我的KD树有什么问题？ （K = 2）

Question

更具体地说，我最近邻搜索有问题。我已经确认树是正确构建的，并且所有辅助函数都按预期工作。

请注意，这与标准KD树略有不同，因为我正在寻找最接近的3点。

struct Node
{
unsigned int axis;
Point* obj;

struct Node* left;
struct Node* right;
};


void _findClosest(struct Node* current, Point* to, Point** first, Point** second, Point** third, unsigned int depth)
{
if(current == NULL) return;

if(current->left == NULL && current->right == NULL)
{
            //_setOrder updates first, second, and third so that they point to
            //objects that are closest to "to", given an object and the current
            //current first, second, and third.
    _setOrder(to, current->obj, first, second, third);
    return;
}

unsigned int axis = depth % 2;
struct Node* other = NULL;

if(to->getValue(axis) < current->obj->getValue(axis))
{
    other = current->right;
    _findClosest(current->left, to, first, second, third, depth+1);
}
else
{
    other = current->left;
    _findClosest(current->right, to, first, second, third, depth+1);
}

_setOrder(to, current->obj, first, second, third);

if(other == NULL) return;

double searchToBest = _largestDistance(to, *first, *second, *third);
double searchToSplit = _distance(to, current->obj, current->axis);

    //This branch is ALWAYS taken, so this devolves into an O(n) search
if(searchToBest >= searchToSplit)
{
    _findClosest(other, to, first, second, third, depth+1);
}
}

我想我只是在某种程度上误解了数据结构/算法。

细节：
- 如果在NULL对象上调用distance函数，它们将返回std :: numberic_limits :: max（）
- 这是在KD树的根上调用的，深度为0，* first = * second = * third = NULL
-axis = 0对应于X，轴= 1对应于Y

问题在于访问每个节点，而不是看到利用树属性的预期减少量。无论缺陷是什么，它都会成为O（n）搜索。

Answer 1

这一行错了：

double searchToSplit = _distance（to，current-＆gt; obj，current-＆gt; axis）;

您需要轴，而不是当前 - ＆gt;轴。