"""
Tobii 单摄 DMS+OMS 架构概念

核心思想：
1. 单摄像头覆盖整个座舱
2. 统一 AI 模型同时处理 DMS 和 OMS
3. 输出统一的人员状态向量
"""

from dataclasses import dataclass
from typing import List, Optional
from enum import Enum


class PersonRole(Enum):
    DRIVER = "driver"
    PASSENGER = "passenger"
    REAR_LEFT = "rear_left"
    REAR_RIGHT = "rear_right"
    UNKNOWN = "unknown"


@dataclass
class PersonState:
    """单个人员状态"""
    role: PersonRole
    present: bool = False
    gaze_pitch: float = 0.0     # 俯仰角 (度)
    gaze_yaw: float = 0.0       # 偏航角 (度)
    gaze_confidence: float = 0.0
    fatigue_level: float = 0.0   # 0-1
    distraction_level: float = 0.0  # 0-1
    posture: str = "normal"      # normal, leaning, turned, slumped
    posture_confidence: float = 0.0
    age_group: str = "adult"     # adult, child, infant
    seatbelt_on: bool = False
    hands_on_wheel: bool = False  # 仅驾驶员
    emotion: str = "neutral"


@dataclass
class CabinState:
    """座舱整体状态（单摄方案输出）"""
    persons: List[PersonState]
    driver_readiness: float = 1.0  # 0-1
    cabin_occupancy: int = 0
    cross_seat_interaction: bool = False  # 驾驶员是否与乘员互动
    timestamp: float = 0.0


class SingleCameraDMSOMS:
    """
    Tobii 单摄 DMS+OMS 概念模型
    
    核心优势：
    1. 统一理解：一个模型理解所有人员
    2. 跨座位交互：检测驾驶员与乘员的互动
    3. 低集成复杂度：单摄像头单接口
    """
    
    def __init__(self):
        self.persons = []
    
    def process_frame(self, image) -> CabinState:
        """
        处理单帧图像，输出座舱状态
        
        步骤：
        1. 人员检测和定位
        2. 每人状态估计（视线、疲劳、姿态）
        3. 跨座位交互分析
        4. 驾驶员就绪度评估
        """
        # Step 1: 检测所有人员
        persons = self._detect_persons(image)
        
        # Step 2: 估计每人状态
        for person in persons:
            person.gaze_pitch, person.gaze_yaw, person.gaze_confidence = \
                self._estimate_gaze(image, person)
            person.fatigue_level = self._estimate_fatigue(image, person)
            person.distraction_level = self._estimate_distraction(
                image, person
            )
            person.posture = self._classify_posture(image, person)
            person.seatbelt_on = self._check_seatbelt(image, person)
        
        # Step 3: 跨座位交互
        driver = next(
            (p for p in persons if p.role == PersonRole.DRIVER), None
        )
        if driver and driver.gaze_yaw > 30:
            # 驾驶员看向侧后方
            cross_interaction = True
        else:
            cross_interaction = False
        
        # Step 4: 驾驶员就绪度
        if driver:
            readiness = self._compute_readiness(driver)
        else:
            readiness = 0.0
        
        return CabinState(
            persons=persons,
            driver_readiness=readiness,
            cabin_occupancy=len(persons),
            cross_seat_interaction=cross_interaction
        )
    
    def _detect_persons(self, image):
        """检测所有人员"""
        # 实际实现使用目标检测模型
        return []
    
    def _estimate_gaze(self, image, person):
        """视线估计"""
        return 0.0, 0.0, 0.0
    
    def _estimate_fatigue(self, image, person):
        """疲劳估计"""
        return 0.0
    
    def _estimate_distraction(self, image, person):
        """分心估计"""
        return 0.0
    
    def _classify_posture(self, image, person):
        """姿态分类"""
        return "normal"
    
    def _check_seatbelt(self, image, person):
        """安全带检测"""
        return False
    
    def _compute_readiness(self, driver: PersonState) -> float:
        """
        计算驾驶员就绪度
        
