WebGL模型拾取——射线法二

Posted 2021-01-20 ccentry

　　这篇文章是对射线法raycaster的补充，上一篇文章主要讲的是raycaster射线法拾取模型的原理，而这篇文章着重讲使用射线法要注意的地方。首先我们来看下图。

　　我来解释一下上图中的originTriangle，这就是Triangle2三角形第一次绘制在空间中的位置，而Triangle2当前的位置是经过一系列空间变换而来的（这些位置姿态变换大多是由用户鼠标交互产生），变换矩阵就是transformMatrix。这下就引出了本文第一个重点，那就是做raycaster的时候要保证线段碰撞模型的时候一定是模型当前所处的空间位置，即已经做过transformMatrix空间（姿态，位置）变换，否则线段如果和模型之前的初始化位置求交显然没有交点，就拾取失败了。这就是做raycaster要注意的第一个重点，即射线一定要和空间变换后的模型求交。

　　接下来我们再看一张图，请看下图。

　　我们看到，射线和模型有2个交点，P2_0和P2_1，分别交四面体的前面于P2_0和交四面体的后面于P2_1。这就是我们要着重关注的本文第二个重点，即raycaster射线拾取模型过程中射线与单个模型有多个交点的问题。处理这个问题其实有很多办法，这里我们采用最简单的方式，就是距离相机（人眼）位置近者胜出的策略。计算交点的算法上一篇文章已经提到，这里不再赘述，但要说明的是，我们计算的每一个面在数据结构中都有自身模型父节点geometry，如果像上图一个四面体的geometry和射线产生了多个面相交，那我们就认为鼠标选中的是该模型geometry离相机（camera）（人眼）最近的交面上的交点。

　　对于上面2点的叙述，配合部分代码展示，是geometry空间变换的，代码如下。

Object.assign(CubeSection.prototype, {
    //重载，每一帧同步数据
    sync: function () {
        if(this._mode === "face") {//根据剖切模式管理鼠标拖拽逻辑
            if (this._selectFace) {
                let camera = this._viewer.getMainCamera();
                let last = this._mousePoints.getLast();
                let lastX = camera.getNormalizedX(last[0]);
                let lastY = camera.getNormalizedY(last[1]);
                let current = this._mousePoints.getCurrent();
                let currentX = camera.getNormalizedX(current[0]);
                let currentY = camera.getNormalizedY(current[1]);

                this._mousePoints.sync();
                let deltaX = currentX - lastX;
                let deltaY = currentY - lastY;
                if (Math.abs(deltaX) < Algorithm.EPSILON && Math.abs(deltaY) < Algorithm.EPSILON)
                    return;

                //如果面被选中，并且有移动量，需要进行剖切面移动处理
                let start = Vec3.MemoryPool.alloc();
                let end = Vec3.MemoryPool.alloc();
                camera.computeScreenToWorldNearFar(lastX, lastY, start, end, true);
                //获取起点与平面的交点
                let plane = this._cubeClip.getClipPlane(this._selectFace.getName());
                let planePt1 = Vec3.MemoryPool.alloc();
                if (Plane.intersectLine(planePt1, start, end, plane)) {
                    //将模型交点再转换到屏幕坐标上，主要为了获取z值给终点
                    let temppt = Vec3.MemoryPool.alloc();
                    camera.computeWorldToScreen(planePt1, temppt);

                    Vec3.set(temppt, currentX, currentY, temppt[2]);

                    let planePt2 = Vec3.MemoryPool.alloc();
                    camera.computeScreenToWorld(temppt, planePt2);
                    Vec3.sub(temppt, planePt2, planePt1);
                    let dist = Vec3.dot(plane, temppt);
                    this.move(dist);
                    Vec3.MemoryPool.free(planePt2);
                    Vec3.MemoryPool.free(temppt);
                }
                Vec3.MemoryPool.free(planePt1);
                Vec3.MemoryPool.free(start);
                Vec3.MemoryPool.free(end);
            }
        } else if(this._mode === "translate"){
            if(this._selectAxis) {
                let camera = this._viewer.getMainCamera();
                let last = this._mousePoints.getLast();//前一帧鼠标的XY坐标
                let lastX = camera.getNormalizedX(last[0]);
                let lastY = camera.getNormalizedY(last[1]);
                let current = this._mousePoints.getCurrent();//目前帧鼠标的XY坐标
                let currentX = camera.getNormalizedX(current[0]);
                let currentY = camera.getNormalizedY(current[1]);

                this._mousePoints.sync();//继续下一帧同步鼠标XY坐标
                let deltaX = currentX - lastX;//X偏移量
                let deltaY = currentY - lastY;//Y偏移量
                if (Math.abs(deltaX) < Algorithm.EPSILON && Math.abs(deltaY) < Algorithm.EPSILON) {
                    //如果XY偏移量都为零，就直接返回，什么操作都不做
                    return;
                }

