MEMORY.md
index 24b07c5..85e62b8 100644
--- a/MEMORY.md
+++ b/MEMORY.md
@@ -236,6 +236,13 @@
 - 线路渲染继续复用 `power_line_tower.altitude_m` 作为高度来源；回填后在 `raw_extra_json.elevation` 写入溯源信息（dataset/sample_distance/sampled_at）。
 - 一期采样策略为 CSV 点集最近邻（非栅格插值），用于先打通管理与回填闭环；默认推荐回填模式 `fill_null_only`，避免覆盖人工高程。
 - 高频通知 topic 为 `admin.elevation`，线路联动通知沿用 `admin.power-lines`。
+- 高程数据文件格式已扩展为：`csv/img/tif/tiff`。
+  - `csv`：继续使用点集最近邻采样。
+  - `img/tif/tiff`：使用栅格像元采样（按杆塔经纬度取值，必要时自动做 CRS 转换）。
+- 栅格实现口径：
+  - 运行依赖 `rasterio`（随 `api` 依赖安装）。
+  - 分析阶段 `sample_count` 使用 `width * height`（超大栅格上限截断为 `2_147_483_647`），`bbox` 直接取源栅格 `bounds`。
+  - 非 WGS84 CRS 会返回告警说明，提醒边界坐标单位可能不是经纬度。
 - 用户名列口径：历史环境存在 `users.username` 与 `users.user_name` 双形态；运行时通过 `USER_USERNAME_COLUMN`（`username`/`user_name`）与目标库对齐，避免启动阶段关系预加载触发 `UndefinedColumn`。
 - 密码列口径：历史环境存在 `users.password` 与 `users.password_hash` 双形态；运行时通过 `USER_PASSWORD_COLUMN`（`password`/`password_hash`）与目标库对齐，避免启动阶段关系预加载触发 `UndefinedColumn`。
 
diff --git a/api/app/services/elevation_service.py b/api/app/services/elevation_service.py
index b566d41..8e4d831 100644
--- a/api/app/services/elevation_service.py
+++ b/api/app/services/elevation_service.py
@@ -4,6 +4,8 @@ import asyncio
 import csv
 import io
 from dataclasses import dataclass
+from pathlib import Path
+from tempfile import NamedTemporaryFile
 from typing import Any
 
 from fastapi import HTTPException, status
@@ -35,6 +37,14 @@ ELEVATION_TOPIC = "admin.elevation"
 POWER_LINES_TOPIC = "admin.power-lines"
 CSV_ENCODINGS = ("utf-8-sig", "utf-8", "gbk", "latin-1")
 NEAREST_MATCH_MAX_DISTANCE_M = 2000.0
+ELEVATION_FILE_EXT_FORMAT_MAP = {
+    ".csv": "csv",
+    ".img": "img",
+    ".tif": "tif",
+    ".tiff": "tiff",
+}
+RASTER_FILE_FORMATS = {"img", "tif", "tiff"}
+MAX_SAMPLE_COUNT_INT = 2_147_483_647
 
 
 @dataclass
@@ -44,6 +54,26 @@ class ElevationSamplePoint:
     altitude_m: float
 
 
+@dataclass
+class _OpenedRasterDataset:
+    rasterio: Any
+    dataset: Any
+    temp_path: str
+
+    def __enter__(self) -> "_OpenedRasterDataset":
+        return self
+
+    def __exit__(self, exc_type: Any, exc: Any, tb: Any) -> bool:
+        try:
+            self.dataset.close()
+        finally:
+            try:
+                Path(self.temp_path).unlink(missing_ok=True)
+            except Exception:
+                pass
+        return False
+
+
 def serialize_dataset(item: ElevationDataset) -> ElevationDatasetSummary:
     return ElevationDatasetSummary(
         id=item.id,
@@ -157,16 +187,18 @@ def create_dataset(
     if get_dataset_by_code(db, payload.code):
         return None
 
-    _ensure_dataset_file_exists(db, mount_code=payload.mount_code, file_path=payload.file_path)
+    normalized_file_path = payload.file_path.strip()
+    file_format = _detect_file_format(normalized_file_path)
+    _ensure_dataset_file_exists(db, mount_code=payload.mount_code, file_path=normalized_file_path)
 
     now = utcnow()
     item = ElevationDataset(
         code=payload.code.strip(),
         name=payload.name.strip(),
         source=_normalize_str(payload.source),
-        file_format="csv",
+        file_format=file_format,
         mount_code=payload.mount_code.strip(),
-        file_path=payload.file_path.strip(),
+        file_path=normalized_file_path,
         resolution_m=payload.resolution_m,
         status="active",
         notes=_normalize_str(payload.notes),
@@ -236,8 +268,7 @@ def analyze_dataset(
     if item.status != "active":
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="高程数据集未启用")
 
