ImageAnalyzer ML Kit 边界框错误排列

如何解决ImageAnalyzer ML Kit 边界框错误排列

我有一个简单的布局：

 <RelativeLayout
    android:id="@+id/myP"
    android:layout_width="match_parent"
    android:layout_height="match_parent"
    >

    <androidx.camera.view.PreviewView
        android:id="@+id/mPreviewView"
        android:layout_width="match_parent"
        android:layout_height="match_parent"/>

    <com.example.dochjavatestimplementation.pkgActivity.ExtendedImageView
        android:id="@+id/imageViewOmgShowIt"
        android:layout_width="match_parent"
        android:layout_height="match_parent"
        android:visibility="visible" />

</RelativeLayout>

previewView 用于显示相机，ExtendedImageView 用于显示显示找到的对象rect。

我的图像分析器：

imageAnalysis = new ImageAnalysis.Builder()
 .setTargetResolution(new Size(mPreviewView.getWidth(),mPreviewView.getHeight()))
 .build();

imageAnalysis.setAnalyzer(executor,new PaperImageAnalyser());

PaperImageAnalyser：

public class PaperImageAnalyser implements ImageAnalysis.Analyzer {


ObjectDetectorOptions options = new ObjectDetectorOptions.Builder()
    .setDetectorMode(ObjectDetectorOptions.STREAM_MODE)
    .enableClassification()  // Optional
    .build();

ObjectDetector objectDetector = ObjectDetection.getClient(options);



@Override
public void analyze(@NonNull ImageProxy imageProxy) {
@SuppressLint("UnsafeExperimentalUsageError") Image mediaimage = imageProxy.getimage();

if (mediaimage != null) {
    Inputimage image =Inputimage.fromMediaimage(mediaimage,imageProxy.getimageInfo().getRotationdegrees());

    objectDetector.process(image)
        .addOnSuccessListener(
            new OnSuccessListener<List<Detectedobject>>() {
                @Override
                public void onSuccess(List<Detectedobject> detectedobjects) {
        
                    for (Detectedobject detectedobject : detectedobjects) {
                        Rect boundingBox = detectedobject.getBoundingBox();
                        Cameraimp.imageViewOmgShowIt.drawFoundobj(boundingBox);
                    }

                }
            })
        .addOnFailureListener(
            new OnFailureListener() {
                @Override
                public void onFailure(@NonNull Exception e) {
                    Log.e("error",""+e.getMessage());
                }
            })
        .addOnCompleteListener(new OnCompleteListener<List<Detectedobject>>() {
            @SuppressLint("UnsafeExperimentalUsageError")
            @Override
            public void onComplete(@NonNull Task<List<Detectedobject>> task) {
                imageProxy.getimage().close();
                imageProxy.close(); 

            }
        });
    }
  }
}

绘制矩形：

public void drawFoundobj(Rect boundingBox) {
    foundYo = boundingBox;
    invalidate(); //call draw
}

和

Paint mPaint = new Paint();
Canvas mCanvas = new Canvas();
@Override
protected void onDraw(Canvas canvas) {

    //clear exsisting rects 
    Xfermode x = mPaint.getXfermode();
    mPaint.setXfermode(new PorterDuffXfermode(PorterDuff.Mode.CLEAR));
    mPaint.setXfermode(x);
    mCanvas.drawPaint(mPaint);


    //set current rect
    Paint paint = new Paint();
    paint.setColor(Color.YELLOW);
    paint.setstrokeWidth(3);

    paint.setAlpha(40);
    canvas.drawRect(foundYo,paint);
}

结果如下：

正如你所看到的，问题是虽然物体以某种方式被识别，但与物体的位置相比，坐标是错误的。

可能是什么原因？

我从 imageanalyzer 接收到的图像的分辨率是否较小，从而导致坐标较小从而错误？

还是我画错了？

解决方法

您需要获取分析图像的宽度和高度，并将边界框转换为合适的显示尺寸。请参考 mlkit vision_quickstart 中的这个示例作为参考。 https://github.com/googlesamples/mlkit/blob/74d5edb101d1e2fb8bd404c41a684b71a06d507a/android/vision-quickstart/app/src/main/java/com/google/mlkit/vision/demo/java/CameraXLivePreviewActivity.java#L421