Question

我想通过它的置信度来丰富PointCloud。

首先，我创建三个FloatBuffer，并根据阈值放置单个点。

private FloatBuffer makeFLoatBuffer(ArrayList<Float> conf) {
    FloatBuffer confFB = FloatBuffer.allocate(conf.size());
    for(Float value : conf){
        confFB.put(value);
    }

    confFB.rewind();
    return confFB;
}

private ArrayList<FloatBuffer> splitPointCloudOnConfidence(PointCloud cloud) {
    ArrayList<Float> conf1 = new ArrayList<>();
    ArrayList<Float> conf2 = new ArrayList<>();
    ArrayList<Float> conf3 = new ArrayList<>();
    ArrayList<FloatBuffer> floatBufferArray = new ArrayList<>();

    FloatBuffer buf_temp = cloud.getPoints();
    int buf_temp_limit = buf_temp.limit();
    for (int i = 0; i < buf_temp_limit / 4; i++) {

        float x = buf_temp.get(i * 4);
        float y = buf_temp.get(i * 4 + 1);
        float z = buf_temp.get(i * 4 + 2);
        float confidence = buf_temp.get(i * 4 + 3);

        if(confidence < 1.0 && confidence > 0.5){
            conf1.add(x);
            conf1.add(y);
            conf1.add(z);
            conf1.add(confidence);
        } else if(confidence < 0.5 && confidence > 0.2) {
            conf2.add(x);
            conf2.add(y);
            conf2.add(z);
            conf2.add(confidence);
        } else {
            conf3.add(x);
            conf3.add(y);
            conf3.add(z);
            conf3.add(confidence);
        }
    }

    floatBufferArray.add(makeFLoatBuffer(conf1));
    floatBufferArray.add(makeFLoatBuffer(conf2));
    floatBufferArray.add(makeFLoatBuffer(conf3));

    return floatBufferArray;
}

我还创建了三个节点，这些节点负责可视化PointCloud。为此，我使用了该类的稍加修改的版本 https://github.com/claywilkinson/arcore-android-sdk/blob/sceneform-samples/samples/cloud_anchor_java/app/src/main/java/com/google/ar/core/examples/java/cloudanchor/sceneform/PointCloudNode.java

pointCloudNodeTestRed = new PointCloudNode_v2(this, color_red);
pointCloudNodeTestRed.setEnabled(true);
pointCloudNodeTestYellow = new PointCloudNode_v2(this, color_yellow);
pointCloudNodeTestYellow.setEnabled(true);
pointCloudNodeTestGreen = new PointCloudNode_v2(this, color_green);
pointCloudNodeTestGreen.setEnabled(true);

fragment.getArSceneView().getScene().addChild(pointCloudNodeTestRed);
fragment.getArSceneView().getScene().addChild( pointCloudNodeTestYellow);
fragment.getArSceneView().getScene().addChild( pointCloudNodeTestGreen);

然后在onUpdateListener中，将特定数据馈送到三个节点。

PointCloud pC = frame.acquirePointCloud();

ArrayList<FloatBuffer> bufHolder = splitPointCloudOnConfidence(pC);

pointCloudNodeTestRed.update(pC, bufHolder.get(2));
pointCloudNodeTestYellow.update(pC,  bufHolder.get(1));
pointCloudNodeTestGreen.update(pC,  bufHolder.get(0));

pC.release();

我的问题一般。有没有更好的方法来实现这种可视化？

Answer 1

我的理解是load不再由Google维护。最好有一种方法可以使用SceneForm和ARCore轻松地完成此操作作为GLES20示例文件的修改版本，如下方的链接，但是所有点都用多种颜色而不是一种颜色

https://github.com/google-ar/arcore-android-sdk/blob/54861e38cceb406021fc8ece495f82387376af62/samples/augmented_faces_java/app/src/main/java/com/google/ar/core/examples/java/common/rendering/PointCloudRenderer.java

通过置信度值对Android Sceneform PointCloud进行分类

1 个答案: