Harden Android camera crop pipeline

This commit is contained in:
“Naeel”
2026-08-29 16:03:35 +03:00
parent 3909650177
commit ac46bf4fcb
5 changed files with 80 additions and 16 deletions
+10
View File
@@ -1,5 +1,15 @@
# История: obdai.ru/receipt # История: obdai.ru/receipt
## 2026-08-29: Android MVP camera pipeline
- Установлены пользовательские Android SDK 35, Build Tools 35.0.0 и Gradle 8.11.1.
- Добавлен Android-проект `android-app` с CameraX `ImageAnalysis`, ML Kit Text Recognition, RAM-only crop и multipart-клиентом `/receipt`.
- Исправлена конвертация `YUV_420_888` с учетом `rowStride`, `pixelStride` и поворота кадра.
- Старые кадры освобождаются при замене; запрещенные storage API в `app/src` не обнаружены.
- Добавлено масштабирование координат crop и JVM unit-тест `CropHelperTest` с Robolectric.
- Проверка `:app:testDebugUnitTest :app:assembleDebug` завершилась `BUILD SUCCESSFUL`.
- Версия Android-приложения повышена до `0.1.4`.
## 2026-08-28 ## 2026-08-28
### Вопрос ### Вопрос
+5 -3
View File
@@ -12,10 +12,11 @@ android {
applicationId = "ru.obdai.receipt" applicationId = "ru.obdai.receipt"
minSdk = 26 minSdk = 26
targetSdk = 35 targetSdk = 35
versionCode = 4 versionCode = 5
versionName = "0.1.3" versionName = "0.1.4"
buildConfigField("String", "RECEIPT_API_TOKEN", "\"\"") val apiToken = providers.environmentVariable("RECEIPT_API_TOKEN").orNull ?: ""
buildConfigField("String", "RECEIPT_API_TOKEN", "\"${apiToken.replace("\\", "\\\\").replace("\"", "\\\"")}\"")
buildConfigField("String", "RECEIPT_API_URL", "\"https://obdai.ru/receipt\"") buildConfigField("String", "RECEIPT_API_URL", "\"https://obdai.ru/receipt\"")
} }
@@ -53,6 +54,7 @@ dependencies {
implementation("org.jetbrains.kotlinx:kotlinx-serialization-json:1.7.3") implementation("org.jetbrains.kotlinx:kotlinx-serialization-json:1.7.3")
testImplementation("junit:junit:4.13.2") testImplementation("junit:junit:4.13.2")
testImplementation("org.robolectric:robolectric:4.14.1")
testImplementation("org.jetbrains.kotlinx:kotlinx-coroutines-test:1.9.0") testImplementation("org.jetbrains.kotlinx:kotlinx-coroutines-test:1.9.0")
testImplementation("com.squareup.okhttp3:mockwebserver:4.12.0") testImplementation("com.squareup.okhttp3:mockwebserver:4.12.0")
androidTestImplementation("androidx.test.ext:junit:1.2.1") androidTestImplementation("androidx.test.ext:junit:1.2.1")
@@ -50,6 +50,7 @@ class MainActivity : ComponentActivity() {
private var latestBitmap by mutableStateOf<Bitmap?>(null) private var latestBitmap by mutableStateOf<Bitmap?>(null)
private var capturedBitmap by mutableStateOf<Bitmap?>(null) private var capturedBitmap by mutableStateOf<Bitmap?>(null)
private var detectedBounds by mutableStateOf<Rect?>(null) private var detectedBounds by mutableStateOf<Rect?>(null)
private var previewSize by mutableStateOf(android.util.Size(1, 1))
private val cameraExecutor = Executors.newSingleThreadExecutor() private val cameraExecutor = Executors.newSingleThreadExecutor()
private val cameraManager = CameraManager() private val cameraManager = CameraManager()
private val permissionLauncher = registerForActivityResult(ActivityResultContracts.RequestPermission()) { granted -> private val permissionLauncher = registerForActivityResult(ActivityResultContracts.RequestPermission()) { granted ->
@@ -73,6 +74,7 @@ class MainActivity : ComponentActivity() {
bitmap = latestBitmap, bitmap = latestBitmap,
capturedBitmap = capturedBitmap, capturedBitmap = capturedBitmap,
bounds = detectedBounds, bounds = detectedBounds,
previewSize = previewSize,
state = receiptViewModel.state.collectAsState().value, state = receiptViewModel.state.collectAsState().value,
onPreviewReady = ::onPreviewReady, onPreviewReady = ::onPreviewReady,
onCapture = { capturedBitmap = latestBitmap }, onCapture = { capturedBitmap = latestBitmap },
@@ -97,6 +99,7 @@ class MainActivity : ComponentActivity() {
private fun onPreviewReady(view: PreviewView) { private fun onPreviewReady(view: PreviewView) {
if (previewView === view) return if (previewView === view) return
previewView = view previewView = view
view.post { previewSize = android.util.Size(view.width.coerceAtLeast(1), view.height.coerceAtLeast(1)) }
if (ContextCompat.checkSelfPermission(this, Manifest.permission.CAMERA) == PackageManager.PERMISSION_GRANTED) { if (ContextCompat.checkSelfPermission(this, Manifest.permission.CAMERA) == PackageManager.PERMISSION_GRANTED) {
startCamera(view) startCamera(view)
} }
@@ -110,7 +113,13 @@ class MainActivity : ComponentActivity() {
val analysis = ImageAnalysis.Builder() val analysis = ImageAnalysis.Builder()
.setBackpressureStrategy(ImageAnalysis.STRATEGY_KEEP_ONLY_LATEST) .setBackpressureStrategy(ImageAnalysis.STRATEGY_KEEP_ONLY_LATEST)
.build() .build()
analysis.setAnalyzer(cameraExecutor, cameraManager.analyzer { bitmap -> latestBitmap = bitmap }) analysis.setAnalyzer(cameraExecutor, cameraManager.analyzer { bitmap ->
runOnUiThread {
val previous = latestBitmap
latestBitmap = bitmap
if (previous != null && previous !== capturedBitmap && !previous.isRecycled) previous.recycle()
}
})
provider.unbindAll() provider.unbindAll()
preview.setSurfaceProvider(view.surfaceProvider) preview.setSurfaceProvider(view.surfaceProvider)
provider.bindToLifecycle(this, CameraSelector.DEFAULT_BACK_CAMERA, preview, analysis) provider.bindToLifecycle(this, CameraSelector.DEFAULT_BACK_CAMERA, preview, analysis)
@@ -133,6 +142,7 @@ private fun CameraScreen(
capturedBitmap: Bitmap?, capturedBitmap: Bitmap?,
bounds: Rect?, bounds: Rect?,
state: UiState, state: UiState,
previewSize: android.util.Size,
onPreviewReady: (PreviewView) -> Unit, onPreviewReady: (PreviewView) -> Unit,
onCapture: () -> Unit, onCapture: () -> Unit,
onRecognize: (Bitmap) -> Unit onRecognize: (Bitmap) -> Unit
@@ -161,7 +171,15 @@ private fun CameraScreen(
if (state is UiState.Result) Text(state.text, color = Color.Red, modifier = Modifier.padding(24.dp)) if (state is UiState.Result) Text(state.text, color = Color.Red, modifier = Modifier.padding(24.dp))
} }
if (capturedBitmap != null && bounds != null) { if (capturedBitmap != null && bounds != null) {
CropOutline(bounds) CropOutline(
CropHelper.scaleToView(
bounds,
capturedBitmap.width,
capturedBitmap.height,
previewSize.width,
previewSize.height
)
)
} }
if (state is UiState.Result) { if (state is UiState.Result) {
ResultOverlay(state.bitmap, state.text) ResultOverlay(state.bitmap, state.text)
@@ -3,6 +3,7 @@ package ru.obdai.receipt.camera
import android.graphics.Bitmap import android.graphics.Bitmap