-    sample_points, warnings = _load_dataset_points(db, item)
-    stats = _compute_dataset_stats(sample_points)
+    stats, warnings = _analyze_dataset_content(db, item)
 
     item.sample_count = stats["sample_count"]
     item.bbox_min_lon = stats["bbox_min_lon"]
@@ -399,18 +430,30 @@ def execute_apply_job(job_id: str) -> None:
             )
             return
 
-        points, warnings = _load_dataset_points(db, dataset)
-        if warnings:
-            warning_note = "; ".join(warnings[:5])
+        file_format = _resolve_dataset_file_format(dataset)
+        if file_format == "csv":
+            points, warnings = _load_dataset_points(db, dataset)
+            stats = _apply_points_to_line_towers(
+                db,
+                line_id=line.id,
+                dataset=dataset,
+                mode=job.mode,
+                points=points,
+            )
+        elif file_format in RASTER_FILE_FORMATS:
+            stats, warnings = _apply_raster_to_line_towers(
+                db,
+                line_id=line.id,
+                dataset=dataset,
+                mode=job.mode,
+            )
         else:
-            warning_note = None
-        stats = _apply_points_to_line_towers(
-            db,
-            line_id=line.id,
-            dataset=dataset,
-            mode=job.mode,
-            points=points,
-        )
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=f"不支持的高程文件格式: {file_format}",
+            )
+
+        warning_note = "; ".join(warnings[:5]) if warnings else None
         job.updated_tower_count = stats["updated_tower_count"]
         job.skipped_tower_count = stats["skipped_tower_count"]
         job.missing_geo_count = stats["missing_geo_count"]
@@ -512,6 +555,22 @@ def _compute_dataset_stats(points: list[ElevationSamplePoint]) -> dict[str, floa
     }
 
 
+def _analyze_dataset_content(
+    db: Session,
+    dataset: ElevationDataset,
+) -> tuple[dict[str, float | int], list[str]]:
+    file_format = _resolve_dataset_file_format(dataset)
+    if file_format == "csv":
+        points, warnings = _load_dataset_points(db, dataset)
+        return _compute_dataset_stats(points), warnings
+    if file_format in RASTER_FILE_FORMATS:
+        return _compute_raster_stats(db, dataset)
+    raise HTTPException(
+        status_code=status.HTTP_400_BAD_REQUEST,
+        detail=f"不支持的高程文件格式: {file_format}",
+    )
+
+
 def _apply_points_to_line_towers(
     db: Session,
     *,
@@ -560,6 +619,7 @@ def _apply_points_to_line_towers(
             "dataset_code": dataset.code,
             "sample_method": "nearest",
             "sample_distance_m": round(distance_m, 3),
+            "sample_distance_source": "computed",
             "sampled_at": utcnow().isoformat(),
         }
         tower.raw_extra_json = raw_extra
@@ -624,6 +684,120 @@ def _haversine_distance_m(
     return 2 * r * math.asin(min(1.0, math.sqrt(h)))
 
