快速的多列文字识别

如何解决快速的多列文字识别

在多于一列的图像上使用文本识别时遇到一些麻烦。例如：

cars:
1 red car       1 blue car
2 red car       2 blue car
3 red car       3 blue car

将返回：

cars:
1 red car 
1 blue car
2 red car
2 blue car
3 red car
3 blue car

我需要它退货的是

cars:
1 red car
2 red car
3 red car
1 blue car
2 blue car
3 blue car

并非我需要分析的每张图像都有2列，但有很多列。

我当前使用的代码如下：

let requestHandler = VNImageRequestHandler(ciImage: image)
let textRecognitionRequest = VNRecognizeTextRequest(completionHandler: recognizeTextHandler)
 
textRecognitionRequest.recognitionLevel = VNRequestTextRecognitionLevel.accurate

do {
     try requestHandler.perform([textRecognitionRequest])
}
catch {
      print(error)
}

func recognizeTextHandler(request: VNRequest,error: Error?) {
          
      let text = ""
      guard let results =
                request.results as? [VNRecognizedTextObservation] else {
            fatalError("Could Not Detect Text")
        }
        
        for visionResult in results {
            let maximumCandidates = 1
            guard let candidate = visionResult.topCandidates(maximumCandidates).first else {
                continue
            }
            let string = candidate.string
            text += "\n"
            text += candidate.string
        }
        print(text)
}

任何帮助将不胜感激。预先感谢！

解决方法

几个月前，我遇到了一个类似的模块，其中我必须从图像中提取文本。我在GitHub上找到了一个项目，这确实很有帮助。不幸的是，我没有该GitHub项目的链接，但我要在这里从我的项目中共享代码。

**我不算学分。该Github项目的所有学分。找到链接后，我将共享该链接。**

import UIKit
import Reachability

创建网络可达性实例。

let reachability = try! Reachability()

从图库中拾取图像并提取文本。

extension ExtractTextViewController: UIImagePickerControllerDelegate,UINavigationControllerDelegate {
    func imagePickerControllerDidCancel(_ picker: UIImagePickerController) {
        picker.dismiss(animated: true)
    }
    
    func imagePickerController(_ picker: UIImagePickerController,didFinishPickingMediaWithInfo info: [UIImagePickerController.InfoKey: Any]) {
        
        picker.dismiss(animated: true)
        if let pickedImage = info[.originalImage] as? UIImage {
            
            // check internet connection first..
            if reachability.connection == .wifi || reachability.connection == .cellular {
                
                // using google api to detect text from images..
                // it will extract text from any image,be it in english,spanish,german etc..
                GoogleAPI.taskDetect(form: pickedImage) { (result,errorString) in
                    guard let result = result else {
                        self.showAlert(title: "Erorr",message: errorString!)
                        return
                    }
                    // Text has been extracted.. It's as simple as that..
                    let extractedText = result.fullTextAnnotation.text.replacingOccurrences(of: "\n",with: " ")
                    print(extractedText)
                }
            }
            else {
                // some alert when internet is not connected..
                self.showAlert(title: "Warning",message: "Please check your internet connection. It is required to extract text from images.")
                
            }
        }
        picker.dismiss(animated: true,completion: nil)
    }
}

GoogleAPI.swift

import Foundation
import UIKit
import Alamofire


class GoogleAPI {
    
    
    enum Endpoint {
        static let endPoint = "https://vision.googleapis.com/v1/images:annotate"
        // This api key was mentioned in that GitHub project. I'm using the same and it's working perfectly fine.
        static let apiKey = "AIzaSyD-IAmWA0NKGNarU0Bxtz-esSp47V5phBo"
        
        case detectImage
        
        var stringValue: String {
            switch self {
            case .detectImage: return "\(Endpoint.endPoint)?key=\(Endpoint.apiKey)"
            }
        }
        
        var url: URL {
            return URL(string: stringValue)!
        }
    }
    
    class func base64EncodeImage(_ image: UIImage) -> String? {
        return image.pngData()?.base64EncodedString(options: .endLineWithCarriageReturn)
    }
    
    class func taskDetect(form image: UIImage,completion: @escaping (Results?,String?) -> Void) {
        guard let base64Image = base64EncodeImage(image) else {
            completion(nil,"Error while base64 encoding image")
            return
        }
        let parameters: Parameters = [
            "requests": [
                [
                    "image": [
                        "content": base64Image
                    ],"features": [
                        [
                            "type": "TEXT_DETECTION"
                        ]
                    ]
                ]
            ]
        ]
        let headers: HTTPHeaders = [
            "X-Ios-Bundle-Identifier": Bundle.main.bundleIdentifier ?? "",]
        AF.request(
            Endpoint.detectImage.url,method: .post,parameters: parameters,encoding: JSONEncoding.default,headers: headers)
            .responseData { response in
                
                guard let data = response.data else {
                    completion(nil,response.error?.localizedDescription)
                    return
                }
                do {
                    let object = try JSONDecoder().decode(GoogleResponse.self,from: data)
                    completion(object.responses[0],nil)
                } catch {
                    print(error.localizedDescription)
                    completion(nil,error.localizedDescription)
                }
        }
    }
}

GoogleResponse.swift

import Foundation
import UIKit

struct Vertex: Codable {
    let x: Int?
    let y: Int?
}

struct BoundingBox: Codable {
    let vertices: [Vertex]
}

struct Annotation: Codable {
    let description: String
    let boundingPoly: BoundingBox
}

struct Full: Codable {
    let text: String
}

struct Results: Codable {
    let textAnnotations: [Annotation]
    let fullTextAnnotation: Full
}

struct GoogleResponse: Codable {
    let responses: [Results]
}