{var%20f='http://v.t.sina.com.cn/share/share.php?appkey=1515056452',u=z||d.location,p=['&url=',e(u),'&title=',e(t||d.title),'&source=',e(r),'&sourceUrl=',e(l),'&content=',c||'gb2312','&pic=',e(p||'')].join('');function%20a(){if(!window.open([f,p].join(''),'mb',['toolbar=0,status=0,resizable=1,width=440,height=430,left=',(s.width-440)/2,',top=',(s.height-430)/2].join('')))u.href=[f,p].join('');};if(/Firefox/.test(navigator.userAgent))setTimeout(a,0);else%20a();})(screen,document,encodeURIComponent,'','','https://www.manongdao.com/data/attach/logo/logo.png', '推荐 We Are One 的问题《GLSL spotlight projection volume》','https://www.manongdao.com/q-799941.html','页面编码gb2312|utf-8默认gb2312'));){kind=link}
In my open source project I have setup a deferred rendering pipeline using Qt3D. So far so good, but now I'd like to move forward by adding spotlights projection volume. (e.g. as if there is smoke in the scene) Like this:
The fragment shader I'm using is at the end of the question. I've read that for each fragment I should do ray marching from the light position and find the intersections with a cone, but I have no idea how to translate this into GLSL. I can easily add a uniform with the depth map (from camera point of view) coming from the GBuffer, but I don't know if that's of any help.
Since my GLSL knowledge is very limited, please reply with actual code, not a lengthy mathematical explanation that I won't be able to understand/translate into code. Please be patient with me.
uniform sampler2D color;
uniform sampler2D position;
uniform sampler2D normal;
uniform vec2 winSize;
out vec4 fragColor;
const int MAX_LIGHTS = 102;
const int TYPE_POINT = 0;
const int TYPE_DIRECTIONAL = 1;
const int TYPE_SPOT = 2;
struct Light {
int type;
vec3 position;
vec3 color;
float intensity;
vec3 direction;
float constantAttenuation;
float linearAttenuation;
float quadraticAttenuation;
float cutOffAngle;
};
uniform Light lightsArray[MAX_LIGHTS];
uniform int lightsNumber;
void main()
{
vec2 texCoord = gl_FragCoord.xy / winSize;
vec4 col = texture(color, texCoord);
vec3 pos = texture(position, texCoord).xyz;
vec3 norm = texture(normal, texCoord).xyz;
vec3 lightColor = vec3(0.0);
vec3 s;
float att;
for (int i = 0; i < lightsNumber; ++i) {
att = 1.0;
if ( lightsArray[i].type != TYPE_DIRECTIONAL ) {
s = lightsArray[i].position - pos;
if (lightsArray[i].constantAttenuation != 0.0
|| lightsArray[i].linearAttenuation != 0.0
|| lightsArray[i].quadraticAttenuation != 0.0) {
float dist = length(s);
att = 1.0 / (lightsArray[i].constantAttenuation + lightsArray[i].linearAttenuation * dist + lightsArray[i].quadraticAttenuation * dist * dist);
}
s = normalize( s );
if ( lightsArray[i].type == TYPE_SPOT ) {
if ( degrees(acos(dot(-s, normalize(lightsArray[i].direction))) ) > lightsArray[i].cutOffAngle)
att = 0.0;
}
} else {
s = normalize(-lightsArray[i].direction);
}
float diffuse = max( dot( s, norm ), 0.0 );
lightColor += att * lightsArray[i].intensity * diffuse * lightsArray[i].color;
}
fragColor = vec4(col.rgb * lightColor, col.a);
}
This is how a spotlight looks like with the original shader above:
[EDIT - SOLVED] Thanks to Rabbid76 excellent answer and precious support
This is the modified code to see the cone projection:
#version 140
uniform sampler2D color;
uniform sampler2D position;
uniform sampler2D normal;
uniform vec2 winSize;
out vec4 fragColor;
const int MAX_LIGHTS = 102;
const int TYPE_POINT = 0;
const int TYPE_DIRECTIONAL = 1;
const int TYPE_SPOT = 2;
struct Light {
int type;
vec3 position;
vec3 color;
float intensity;
vec3 direction;
float constantAttenuation;
float linearAttenuation;
float quadraticAttenuation;
float cutOffAngle;
};
uniform Light lightsArray[MAX_LIGHTS];
uniform int lightsNumber;
uniform mat4 inverseViewMatrix; // defined by camera position, camera target and up vector
void main()
{
vec2 texCoord = gl_FragCoord.xy / winSize;
vec4 col = texture(color, texCoord);
vec3 pos = texture(position, texCoord).xyz;
vec3 norm = texture(normal, texCoord).xyz;
vec3 lightColor = vec3(0.0);
vec3 s;
// calculate unprojected fragment position on near plane and line of sight relative to view
float nearZ = -1.0;
vec3 nearPos = vec3( (texCoord.x - 0.5) * winSize.x / winSize.y, texCoord.y - 0.5, nearZ ); // 1.0 is camera near
vec3 los = normalize( nearPos );
// ray definition
vec3 O = vec3( inverseViewMatrix * vec4( 0.0, 0.0, 0.0, 1.0 ) ); // translation part of the camera matrix, which is equal to the camera position
vec3 D = (length(pos) > 0.0) ? normalize(pos - O) : (mat3(inverseViewMatrix) * los);
for (int i = 0; i < lightsNumber; ++i)