 
+def _detect_file_format(file_path: str) -> str:
+    extension = Path(file_path).suffix.lower()
+    file_format = ELEVATION_FILE_EXT_FORMAT_MAP.get(extension)
+    if not file_format:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=f"不支持的高程文件类型: {extension or 'unknown'}，仅支持 .csv/.img/.tif/.tiff",
+        )
+    return file_format
+
+
+def _resolve_dataset_file_format(dataset: ElevationDataset) -> str:
+    declared = (dataset.file_format or "").strip().lower()
+    detected = _detect_file_format(dataset.file_path)
+    if declared and declared in ELEVATION_FILE_EXT_FORMAT_MAP.values():
+        if declared == detected:
+            return declared
+        if declared in {"img", "tif", "tiff"} and detected in RASTER_FILE_FORMATS:
+            return detected
+    return detected
+
+
+def _open_raster_dataset(
+    db: Session,
+    dataset: ElevationDataset,
+) -> _OpenedRasterDataset:
+    file_format = _resolve_dataset_file_format(dataset)
+    if file_format not in RASTER_FILE_FORMATS:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=f"当前文件不是栅格高程文件: {dataset.file_path}",
+        )
+
+    try:
+        import rasterio
+    except ImportError as exc:
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="服务未安装 rasterio，暂不支持 IMG/TIF 高程文件",
+        ) from exc
+
+    mount = _require_mount(db, dataset.mount_code)
+    driver = _build_driver_or_400(mount)
+    try:
+        read_result = driver.read_file(dataset.file_path)
+    except StoragePathNotFoundError as exc:
+        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=f"高程数据文件不存在: {dataset.file_path}") from exc
+    except StorageInvalidPathError as exc:
+        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from exc
+
+    suffix = Path(dataset.file_path).suffix.lower() or ".img"
+    temp_path = ""
+    try:
+        with NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
+            tmp.write(read_result.content)
+            temp_path = tmp.name
+        opened = rasterio.open(temp_path)
+        return _OpenedRasterDataset(
+            rasterio=rasterio,
+            dataset=opened,
+            temp_path=temp_path,
+        )
+    except Exception as exc:
+        if temp_path:
+            try:
+                Path(temp_path).unlink(missing_ok=True)
+            except Exception:
+                pass
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=f"高程栅格文件解析失败: {dataset.file_path}",
+        ) from exc
+
+
+def _append_non_wgs84_bounds_warning(*, rasterio: Any, src: Any) -> str | None:
+    src_crs = src.crs
+    if src_crs is None:
+        return "栅格缺少 CRS 定义，默认按 WGS84 经度/纬度采样"
+    try:
+        src_crs_obj = rasterio.crs.CRS.from_user_input(src_crs)
+    except Exception:
+        return "栅格 CRS 无法识别，默认按 WGS84 经度/纬度采样，建议先校验源数据"
+    if src_crs_obj.to_string() in {"EPSG:4326", "OGC:CRS84"}:
+        return None
+    if bool(getattr(src_crs_obj, "is_geographic", False)):
+        return None
+    return (
+        f"栅格 CRS 为 {src_crs_obj.to_string()}，数据集边界框基于该投影坐标，"
+        "回填时会自动从 WGS84 坐标转换后采样"
+    )
+
+
+def _is_masked_value(value: Any) -> bool:
+    try:
+        import numpy as np
+    except ImportError:
+        return False
+    return bool(np.ma.is_masked(value))
+
+
+def _almost_equal(a: float, b: float) -> bool:
+    return abs(a - b) <= 1e-6
+
+
+def _is_finite_number(value: float) -> bool:
+    import math
+
+    return math.isfinite(value)
+
+
+def _is_point_within_bounds(*, x: float, y: float, left: float, right: float, bottom: float, top: float) -> bool:
+    return left <= x <= right and bottom <= y <= top
+
+
 def _pick_float(row: dict[str, Any], keys: list[str]) -> float | None:
     for key in keys:
         value = row.get(key)
@@ -642,6 +816,151 @@ def _decode_csv_bytes(content: bytes) -> str:
     raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="CSV 编码不受支持")
 
