Fruit Ripeness Detector: Near Infrared Sensor for Harvest

Judging fruit ripeness accurately is critical for both maximum yield and premium pricing at Indian mandis. A fruit ripeness detector using near infrared sensors provides non-destructive, objective measurement of internal sugar content, moisture, and maturity – eliminating guesswork and reducing post-harvest losses that cost Indian farmers an estimated Rs 90,000 crore annually. This guide explains the science and shows how to build an affordable NIR-based ripeness detection system using Indian hobbyist components.

Table of Contents

• Science of NIR Fruit Detection
• Hardware Setup
• Available NIR Sensor Modules
• Measurement and Analysis Code
• Crop-Specific Guide for India
• Post-Harvest Integration
• Frequently Asked Questions

Science of NIR Fruit Detection

Near infrared spectroscopy (700-2500nm wavelength range) is the industry standard for non-destructive fruit quality assessment. Key chemical bonds in fruit absorb NIR light at specific wavelengths:

• Sugar (Brix content): Glucose and fructose absorb at 1450nm and 2100nm
• Water content: Strong absorption at 970nm and 1450nm
• Starch to sugar conversion: Shift in absorbance at 850-1000nm as ripening progresses
• Lycopene (tomato ripeness): Absorbance shift at 475nm (visible) and 700nm

DIY NIR detectors using broadband NIR LEDs (850nm or 940nm) and photodiodes can detect relative changes in these parameters with surprising accuracy – sufficient for ripeness classification (unripe, nearly ripe, ripe, overripe) even if absolute Brix values require calibrated laboratory instruments.

Hardware Setup

A basic DIY NIR ripeness detector needs:

• NIR LED (850nm or 940nm, 5mm) – illumination source (Rs 5-15 each)
• NIR photodiode (SFH213 or BPW34) – reflected light detector (Rs 30-80)
• Transimpedance amplifier (TIA) with op-amp MCP6001 – converts photodiode current to measurable voltage
• Arduino Uno or Nano – ADC reading of amplified signal
• Dark measurement chamber – 3D-printed box to eliminate ambient light interference
• Optional: AS7265x spectral sensor (Adafruit) – 18-channel spectral sensing from 410-940nm for more accurate classification

For a more advanced system: the AS7265x trimodal spectral sensor covers 410-940nm in 18 channels and interfaces via I2C to Arduino or Raspberry Pi. It costs approximately Rs 2,500-4,000 on Amazon India and provides chemical-grade spectral data without any custom optics.

Available NIR Sensor Modules

Measurement and Analysis Code

// Single-band NIR reflectance ripeness detector
// Using BPW34 photodiode + 850nm LED + TIA amplifier

#define NIR_LED_PIN 9       // PWM pin for LED control
#define SENSOR_PIN A0       // Amplified photodiode signal

// Calibration values (measure white reference card)
float white_ref = 900.0;   // Adjust after calibration
float dark_ref = 50.0;     // With LED off

float measureReflectance() {
  // Dark reading
  digitalWrite(NIR_LED_PIN, LOW);
  delay(100);
  float dark = analogRead(SENSOR_PIN);
  
  // Illuminated reading
  analogWrite(NIR_LED_PIN, 200); // 78% duty cycle
  delay(100);
  float illuminated = analogRead(SENSOR_PIN);
  
  // Relative reflectance %
  float refl = ((illuminated - dark) / (white_ref - dark_ref)) * 100.0;
  return constrain(refl, 0, 100);
}

void classifyRipeness(float refl) {
  // Classification thresholds for mango (Alphonso variety)
  // Calibrate for each variety and crop
  if(refl < 25) Serial.println("UNRIPE - harvest in 5-7 days");
  else if(refl < 45) Serial.println("NEARLY RIPE - harvest in 2-3 days");
  else if(refl < 65) Serial.println("RIPE - harvest now");
  else Serial.println("OVERRIPE - immediate harvest or processing");
}

void setup() { 
  Serial.begin(9600); 
  pinMode(NIR_LED_PIN, OUTPUT);
}

void loop() {
  float r = measureReflectance();
  Serial.print("NIR Reflectance: "); Serial.print(r,1); Serial.println("%");
  classifyRipeness(r);
  delay(2000);
}

Crop-Specific Guide for India

Calibration parameters for major Indian fruit crops:

• Mango (Alphonso, Dasheri, Kesar): NIR reflectance at 850nm decreases 15-20% from unripe to ripe as starch converts to sugar. Skin colour alone is misleading – especially for Alphonso which stays green when ripe. NIR consistently outperforms visual assessment.
• Banana (Cavendish, Robusta): 940nm absorbance correlates strongly with Brix increase from 14 (green) to 23+ (fully ripe). Government-operated ripening chambers in Maharashtra use NIR inline for quality grading.
• Tomato: 700nm reflectance increases dramatically as lycopene accumulates during ripening. The colour change is visible but NIR allows detection of internal ripeness in artificially coloured tomatoes.
• Guava: 850nm reflectance peaks at optimal harvest maturity then declines as over-softening begins – useful for table variety grading.

Post-Harvest Integration

Integrate the ripeness detector into sorting lines:

• Mount a dark measurement chamber on a conveyor belt with a NIR sensor above each fruit
• Arduino or Raspberry Pi classifies each fruit as it passes
• Servo-driven sorting gates direct fruit to graded bins (Grade A, B, reject)
• Processing rate: 1-2 fruits per second with a single-point sensor
• Cost of automated sorting line: Rs 15,000-30,000 depending on belt length and servo count

Frequently Asked Questions

How accurate is a DIY NIR detector compared to a commercial refractometer?

A calibrated DIY single-band NIR system typically classifies fruit into 4 ripeness categories with 80-90% accuracy for a single variety grown under consistent conditions. A commercial NIR spectrophotometer achieves 95-98% accuracy with absolute Brix values to 0.1 precision. The DIY system is sufficient for farm-level sorting to avoid harvesting too early or letting fruit overripen on the tree.

Can the same sensor work for different fruit varieties?

The hardware works for any fruit, but calibration is variety-specific. A sensor calibrated for Alphonso mango will not directly apply to Dasheri mango. You need to collect training samples (measure 50+ fruits across the ripeness spectrum along with destructive Brix measurements) and build a new calibration curve for each variety. The AS7265x multi-channel sensor transfers calibration better across varieties due to its richer spectral data.

Will the sensor work through the fruit skin?

NIR at 850-940nm penetrates 3-8mm into fruit tissue depending on skin thickness and water content. For thin-skinned fruit (tomato, banana), this accesses the sub-epidermal flesh where ripening chemistry occurs. For thick-skinned fruit (mango, citrus), use a contact probe that presses against the natural skin indentation or the shoulder where skin is thinner.

Is this technology already used in Indian agriculture?

Yes – Mahagrape and Maharashtra State Horticulture Board have trialled NIR-based mango grading. Several private exporters (especially for Alphonso to Middle East and Europe) use imported NIR sorters costing Rs 15-40 lakh. DIY versions make the core technology accessible to small and medium farmers who cannot afford commercial systems.

Can the system detect pesticide residue on fruit surface?

Simple NIR reflectance cannot reliably detect specific pesticide residues – that requires hyperspectral imaging (400-2500nm, 100+ bands) or chemical analysis. Surface contamination affects overall NIR reflectance, but distinguishing pesticide from normal wax bloom or dust requires more sophisticated spectral analysis than a DIY Arduino system can provide.

Shop Agriculture & Smart Farming at Zbotic