Spawn crops/3D game objects randomly on a mesh/plane surface

Question

I am stuck on a problem that might be quite simple yet it's getting on my nerves. I want to spawn crops on different plane shapes like in the attached screenshot.

So far my progress is that I am able to spawn crops with random distance with a certain number of rows and columns.

[ContextMenu("Generate Field")]
public void GenerateField()
    {
        GameObject wheatField = new GameObject();
        wheatField.transform.position = transform.position;
        int counter = 0;
        for(int i = 0; i < rows; i++)
        {
            for (int j = 0; j < column; j++)
            {
                xDelta = Random.Range(0.4f, 0.5f);
                yDelta = Random.Range(0.4f, 0.6f);
                float yPos = Random.Range(-0.25f, 0.25f);
                GameObject tempWheat = Instantiate(wheatPrefab, new Vector3(xDelta * i, 0, yDelta * j), Quaternion.identity
                    ,wheatField.transform);
                counter++;
            }
        }
        Debug.Log("TOTAL CROPS PLACED = " + counter);
    }

I want to do this in editor time, not runtime.

Below is the sample mesh I have where Orange, White, and Green are separate meshes.

Answer 1

You can use rejection sampling for this.

Iterate over the area covered by the mesh's bounding box, placing your rows of crops.

Each time you try to place a plant, using code similar to what you have now, first check to see whether it's on top of the mesh. You could do this by...

• Adding a MeshCollider, and calling ClosestPoint to snap the randomized crop position to a point on the mesh shape. If the snapped point is farther from your randomly chosen point than a threshold you set, consider this "outside" the mesh and skip to the next row/column cell.

• Iterating over the triangles of the mesh, and checking whether your (x,z) point is inside any of those triangles, using conventional point-in-triangle tests. If the point is not on any triangles, skip to the next row/column cell.

  You can accelerate this check with a spatial partition if needed, but for low-poly shapes like you've shown in your example that optmization might not be necessary.