 
+def _compute_raster_stats(
+    db: Session,
+    dataset: ElevationDataset,
+) -> tuple[dict[str, float | int], list[str]]:
+    warnings: list[str] = []
+    with _open_raster_dataset(db, dataset) as opened:
+        rasterio = opened.rasterio
+        src = opened.dataset
+        bounds = src.bounds
+        warning_text = _append_non_wgs84_bounds_warning(rasterio=rasterio, src=src)
+        if warning_text:
+            warnings.append(warning_text)
+
+        width = int(src.width or 0)
+        height = int(src.height or 0)
+        sample_count = width * height
+        if sample_count > MAX_SAMPLE_COUNT_INT:
+            sample_count = MAX_SAMPLE_COUNT_INT
+
+        return (
+            {
+                "sample_count": sample_count,
+                "bbox_min_lon": float(bounds.left),
+                "bbox_max_lon": float(bounds.right),
+                "bbox_min_lat": float(bounds.bottom),
+                "bbox_max_lat": float(bounds.top),
+            },
+            warnings,
+        )
+
+
+def _apply_raster_to_line_towers(
+    db: Session,
+    *,
+    line_id: str,
+    dataset: ElevationDataset,
+    mode: str,
+) -> tuple[dict[str, int], list[str]]:
+    towers = db.execute(
+        select(LineTower)
+        .where(LineTower.line_id == line_id)
+        .order_by(LineTower.seq_no.asc(), LineTower.id.asc())
+    ).scalars().all()
+
+    updated_tower_count = 0
+    skipped_tower_count = 0
+    missing_geo_count = 0
+    unmatched_count = 0
+    warnings: list[str] = []
+
+    with _open_raster_dataset(db, dataset) as opened:
+        rasterio = opened.rasterio
+        src = opened.dataset
+        warning_text = _append_non_wgs84_bounds_warning(rasterio=rasterio, src=src)
+        if warning_text:
+            warnings.append(warning_text)
+        src_crs = src.crs
+        band_nodata = src.nodatavals[0] if src.nodatavals else None
+
+        for tower in towers:
+            if tower.longitude is None or tower.latitude is None:
+                missing_geo_count += 1
+                continue
+            if mode == "fill_null_only" and tower.altitude_m is not None:
+                skipped_tower_count += 1
+                continue
+
+            lon = float(tower.longitude)
+            lat = float(tower.latitude)
+            transformed_lon = lon
+            transformed_lat = lat
+
+            if src_crs and str(src_crs) not in {"EPSG:4326", "OGC:CRS84"}:
+                try:
+                    xs, ys = rasterio.warp.transform(
+                        "EPSG:4326",
+                        src_crs,
+                        [lon],
+                        [lat],
+                    )
+                    transformed_lon = float(xs[0])
+                    transformed_lat = float(ys[0])
+                except Exception:
+                    unmatched_count += 1
+                    continue
+
+            if not _is_point_within_bounds(
+                x=transformed_lon,
+                y=transformed_lat,
+                left=float(src.bounds.left),
+                right=float(src.bounds.right),
+                bottom=float(src.bounds.bottom),
+                top=float(src.bounds.top),
+            ):
+                unmatched_count += 1
+                continue
+
+            try:
+                sampled = next(src.sample([(transformed_lon, transformed_lat)], masked=True), None)
+            except Exception:
+                sampled = None
+
+            if sampled is None or len(sampled) == 0:
+                unmatched_count += 1
+                continue
+
+            value = sampled[0]
+            if _is_masked_value(value):
+                unmatched_count += 1
+                continue
+            if band_nodata is not None and _almost_equal(float(value), float(band_nodata)):
+                unmatched_count += 1
+                continue
+
+            altitude = float(value)
+            if not _is_finite_number(altitude):
+                unmatched_count += 1
+                continue
+
+            tower.altitude_m = round(altitude, 3)
+            raw_extra = dict(tower.raw_extra_json or {})
+            raw_extra["elevation"] = {
+                "dataset_id": dataset.id,
+                "dataset_code": dataset.code,
+                "sample_method": "raster_pixel",
+                "sample_distance_m": 0.0,
+                "sample_distance_source": "pixel_lookup",
