Update lidar_tracker howto docs

docs/howto/how_to_add_a_new_lidar_detector_algorithm_cn.md

@@ -45,7 +45,7 @@ class NewLidarDetector : public BaseLidarDetector {
 }  // namespace apollo
 ```
 
-基类`base_lidar_detector`已定义好各虚函数签名，接口信息如下：
+基类 `base_lidar_detector` 已定义好各虚函数签名，接口信息如下：
 
 ```c++
 struct LidarDetectorInitOptions {

docs/howto/how_to_add_a_new_lidar_tracker_algorithm_cn.md

@@ -1,5 +1,13 @@
 # 如何添加新的lidar匹配算法
 
+Perception中的lidar数据流如下：
+![](https://github.com/ApolloAuto/apollo/blob/master/docs/specs/images/lidar_perception_data_flow.png)
+
+本篇文档所介绍的lidar检测算法位于图中的Recognition Component中。当前Recognition Component的架构如下：
+![lidar recognition](images/lidar_recognition.png)
+
+从以上结构中可以清楚地看到lidar匹配算法是位于Recognition Component的 `base_lidar_obstacle_tracking` 中的抽象成员类 `base_multi_target_tracker` 的派生类。下面将详细介绍如何基于当前结构添加新的lidar匹配算法。
+
 Apollo默认的lidar匹配算法为MlfEngine，它可以轻松更改或替换为不同的算法。本篇文档将介绍如何引入新的lidar匹配算法，添加新算法的步骤如下：
 
 1. 定义一个继承基类 `base_multi_target_tracker` 的类
@@ -11,7 +19,7 @@ Apollo默认的lidar匹配算法为MlfEngine，它可以轻松更改或替换为
 
 ## 定义一个继承基类 `base_multi_target_tracker` 的类
 
-所有的lidar匹配算法都必须继承基类`base_multi_target_tracker`，它定义了一组接口。 以下是匹配算法继承基类的示例:
+所有的lidar匹配算法都必须继承基类 `base_multi_target_tracker`，它定义了一组接口。 以下是匹配算法继承基类的示例:
 
 ```c++
 namespace apollo {
@@ -37,6 +45,48 @@ class NewLidarTracker : public BaseMultiTargetTracker {
 }  // namespace apollo
 ```
 
+基类 `base_multi_target_tracker` 已定义好各虚函数签名，接口信息如下：
+
+```c++
+struct MultiTargetTrackerInitOptions {};
+
+struct MultiTargetTrackerOptions {};
+
+struct LidarFrame {
+  // point cloud
+  std::shared_ptr<base::AttributePointCloud<base::PointF>> cloud;
+  // world point cloud
+  std::shared_ptr<base::AttributePointCloud<base::PointD>> world_cloud;
+  // timestamp
+  double timestamp = 0.0;
+  // lidar to world pose
+  Eigen::Affine3d lidar2world_pose = Eigen::Affine3d::Identity();
+  // lidar to world pose
+  Eigen::Affine3d novatel2world_pose = Eigen::Affine3d::Identity();
+  // hdmap struct
+  std::shared_ptr<base::HdmapStruct> hdmap_struct = nullptr;
+  // segmented objects
+  std::vector<std::shared_ptr<base::Object>> segmented_objects;
+  // tracked objects
+  std::vector<std::shared_ptr<base::Object>> tracked_objects;
+  // point cloud roi indices
+  base::PointIndices roi_indices;
+  // point cloud non ground indices
+  base::PointIndices non_ground_indices;
+  // secondary segmentor indices
+  base::PointIndices secondary_indices;
+  // sensor info
+  base::SensorInfo sensor_info;
+  // reserve string
+  std::string reserve;
+
+  void Reset();
+
+  void FilterPointCloud(base::PointCloud<base::PointF> *filtered_cloud,
+                        const std::vector<uint32_t> &indices);
+};
+```
+
 ## 实现新类 `NewLidarTracker`
 
 为了确保新的匹配算法能顺利工作，`NewLidarTracker`至少需要重写`base_multi_target_tracker`中定义的接口Init(),Track()和Name()。其中Init()函数负责完成加载配置文件，初始化类成员等工作；而Track()则负责实现算法的主体流程。一个具体的`NewLidarTracker.cc`实现示例如下：

docs/howto/how_to_add_a_new_lidar_tracker_algorithm_en.md

@@ -1,5 +1,13 @@
 # How to add a new lidar tracker algorithm
 
+The processing flow of lidar perception module is shown below: ：
+![](https://github.com/ApolloAuto/apollo/blob/master/docs/specs/images/lidar_perception_data_flow.png)
+
+The tracker algorithm introduced by this document is located at Recognition Component listed below. Current architecture of Recognition Component is shown：
+![lidar recognition](images/lidar_recognition.png)
+
+As we can see from above structure, lidar tracker algorithm, such as MlfEngine, is the derived class of `base_multi_target_tracker` which acts as a abstract class member of `base_lidar_obstacle_tracking` located in Recognition Component. Next, We will introduce how to add a new lidar tracker algorithm.
+
 The default tracking algorithm of Apollo is MlfEngine，which cloud be easily changed or replaced by other algorithms. This document will introduce how to add a new lidar tracker algorithm, the basic task sequence is listed below：
 
 1. Define a class that inherits `base_multi_target_tracker` 
@@ -37,6 +45,48 @@ class NewLidarTracker : public BaseMultiTargetTracker {
 }  // namespace apollo
 ```
 
+The function signature of `base_multi_target_tracker` is pre-defined：
+
+```c++
+struct MultiTargetTrackerInitOptions {};
+
+struct MultiTargetTrackerOptions {};
+
+struct LidarFrame {
+  // point cloud
+  std::shared_ptr<base::AttributePointCloud<base::PointF>> cloud;
+  // world point cloud
+  std::shared_ptr<base::AttributePointCloud<base::PointD>> world_cloud;
+  // timestamp
+  double timestamp = 0.0;
+  // lidar to world pose
+  Eigen::Affine3d lidar2world_pose = Eigen::Affine3d::Identity();
+  // lidar to world pose
+  Eigen::Affine3d novatel2world_pose = Eigen::Affine3d::Identity();
+  // hdmap struct
+  std::shared_ptr<base::HdmapStruct> hdmap_struct = nullptr;
+  // segmented objects
+  std::vector<std::shared_ptr<base::Object>> segmented_objects;
+  // tracked objects
+  std::vector<std::shared_ptr<base::Object>> tracked_objects;
+  // point cloud roi indices
+  base::PointIndices roi_indices;
+  // point cloud non ground indices
+  base::PointIndices non_ground_indices;
+  // secondary segmentor indices
+  base::PointIndices secondary_indices;
+  // sensor info
+  base::SensorInfo sensor_info;
+  // reserve string
+  std::string reserve;
+
+  void Reset();
+
+  void FilterPointCloud(base::PointCloud<base::PointF> *filtered_cloud,
+                        const std::vector<uint32_t> &indices);
+};
+```
+
 ## Implement the class `NewLidarTracker`
 
 To ensure the new tracker could function properly, `NewLidarTracker` should at least override the interface Init(), Track(), Name() defined in `base_multi_target_tracker`. Init() is resposible for config loading, class member initialization, etc. And Track() will implement the basic logic of algorithm. A concrete `NewLidarTracker.cc` example is shown：