{
float att = 1.0;
if ( lightsArray[i].type == TYPE_DIRECTIONAL )
{
s = normalize( -lightsArray[i].direction );
}
else
{
s = lightsArray[i].position - pos;
if (lightsArray[i].type != TYPE_SPOT
&& (lightsArray[i].constantAttenuation != 0.0
|| lightsArray[i].linearAttenuation != 0.0
|| lightsArray[i].quadraticAttenuation != 0.0))
{
float dist = length(s);
att = 1.0 / (lightsArray[i].constantAttenuation + lightsArray[i].linearAttenuation * dist + lightsArray[i].quadraticAttenuation * dist * dist);
}
s = normalize( s );
if ( lightsArray[i].type == TYPE_SPOT )
{
// cone definition
vec3 C = lightsArray[i].position;
vec3 V = normalize(lightsArray[i].direction);
float cosTh = cos( radians(lightsArray[i].cutOffAngle) );
// ray - cone intersection
vec3 CO = O - C;
float DdotV = dot( D, V );
float COdotV = dot( CO, V );
float a = DdotV * DdotV - cosTh * cosTh;
float b = 2.0 * (DdotV * COdotV - dot( D, CO ) * cosTh * cosTh);
float c = COdotV * COdotV - dot( CO, CO ) * cosTh * cosTh;
float det = b * b - 4.0 * a * c;
// find intersection
float isIsect = 0.0;
vec3 isectP = vec3(0.0);
if ( det >= 0.0 )
{
vec3 P1 = O + (-b - sqrt(det)) / (2.0 * a) * D;
vec3 P2 = O + (-b + sqrt(det)) / (2.0 * a) * D;
float isect1 = step( 0.0, dot(normalize(P1 - C), V) );
float isect2 = step( 0.0, dot(normalize(P2 - C), V) );
if ( isect1 < 0.5 )
{
P1 = P2;
isect1 = isect2;
}
if ( isect2 < 0.5 )
{
P2 = P1;
isect2 = isect1;
}
isectP = (length(P1 - O) < length(P2 - O)) ? P1 : P2;
isIsect = mix( isect2, 1.0, isect1 );
if ( length(pos) != 0.0 && length(isectP - O) > length(pos - O))
isIsect = 0.0;
}
float dist = length( isectP - C.xyz );
float limit = degrees(acos(dot(-s, normalize(lightsArray[i].direction))) );
if (isIsect > 0 || limit <= lightsArray[i].cutOffAngle)
{
att = 1.0 / dot( vec3( 1.0, dist, dist * dist ),
vec3(lightsArray[i].constantAttenuation,
lightsArray[i].linearAttenuation,
lightsArray[i].quadraticAttenuation) );
}
else
att = 0.0;
}
}
float diffuse = max( dot( s, norm ), 0.0 );
lightColor += att * lightsArray[i].intensity * diffuse * lightsArray[i].color;
}
fragColor = vec4(col.rgb * lightColor, col.a);
}
Uniforms passed to the shader are:
qml: lightsArray[0].type = 0
qml: lightsArray[0].position = QVector3D(0, 10, 0)
qml: lightsArray[0].color = #ffffff
qml: lightsArray[0].intensity = 0.8
qml: lightsArray[0].constantAttenuation = 1
qml: lightsArray[0].linearAttenuation = 0
qml: lightsArray[0].quadraticAttenuation = 0
qml: lightsArray[1].type = 2
qml: lightsArray[1].position = QVector3D(0, 3, 0)
qml: lightsArray[1].color = #008000
qml: lightsArray[1].intensity = 0.5
qml: lightsArray[1].constantAttenuation = 2
qml: lightsArray[1].linearAttenuation = 0
qml: lightsArray[1].quadraticAttenuation = 0
qml: lightsArray[1].direction = QVector3D(-0.573576, -0.819152, 0)
qml: lightsArray[1].cutOffAngle = 15
qml: lightsNumber = 2
Screenshot:
For a primitive visualization of the light cone of a spot light, you have to do a intersection of the line of sight and the light cone.
The following algorithm works in a perpectiv view and the caluclations ar done in view (eye) sapce. The algorithm does not care about the geometry of the scene and does not do any depth test or shadow test, it only is a overlayerd visualization of the light cone.
The line of sight in a perspective view can be deifned by a points and a direction. Since the calculations is done in view (eye) space, the point is the point of view (the origin of the view frustum) which is
vec3(0.0).The direction can easily be determined, by the intersection of the line of sight and the near plane of the camera frustum. This can easily be calculated if the projected XY-coordinate of the fragment is known in normalized device coordinates (the lower left point is (-1,-1) and the upper right point is (1,1) see the answer to this question).
The light cone is defined by the origin of the light source, the direction where the light source points to, and the full angle of the light cone. The position and the direction have to be up in view space. The angle has to be set up in radians.
How to calculate the intersection point(s) of a ray and a cone can be found in the answer to Stackoverflow question Points of intersection of vector with cone and in the following paper: Intersection of a ray and a cone.
The following code calculates a intersection of a ray and a cone as defined above. The result point is stored in
isectP. The variableisIsectof the typefloatis set to 1.0 if there is a intersection, and is set to 0.0 else.For the full GLSL code, see the following WebGL example: