89 lines
3.1 KiB
Kotlin
89 lines
3.1 KiB
Kotlin
package com.guiagent.ocr
|
||
|
||
import android.graphics.BitmapFactory
|
||
import com.google.gson.Gson
|
||
import fi.iki.elonen.NanoHTTPD
|
||
import java.io.ByteArrayOutputStream
|
||
|
||
class OcrHttpServer(port: Int = 18900) : NanoHTTPD(port) {
|
||
|
||
private val gson = Gson()
|
||
private val defaultPath = "/sdcard/ocr_screen.png"
|
||
|
||
override fun serve(session: IHTTPSession): Response {
|
||
return when (session.uri) {
|
||
"/ocr" -> handleOcr(session)
|
||
"/snap" -> handleSnap(session)
|
||
"/health" -> jsonResponse(mapOf("status" to "ok", "engine" to "mlkit-chinese"))
|
||
else -> newFixedLengthResponse(Response.Status.NOT_FOUND, MIME_PLAINTEXT, "404")
|
||
}
|
||
}
|
||
|
||
/** 读文件方式 OCR */
|
||
private fun handleOcr(session: IHTTPSession): Response {
|
||
val params = session.parms ?: emptyMap()
|
||
val imagePath = params["path"] ?: defaultPath
|
||
return doOcr(params["text"]) { OcrEngine.recognize(imagePath) }
|
||
}
|
||
|
||
/** POST 图片数据直接 OCR，不存文件 */
|
||
private fun handleSnap(session: IHTTPSession): Response {
|
||
val params = session.parms ?: emptyMap()
|
||
|
||
if (session.method == Method.POST) {
|
||
// NanoHTTPD parseBody 将 binary data 存到临时文件
|
||
val bodyFiles = HashMap<String, String>()
|
||
session.parseBody(bodyFiles)
|
||
|
||
// postData 键对应临时文件路径
|
||
val tmpPath = bodyFiles["postData"]
|
||
if (tmpPath != null) {
|
||
val imageBytes = java.io.File(tmpPath).readBytes()
|
||
val bitmap = BitmapFactory.decodeByteArray(imageBytes, 0, imageBytes.size)
|
||
if (bitmap != null) {
|
||
return doOcr(params["text"]) { OcrEngine.recognizeBitmap(bitmap) }
|
||
}
|
||
return jsonResponse(mapOf("error" to "decode failed", "size" to imageBytes.size, "count" to 0))
|
||
}
|
||
return jsonResponse(mapOf("error" to "no body received", "count" to 0))
|
||
}
|
||
|
||
// GET: 读文件方式 fallback
|
||
return handleOcr(session)
|
||
}
|
||
|
||
private fun doOcr(query: String?, recognize: () -> List<TextBox>): Response {
|
||
val startTime = System.currentTimeMillis()
|
||
var results = recognize()
|
||
|
||
if (!query.isNullOrBlank()) {
|
||
results = results.filter { it.text.contains(query) }
|
||
}
|
||
|
||
val elapsed = System.currentTimeMillis() - startTime
|
||
|
||
val response = mapOf(
|
||
"results" to results.map { box ->
|
||
mapOf(
|
||
"text" to box.text,
|
||
"x" to box.x,
|
||
"y" to box.y,
|
||
"w" to box.w,
|
||
"h" to box.h,
|
||
"cx" to box.cx,
|
||
"cy" to box.cy,
|
||
"confidence" to box.confidence
|
||
)
|
||
},
|
||
"count" to results.size,
|
||
"elapsed_ms" to elapsed
|
||
)
|
||
return jsonResponse(response)
|
||
}
|
||
|
||
private fun jsonResponse(data: Any): Response {
|
||
val json = gson.toJson(data)
|
||
return newFixedLengthResponse(Response.Status.OK, "application/json", json)
|
||
}
|
||
}
|