                //坐标系轴被选中，并且有偏移量，就要移动整个包围盒子
                let start = Vec3.MemoryPool.alloc();
                let end = Vec3.MemoryPool.alloc();
                //把屏幕上的XY坐标换算到视棱台near，far截面上的XY坐标
                camera.computeScreenToWorldNearFar(lastX, lastY, start, end, true);
                //当前pick的坐标轴
                let axis = this._selectAxis;
                //near-far线段截axis坐标轴的交点
                let intersectPoint1 = Vec3.MemoryPool.alloc();
                //射线碰撞
                let intersections = this._drawActor.linesegmentIntersect(start, end);//对场景中的所有物体进行线段碰撞检测
                //遍历intersections列表，按照离相机从远到近排列
                for (let i = 0; i < intersections.length; i++) {
                    let geometry = intersections[i].getDrawable().getGeometry();
                    if (geometry && new String(geometry._name).substring(0, 4) === "axis") {
                        intersectPoint1 = intersections[i]._point;//获取到near-far线段和坐标轴的交点
                        break;
                    }
                }
                //将near-far和坐标轴的交点再转换到屏幕坐标上，主要为了获取z值给终点
                let screenPoint = Vec3.MemoryPool.alloc();
                camera.computeWorldToScreen(intersectPoint1, screenPoint);
                //screePoint(currentX, currentY, screenPoint.z)
                Vec3.set(screenPoint, currentX, currentY, screenPoint[2]);
                //鼠标移动的第二个场景坐标系里的点坐标
                let intersectPoint2 = Vec3.MemoryPool.alloc();
                //把屏幕归一化坐标转化为场景世界坐标
                camera.computeScreenToWorld(screenPoint, intersectPoint2);
                Vec3.sub(screenPoint, intersectPoint2, intersectPoint1);
                let dist = 0;
                if(this._selectAxis._name === "axisX"){
                    dist = screenPoint[0];
                }else if(this._selectAxis._name === "axisY"){
                    dist = screenPoint[1];
                }else if(this._selectAxis._name === "axisZ"){
                    dist = screenPoint[2];
                }
                this.move(dist);
                //析构向量
                Vec3.MemoryPool.free(intersectPoint1);
                Vec3.MemoryPool.free(intersectPoint2);
                Vec3.MemoryPool.free(screenPoint);
                Vec3.MemoryPool.free(start);
                Vec3.MemoryPool.free(end);
            }
        } else if(this._mode === "rotate"){

        } else if(this._mode === "scale"){

        }
    },

updateTransform: function () {
        let mat = this._cubeRoot.getMatrix();

        //重新计算坐标系模型的_matrix
        this._coordinateSection.update(this._clipBox, this._scale, this._translate, this._rotate, this._scaleMatrix, this._translateMatrix, this._rotateMatrix, mat);

        Mat4.fromScaling(this._scaleMatrix, this._scale);
        Mat4.fromTranslation(this._translateMatrix, this._translate);
        Mat4.fromQuat(this._rotateMatrix, this._rotate);

        Mat4.mul(mat, this._translateMatrix, this._rotateMatrix);
        Mat4.mul(mat, mat, this._scaleMatrix);

        //剖切面数据的变换
        this._cubeClip.resetClipPlane();
        this._cubeClip.transformClipPlane(mat);

        //包围盒子更新
        this._clipBox.setMaxValue(0.5, 0.5, 0.5);
        this._clipBox.setMinValue(-0.5, -0.5, -0.5);
        this._clipBox.transformMat4(mat);
    },

接下来是选取离相机近的交点，代码如下

//拾取物体，根据当前剖切模式选择intersections列表中的碰撞对象
    pick: function (x, y) {
        let camera = this._viewer.getMainCamera();
        let start = Vec3.MemoryPool.alloc();
        let end = Vec3.MemoryPool.alloc();
        camera.computeScreenToWorldNearFar(x, y, start, end);
        let intersections = this._drawActor.linesegmentIntersect(start, end);
        let l = intersections.length;
        if (l !== 0) {
            switch(this._mode){
                case "face" : {//面剖切
                    let intersection = intersections[0];//LineSegmentIntersection
                    let geometry = intersection.getDrawable().getGeometry();
                    if (geometry) {
                        this._selectFace = geometry;
                        this._selectFace.setStateSet(this._selectState);
                        return true;
                    }
                }
                case "translate" : {//平移剖切
                    //遍历intersections列表，按照离相机从远到近排列
                    for(var i=0; i<l; i++){
                        let geometry = intersections[i].getDrawable().getGeometry();
                        if(geometry && new String(geometry._name).substring(0, 4) === "axis"){
                            this._selectAxis = geometry;
                            this._selectAxis.setStateSet(this._selectStateAxis);
                            break;
                        }
                    }
                    return true;
                }
                case "rotate" : {//旋转剖切
                    //遍历intersections列表，按照离相机从远到近排列
                    for(var i=0; i<l; i++){
                        let geometry = intersections[i].getDrawable().getGeometry();
                        if(geometry && new String(geometry._name).substring(0, 4) === "face"){
                            this._selectAxisFace = geometry;
                            this._selectAxisFace.setStateSet(this._selectStateAxisFace);
                            break;
                        }
                    }
                    return true;
                }
                case "scale" : {//缩放剖切
                    //遍历intersections列表，按照离相机从远到近排列
                    for(var i=0; i<l; i++){
                        let geometry = intersections[i].getDrawable().getGeometry();
                        if(geometry && new String(geometry._name).substring(0, 4) === "axis"){
                            this._selectAxis = geometry;
                            this._selectAxis.setStateSet(this._selectStateAxis);
                            break;
                        }
                    }
                    return true;
                }
            }
        }
        return false;
    },

其中intersections[]交点列表是按照离相机由远到近距离排序的，intersection[i]交点离相机距离比intersection[i+1]交点离相机距离要近。这就是我们采取的离相机近交点胜出原则。

　　好了，以上就是raycaster射线拾取模型要注意的地方，如有错误，希望读者斧正，欢迎诸位同学留言。如需转载本文，请注明出处：https://www.cnblogs.com/ccentry/p/9977490.html