Complexity

Research Article

Development of ANPR Framework for Pakistani Vehicle Number Plates Using Object Detection and OCR

Table 1

Summary of the comparative study of the existing methods.
Ref no.Plate’s typeProposed methodPerformance rate/accuracyDiscussion/remarks
[2]Real-time imagesNiblack threshold, blob-coloring, neural network -based OCR86.1% for recognitionAddressing low-resolution images with an average computation time is 1.5 seconds
[3]Nepali number platesGrayscale, morphological operation, median filter, phase correlation, cross-correlation in template matching67.98% for cross-correlation, 63.46% for phase correlationDue to template matching the average accuracy is low
[4] Indian number platesBasic preprocessing, PCA for feature extraction, CNN classifier for recognitionSuccessful execution is done by using Raspberry PiSuitable resources are discussed in it
[5]Real-time imagesYOLOv2, Warped Planner Object Detection Network (WPOD-NET) for detection, OCR for recognitionFor detection is 76.8% and for recognition is 75%Focus on unconstrained images having single-row number plates
[6]Real-time imagesGrayscale, binarization, masking for plate detection, distinguishing definite characters by SVM (deployed using MATLAB 2010a)92% accuracy for recognitionCannot recognize motion blurred, overlapped, skewed, and plate with a different language
[7]Indian number platesGrayscale, binarization, contrast extension, median filter, MATLAB region props function for segmentation, zonal function for feature extraction, template matching for recognitionThe recognition rate lies between 75% and 85%Addressing low resolution, unskew and clear images
[8]Qatar number platesRescaling, morphological operation, connected component analysis (CCA), vector crossing, zoning, template matchingRecognition rate is 99.5% with 0.63 ms computation timeHigh-resolution and single-row images are addressed
[9]Ghanaian number platesGrayscale, Gaussian kernel, Sobel edge detector, CCA on a binarized image, Tesseract OCR for character recognitionRecognition rate is 60% with 0.2 s computation timeUp to a distance of 5 meters, the detection algorithm performs fairly efficiently
[10]Real-time imagesDesaturation, segmentation, plate recognition using Raspberry PiRecognition rate is 85% with a 3-second delayThe system manages to deliver good results when the subject is within 2 meters from the camera
[11]Indian number platesYOLOv3 for detecting and recognition100% for detection, 91% for recognitionHigh-resolution and single-row number plates are focused
[12] Indonesian number platesThresholding, morphological operations, KNN for recognition98% accuracy for recognitionDamage and cut-off characters are unidentifiable
[13] Malaysian number platesSauvola threshold, template matching for character identification83.17% average accuracy for recognitionHigh-resolution images capture from the distance of 1.5 meters to 2 meters
[14]European number platesUndersampling, quantization, binary masking, Tesseract OCR for character segmentation and recognition90–100% average precision for recognitionHigh-resolution single-row number plates are focused
[15]Real-time imagesBinarization, minimum filter to enhance dark values, Roberts edge detection, bounding boxes, template matchingNo accuracy information because they preferred plate label managementSuitable resources are discussed in it
[16]Indian number platesMorphological operations, Gaussian filter, deskewing, KNN96.22% for recognitionFocused on the single-row number plate
[17]Indonesian number platesHistogram equalization, canny and Laplacian or Sobel and Laplacian edge detection, bounding box, cropping, OCR using eigenvectorNo accuracy information because they preferred hardware configurationSuitable resources are discussed in it
[20]Pakistani number platesHistogram equalization, distinct feature matching93% accuracy for recognitionMedium resolution with the single-row number plate
[21]Pakistani number platesGrayscale, Gaussian filter, canny edge detector, KNN93% accuracy for recognitionMainly focus on a car number plate