diff --git a/data/__init__.py b/data/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/qodana.yaml b/qodana.yaml
new file mode 100644
index 0000000000..38e12241aa
--- /dev/null
+++ b/qodana.yaml
@@ -0,0 +1,41 @@
+#-------------------------------------------------------------------------------#
+#               Qodana analysis is configured by qodana.yaml file               #
+#             https://www.jetbrains.com/help/qodana/qodana-yaml.html            #
+#-------------------------------------------------------------------------------#
+version: "1.0"
+
+#Specify inspection profile for code analysis
+profile:
+  name: qodana.starter
+
+#Enable inspections
+#include:
+#  - name: <SomeEnabledInspectionId>
+
+#Disable inspections
+#exclude:
+#  - name: <SomeDisabledInspectionId>
+#    paths:
+#      - <path/where/not/run/inspection>
+
+#Execute shell command before Qodana execution (Applied in CI/CD pipeline)
+#bootstrap: sh ./prepare-qodana.sh
+
+#Install IDE plugins before Qodana execution (Applied in CI/CD pipeline)
+#plugins:
+#  - id: <plugin.id> #(plugin id can be found at https://plugins.jetbrains.com)
+
+# Quality gate. Will fail the CI/CD pipeline if any condition is not met
+# severityThresholds - configures maximum thresholds for different problem severities
+# testCoverageThresholds - configures minimum code coverage on a whole project and newly added code
+# Code Coverage is available in Ultimate and Ultimate Plus plans
+#failureConditions:
+#  severityThresholds:
+#    any: 15
+#    critical: 5
+#  testCoverageThresholds:
+#    fresh: 70
+#    total: 50
+
+#Specify Qodana linter for analysis (Applied in CI/CD pipeline)
+linter: jetbrains/qodana-<linter>:2025.2
diff --git a/src/Car_Perception_Automatic_Sensing/.gitignore b/src/Car_Perception_Automatic_Sensing/.gitignore
new file mode 100644
index 0000000000..f3285da82c
--- /dev/null
+++ b/src/Car_Perception_Automatic_Sensing/.gitignore
@@ -0,0 +1,30 @@
+# Python 相关
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.venv/
+env/
+*.env
+
+# PyCharm 相关
+.idea/
+*.iml
+*.iws
+*.ipr
+out/
+
+# 项目无关目录/文件
+../../.idea/
+
+# 数据与模型（后续会用到）
+data/raw/
+data/kitti/
+*.npy
+*.pth
+train/checkpoints/
+infer/results/
+
+# OS 相关
+.DS_Store
+Thumbs.db
\ No newline at end of file
diff --git a/src/Car_Perception_Automatic_Sensing/README.md b/src/Car_Perception_Automatic_Sensing/README.md
index a93adc2a1c..7602125d1c 100644
--- a/src/Car_Perception_Automatic_Sensing/README.md
+++ b/src/Car_Perception_Automatic_Sensing/README.md
@@ -1,2 +1,28 @@
-无人车 无人车动态障碍物避障算法  无人车动态障碍物避障的核心目标是：在实时感知环境中移动障碍物（行人和车辆）的基础上，快速规划出无碰撞、平滑且符合运动学约束的安全路径
-环境：python3.13
\ No newline at end of file
+# Car_Perception_Automatic_Sensing
+自动驾驶环境感知自动感知系统，支持车辆、行人、车道线等目标的检测与分割。
+
+## 功能模块
+- 数据加载与预处理（支持 KITTI/Waymo 数据集）
+- 深度学习模型搭建（CNN/Transformer 骨干网络）
+- 模型训练与验证
+- 推理可视化与结果导出## 
+- 目录结构
+Car_Perception_Automatic_Sensing/
+- ├── data/ 
+- ├── model/ 
+- ├── loss/ 
+- ├── train/ # 训练脚本目录
+- ├── infer/ # 推理脚本目录
+- ├── utils/ # 工具类目录
+- ├── test/ # 测试用例目录
+- ├── requirements.txt # 依赖配置
+- └── README.md # 项目文档
+
+## 环境要求
+Python >= 3.8，PyTorch >= 2.0.0
+
+## 快速开始
+1. 克隆仓库
+2. 安装依赖
+3. 准备数据集
+4. 运行训练脚本
\ No newline at end of file