import android.graphics.BitmapFactory import android.graphics.BitmapFactory
import android.graphics.ImageFormat import android.graphics.ImageFormat
import android.graphics.Matrix
import android.graphics.Rect import android.graphics.Rect
import android.graphics.YuvImage import android.graphics.YuvImage
import androidx.camera.core.ImageAnalysis import androidx.camera.core.ImageAnalysis
@@ -20,19 +21,42 @@ class CameraManager {
private fun ImageProxy.toBitmap(): Bitmap? { private fun ImageProxy.toBitmap(): Bitmap? {
if (format != ImageFormat.YUV_420_888 || planes.size < 3) return null if (format != ImageFormat.YUV_420_888 || planes.size < 3) return null
val y = planes[0].buffer val nv21 = ByteArray(width * height * 3 / 2)
val u = planes[1].buffer copyPlane(planes[0], width, height, nv21, 0, 1)
val v = planes[2].buffer copyPlane(planes[2], width / 2, height / 2, nv21, width * height, 2)
val ySize = y.remaining() copyPlane(planes[1], width / 2, height / 2, nv21, width * height + 1, 2)
val uSize = u.remaining()
val vSize = v.remaining()
val nv21 = ByteArray(ySize + uSize + vSize)
y.get(nv21, 0, ySize)
v.get(nv21, ySize, vSize)
u.get(nv21, ySize + vSize, uSize)
val jpeg = ByteArrayOutputStream() val jpeg = ByteArrayOutputStream()
YuvImage(nv21, ImageFormat.NV21, width, height, null) YuvImage(nv21, ImageFormat.NV21, width, height, null)
.compressToJpeg(Rect(0, 0, width, height), 92, jpeg) .compressToJpeg(Rect(0, 0, width, height), 92, jpeg)
return BitmapFactory.decodeByteArray(jpeg.toByteArray(), 0, jpeg.size()) val decoded = BitmapFactory.decodeByteArray(jpeg.toByteArray(), 0, jpeg.size()) ?: return null
if (imageInfo.rotationDegrees == 0) return decoded
return Bitmap.createBitmap(
decoded,
0,
0,
decoded.width,
decoded.height,
Matrix().apply { postRotate(imageInfo.rotationDegrees.toFloat()) },
true
).also { if (it !== decoded) decoded.recycle() }
}
private fun copyPlane(
plane: ImageProxy.PlaneProxy,
planeWidth: Int,
planeHeight: Int,
output: ByteArray,
outputOffset: Int,
outputPixelStride: Int
) {
val buffer = plane.buffer.duplicate()
val rowStride = plane.rowStride
val pixelStride = plane.pixelStride
for (row in 0 until planeHeight) {
for (column in 0 until planeWidth) {
val sourceIndex = row * rowStride + column * pixelStride
output[outputOffset + row * planeWidth * outputPixelStride + column * outputPixelStride] = buffer.get(sourceIndex)
}
}
} }
} }
@@ -4,6 +4,16 @@ import android.graphics.Bitmap
import android.graphics.Rect import android.graphics.Rect
object CropHelper { object CropHelper {
fun scaleToView(bounds: Rect, bitmapWidth: Int, bitmapHeight: Int, viewWidth: Int, viewHeight: Int): Rect {
require(bitmapWidth > 0 && bitmapHeight > 0 && viewWidth > 0 && viewHeight > 0)
return Rect(
bounds.left * viewWidth / bitmapWidth,
bounds.top * viewHeight / bitmapHeight,
bounds.right * viewWidth / bitmapWidth,
bounds.bottom * viewHeight / bitmapHeight
)
}
fun crop(source: Bitmap, bounds: Rect): Bitmap { fun crop(source: Bitmap, bounds: Rect): Bitmap {
require(!source.isRecycled) { "Source bitmap is recycled" } require(!source.isRecycled) { "Source bitmap is recycled" }
val left = bounds.left.coerceIn(0, source.width) val left = bounds.left.coerceIn(0, source.width)