        结合疲劳、分心、视线、姿态
        """
        score = 1.0
        score -= driver.fatigue_level * 0.4
        score -= driver.distraction_level * 0.3
        if driver.posture in ['slumped', 'turned']:
            score -= 0.2
        if not driver.hands_on_wheel:
            score -= 0.1
        return max(score, 0.0)


# ========== 测试场景 ==========

if __name__ == "__main__":
    model = SingleCameraDMSOMS()
    
    # 模拟场景：驾驶员疲劳+后座有儿童
    driver = PersonState(
        role=PersonRole.DRIVER,
        present=True,
        gaze_pitch=-5, gaze_yaw=10,
        gaze_confidence=0.92,
        fatigue_level=0.65,
        distraction_level=0.15,
        posture="normal",
        hands_on_wheel=True,
        emotion="tired"
    )
    
    child = PersonState(
        role=PersonRole.REAR_LEFT,
        present=True,
        gaze_pitch=0, gaze_yaw=0,
        posture="normal",
        age_group="child",
        seatbelt_on=True
    )
    
    cabin = CabinState(
        persons=[driver, child],
        driver_readiness=model._compute_readiness(driver),
        cabin_occupancy=2,
        cross_seat_interaction=False
    )
    
    print("=== 单摄 DMS+OMS 场景 ===")
    print(f"  乘员数: {cabin.cabin_occupancy}")
    print(f"  驾驶员就绪度: {cabin.driver_readiness:.2f}")
    for p in cabin.persons:
        print(f"  {p.role.value}: "
              f"疲劳={p.fatigue_level:.0%} "
              f"分心={p.distraction_level:.0%} "
              f"姿态={p.posture}")

"""
全息感知：舱内 + 舱外 ADAS 融合

这是 Tobii 描述的 Evolving Cabin 终极形态
"""

class HolisticSensing:
    """
    全息感知引擎
    
    融合舱内感知 + 舱外 ADAS 上下文
    判断驾驶员行为的"含义"而非仅检测行为
    """
    
    def interpret_behavior(self, driver_state, adas_context):
        """
        行为解释：不是所有"不看路"都是分心
        
        场景1：驾驶员看向右后视镜 + ADAS检测到右侧有车
               → 判定：合理检查行为，不是分心
        
        场景2：驾驶员看向右后视镜 + ADAS无变道意图
               → 判定：可能分心，需要警告
        
        场景3：驾驶员闭眼0.5秒 + 高速直路
               → 判定：微睡眠，一级警告
        
        场景4：驾驶员闭眼0.5秒 + 前方紧急制动
               → 判定：危险！紧急接管
        """
        gaze_yaw = driver_state.gaze_yaw
        has_lane_change_intent = adas_context.get('lane_change_intent', False)
        forward_collision_risk = adas_context.get('collision_risk', 0)
        
        if abs(gaze_yaw) > 30:
            if has_lane_change_intent:
                return {
                    'behavior': 'mirror_check',
                    'is_safe': True,
                    'action': 'none',
                    'reason': '合理的并道前检查'
                }
            else:
                return {
                    'behavior': 'distraction',
                    'is_safe': False,
                    'action': 'warning_l1',
                    'reason': '无变道意图的视线偏离'
                }
        
        if driver_state.fatigue_level > 0.5:
            if forward_collision_risk > 0.5:
                return {
                    'behavior': 'fatigue_emergency',
                    'is_safe': False,
                    'action': 'emergency_takeover',
                    'reason': '疲劳+前方危险，紧急接管'
                }
            else:
                return {
                    'behavior': 'fatigue_warning',
                    'is_safe': False,
                    'action': 'warning_l2',
                    'reason': '疲劳但无即时危险'
                }
        
        return {
            'behavior': 'normal',
            'is_safe': True,
            'action': 'none',
            'reason': '正常驾驶'
        }