+                "sampled_at": utcnow().isoformat(),
+            }
+            tower.raw_extra_json = raw_extra
+            tower.update_date = utcnow()
+            updated_tower_count += 1
+
+    db.commit()
+    return (
+        {
+            "updated_tower_count": updated_tower_count,
+            "skipped_tower_count": skipped_tower_count,
+            "missing_geo_count": missing_geo_count,
+            "unmatched_count": unmatched_count,
+        },
+        warnings,
+    )
+
+
 def _normalize_str(value: Any) -> str | None:
     if value is None:
         return None
diff --git a/api/pyproject.toml b/api/pyproject.toml
index 865c657..76bd6be 100644
--- a/api/pyproject.toml
+++ b/api/pyproject.toml
@@ -15,4 +15,5 @@ dependencies = [
   "python-multipart>=0.0.20,<1.0.0",
   "boto3>=1.40.0,<2.0.0",
   "celery[redis]>=5.5.0,<6.0.0",
+  "rasterio>=1.4.0,<2.0.0",
 ]
diff --git a/api/requirements.txt b/api/requirements.txt
index 665dc61..8a30421 100644
--- a/api/requirements.txt
+++ b/api/requirements.txt
@@ -15,3 +15,4 @@ python-multipart==0.0.20
 boto3==1.40.59
 httpx==0.28.1
 celery[redis]==5.5.3
+rasterio==1.4.3
diff --git a/memory/2026-05-01.md b/memory/2026-05-01.md
index dc51286..0e7b137 100644
--- a/memory/2026-05-01.md
+++ b/memory/2026-05-01.md
@@ -385,3 +385,37 @@
   - 当前实现使用 CSV 点集“最近邻采样”，适合先跑通管理与回填流程；不是严格栅格插值方案。
   - 未引入 GDAL/rasterio，部署更稳，但精度依赖 CSV 样本密度与坐标质量。
   - 回填默认允许 `overwrite_all`，存在覆盖人工高程风险；前端默认展示“仅填空（推荐）”。
+
+## Work Log - 高程管理支持 IMG/TIF 导入与回填（2026-05-01）
+
+- 背景：
+  - 用户上传的高程数据为 `.img`（栅格），现有实现仅支持 CSV 点集，无法直接用于高程回填任务。
+
+- 本次改动（最小闭环）：
+  - 后端服务能力扩展（文件格式识别 + 栅格采样）：
+    - 文件：`api/app/services/elevation_service.py`
+    - 数据集创建时按扩展名自动识别并落库 `file_format`（支持 `.csv/.img/.tif/.tiff`）。
+    - 新增“按格式分流”执行：
+      - `csv`：沿用现有最近邻点集采样逻辑。
+      - `img/tif/tiff`：新增栅格像元采样逻辑，按杆塔经纬度写回 `power_line_tower.altitude_m`。
+    - 新增栅格分析逻辑：
+      - 从栅格读取 `width/height/bounds` 回写 `sample_count/bbox`。
+      - 对非 WGS84 CRS 增加告警，并在回填时自动执行坐标转换（WGS84 -> 栅格 CRS）。
+    - 回填溯源扩展：
+      - `raw_extra_json.elevation.sample_method` 对栅格标记为 `raster_pixel`。
+      - 增加 `sample_distance_source` 字段（CSV 为 `computed`，栅格为 `pixel_lookup`）。
+  - 依赖补齐：
+    - 文件：`api/requirements.txt`，新增 `rasterio==1.4.3`。
+    - 文件：`api/pyproject.toml`，新增 `rasterio>=1.4.0,<2.0.0`。
+  - 前端文案更新：
+    - 文件：`web/src/app/admin/elevation/page.tsx`
+    - 页面提示更新为支持 `CSV/IMG/TIF/TIFF`，空态与示例路径同步为栅格可用口径。
+
+- 验证：
+  - `python3 -m compileall api/app` -> 通过。
+  - `npm run build:web` -> 通过。
+  - 构建产物包含路由 `/admin/elevation`。
+
+- 风险与影响：
+  - `.img/.tif` 回填依赖 `rasterio`（及底层 GDAL 运行时），部署环境需确保镜像能成功安装该依赖。
+  - 栅格 bbox 直接来自源栅格 CRS；非经纬度坐标系场景会返回告警，便于识别与后续治理。
diff --git a/web/src/app/admin/elevation/page.tsx b/web/src/app/admin/elevation/page.tsx
index 3717b59..64e1e60 100644
--- a/web/src/app/admin/elevation/page.tsx
+++ b/web/src/app/admin/elevation/page.tsx
@@ -451,8 +451,15 @@ export default function AdminElevationPage() {
           </Space>
         )}
       >
+        <Alert
+          type="info"
+          showIcon
+          message="支持文件格式：CSV（点集）/ IMG / TIF / TIFF（栅格）"
+          description="CSV 需包含经度、纬度、高程列；IMG/TIF 会按塔杆经纬度直接采样。"
+          className="mb-4"
+        />
         {datasets.length === 0 ? (
-          <Empty description="暂无高程数据集，请先上传 CSV 并创建数据集。" />
+          <Empty description="暂无高程数据集，请先上传 CSV/IMG/TIF 并创建数据集。" />
         ) : (
           <Table<ElevationDatasetSummary>
             rowKey={(row) => row.id}
@@ -510,7 +517,7 @@ export default function AdminElevationPage() {
             <Input placeholder="dem_china_90m_v1" />
           </Form.Item>
           <Form.Item name="name" label="名称" rules={[{ required: true, message: "请输入名称" }]}>
-            <Input placeholder="中国90米DEM（CSV）" />
+            <Input placeholder="中国90米DEM（IMG）" />
           </Form.Item>
           <Form.Item name="source" label="来源">
             <Input placeholder="中科院地理空间数据云" />
@@ -519,7 +526,7 @@ export default function AdminElevationPage() {
             <Input placeholder="main" />
           </Form.Item>
           <Form.Item name="file_path" label="文件路径" rules={[{ required: true, message: "请输入文件路径" }]}>
-            <Input placeholder="/elevation/datasets/china90m.csv" />
+            <Input placeholder="/elevation/datasets/china90m.img" />
           </Form.Item>
           <Form.Item name="resolution_m" label="分辨率（米）">
             <InputNumber className="w-full" min={1} max={10000} />