The European "Green Deal" requires the increase of yield and product purity in waste treatment plants and the implementation of new technologies to increase the recycling rate of plastic packaging from the current 25% to 55% by 2030. This paper deals with the use of sensor-based material flow monitoring (SBMM), also known as "material flow monitoring", in plastic sorting plants. For this purpose, a survey of the state of the art of plastics sorting in Belgium and Austria was first carried out and a comparison of the two countries was made to determine the current technological standard. The aim is to investigate whether the number of pixels of a plastic type determined by using near infrared (NIR) sensors can be used to predict its mass fraction in the material stream. For this purpose, several tests with NIR sensors were carried out in a P+MD (plastic bottels, metal, drink cartons) sorting plant in Belgium. A check of the correlation of the pixel data measured by using the NIR sensor with the corresponding masses of selected plastics was carried out based on the results of two series of tests that were performed. In test series 1, two material streams of lightweight packaging were fed onto conveyor belts and examined for their material composition using a near-infrared sensor (EVK Kerschhaggl GmbH, wavelength range 0.9 ─ 1.7 µm). In addition, the impurities were sorted manually with a scale that generated time-resolved data. The area fraction of the impurities determined by SBMM was compared with the mass fraction determined by hand sorting and checked for their correlation. In test series 2, the teach-ins regarding to yield and purity of two NIR sorters (SBS), which were positioned in front of the sensors in the sorting chain, were changed to investigate the occurring effects in the pixel data of the sensors. The results of the first set of experiments indicated that the sensors performed better at low belt loading and low amount of transparency 2D material (case 1) than at high belt loading with a lot of objects overlapping and a high film content (case 2). The interfering weight and interfering pixel data in the PS line (case 1) show a correlation of 0.81 while the same data sets in the PP line (case 2) show a correlation 0.52. This indicates a higher resilience of the SBMM data in the PS line. In the results of the second set of experiments, differences in the SBMM data after modified SBS teach-in were observed. The results also indicated that the SBS sensor data could also be used for a possible quantitative analysis of the material stream composition and product quantity. However, further experiments would need to be conducted to validate this claim.