diff --git a/src/Car_Perception_Automatic_Sensing/data/__init__.py b/src/Car_Perception_Automatic_Sensing/data/__init__.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/src/Car_Perception_Automatic_Sensing/data/data_loader.py b/src/Car_Perception_Automatic_Sensing/data/data_loader.py
new file mode 100644
index 0000000000..f600a3e5e9
--- /dev/null
+++ b/src/Car_Perception_Automatic_Sensing/data/data_loader.py
@@ -0,0 +1,80 @@
+"""
+数据加载器核心模块
+支持自动驾驶数据集的加载、样本管理与批次分发
+"""
+import os
+import numpy as np
+import torch
+from torch.utils.data import Dataset, DataLoader
+from typing import Optional, List
+
+
+class AutoDriveDataset(Dataset):
+    """自动驾驶感知数据集基类（所有自定义数据集需继承此类）"""
+    def __init__(self, data_root: str, split: str = "train", transform: Optional[object] = None):
+        self.data_root = os.path.abspath(data_root)
+        self.split = split
+        self.transform = transform
+        self.sample_paths: List[str] = []  # 存储所有样本路径
+        self._load_sample_paths()  # 加载样本列表
+
+    def _load_sample_paths(self):
+        """加载样本路径（抽象方法，子类实现）"""
+        raise NotImplementedError("请在子类中实现 _load_sample_paths 方法")
+
+    def __len__(self) -> int:
+        """返回样本总数"""
+        return len(self.sample_paths)
+
+    def __getitem__(self, idx: int):
+        """获取单个样本（抽象方法，子类实现）"""
+        raise NotImplementedError("请在子类中实现 __getitem__ 方法")
+
+
+def build_dataloader(
+    dataset: Dataset,
+    batch_size: int,
+    shuffle: bool = True,
+    num_workers: int = 4,
+    drop_last: bool = False
+) -> DataLoader:
+    """构建PyTorch DataLoader（跨系统兼容）"""
+    # Windows系统默认关闭多进程
+    if os.name == "nt":
+        num_workers = 0
+
+    return DataLoader(
+        dataset=dataset,
+        batch_size=batch_size,
+        shuffle=shuffle,
+        num_workers=num_workers,
+        pin_memory=True,
+        drop_last=drop_last
+    )
+
+
+if __name__ == "__main__":
+    # 框架测试
+    pass
+# 新增导入
+import cv2
+from typing import Tuple
+
+# 在 AutoDriveDataset 类中新增方法
+def _load_image(self, img_rel_path: str) -> Tuple[np.ndarray, Tuple[int, int, int]]:
+    """
+    读取图像并转换为RGB格式
+    :param img_rel_path: 图像相对路径（基于data_root）
+    :return: (RGB图像数组, 图像形状(h, w, c))
+    """
+    img_abs_path = os.path.join(self.data_root, img_rel_path)
+    # 路径校验
+    if not os.path.exists(img_abs_path):
+        raise FileNotFoundError(f"图像不存在：{img_abs_path}")
+    # 读取图像（忽略透明通道）
+    img_bgr = cv2.imread(img_abs_path, cv2.IMREAD_COLOR)
+    if img_bgr is None:
+        raise ValueError(f"无法读取图像（损坏/格式不支持）：{img_abs_path}")
+    # BGR转RGB
+    img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
+    return img_rgb, img_rgb.shape
\ No newline at end of file
diff --git a/src/Car_Perception_Automatic_Sensing/data/preprocess.py b/src/Car_Perception_Automatic_Sensing/data/preprocess.py
new file mode 100644
index 0000000000..7e3b4ab936
--- /dev/null
+++ b/src/Car_Perception_Automatic_Sensing/data/preprocess.py
@@ -0,0 +1,40 @@
+"""数据预处理工具：归一化、标准化"""
+import numpy as np
+
+
+def normalize_image(img: np.ndarray, target_range: tuple = (0, 1)) -> np.ndarray:
+    """
+    图像像素值归一化到指定范围
+    :param img: 输入图像（np.ndarray, uint8/float32）
+    :param target_range: 目标范围，默认(0,1)
+    :return: 归一化后的图像（float32）
+    """
+    img_float = img.astype(np.float32)
+    min_val = img_float.min()
+    max_val = img_float.max()
+    # 避免除零
+    if max_val - min_val < 1e-6:
+        return np.zeros_like(img_float) + target_range[0]
+    # 归一化计算
+    normalized = (img_float - min_val) / (max_val - min_val)
+    normalized = normalized * (target_range[1] - target_range[0]) + target_range[0]
+    return normalized
+
+
+def standardize_image(img: np.ndarray, mean: list = None, std: list = None) -> np.ndarray:
+    """
+    图像标准化（减均值、除标准差）
+    :param img: 输入RGB图像（np.ndarray, (h,w,3)）
+    :param mean: 通道均值，默认ImageNet均值
+    :param std: 通道标准差，默认ImageNet标准差
+    :return: 标准化后的图像
+    """
+    if mean is None:
+        mean = [0.485, 0.456, 0.406]
+    if std is None:
+        std = [0.229, 0.224, 0.225]
+
+    img_float = img.astype(np.float32) / 255.0  # 先归一化到0-1
+    for i in range(3):
+        img_float[..., i] = (img_float[..., i] - mean[i]) / std[i]
+    return img_float
\ No newline at end of file
diff --git a/src/Car_Perception_Automatic_Sensing/data_loader.py b/src/Car_Perception_Automatic_Sensing/data_loader.py
deleted file mode 100644
index eba4500e65..0000000000
--- a/src/Car_Perception_Automatic_Sensing/data_loader.py
+++ /dev/null
@@ -1,73 +0,0 @@
-"""
-数据加载器模块
-负责自动驾驶感知任务的数据集加载、数据预处理分发等功能
-支持图像类数据集（如KITTI、COCO自动驾驶子集）的加载
-"""
-import numpy as np
-import torch
-from torch.utils.data import Dataset, DataLoader
-
-
-class AutoDriveDataset(Dataset):
-    """
-    自动驾驶感知数据集基类
-    所有自定义数据集需继承此类并实现抽象方法
-    """
-    def __init__(self, data_root: str, split: str = "train", transform=None):
-        """
-        初始化数据集
-        :param data_root: 数据集根目录路径
-        :param split: 数据集划分（train/val/test）
-        :param transform: 数据增强/预处理变换
-        """
-        self.data_root = data_root
-        self.split = split
-        self.transform = transform
-        self.sample_list = []  # 存储样本路径/索引的列表
-
-        # 后续将实现：加载样本列表
-        self._load_sample_list()
-
-    def _load_sample_list(self):
-        """
-        加载数据集样本列表（抽象方法，需子类实现）
-        """
-        raise NotImplementedError("子类必须实现 _load_sample_list 方法")
-
-    def __len__(self):
-        """
-        返回数据集样本总数
-        """
-        return len(self.sample_list)
-
-    def __getitem__(self, idx: int):
-        """
-        根据索引获取单个样本（抽象方法，需子类实现）
-        :param idx: 样本索引
-        :return: 处理后的图像和标注数据
-        """
-        raise NotImplementedError("子类必须实现 __getitem__ 方法")
-
-
-def build_dataloader(dataset: Dataset, batch_size: int, shuffle: bool = True, num_workers: int = 4):
-    """
-    构建数据加载器
-    :param dataset: 实例化的 Dataset 对象
-    :param batch_size: 批次大小
-    :param shuffle: 是否打乱样本顺序
-    :param num_workers: 数据加载进程数
-    :return: DataLoader 对象
-    """
-    dataloader = DataLoader(
-        dataset=dataset,
-        batch_size=batch_size,
-        shuffle=shuffle,
-        num_workers=num_workers,
-        pin_memory=True  # 加速GPU数据传输
-    )
-    return dataloader
-
-
-if __name__ == "__main__":
-    # 测试代码框架（后续可完善）
-    pass
\ No newline at end of file
diff --git a/src/Car_Perception_Automatic_Sensing/main.py b/src/Car_Perception_Automatic_Sensing/main.py
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/src/__init__.py b/src/__init__.py
new file mode 100644
index 0000000000..e69de29bb2