Fruit sizing systems are an emerging technology in the agricultural industry, specifically orchard management. Fruit sizing systems employ various computer vision approaches such as those based around thermal cameras, RBG-D cameras, and LiDAR. Most of the current systems for fruit sizing occur post-harvest and the current On-Tree or in-field fruit sizing is commonly done manually which costs valuable time, labour and can be costly in cases where orchard owners outsource the work to alternative expert services. Furthermore, the reliance on said services can cause issues to arise such as miscommunication, missing appointments for sampling periods, or general lack of control. For this reason, this thesis proposes a novel proof-of-concept application that involves the employment of a LiDAR-based IOS device for On-Tree fruit sizing. This thesis documents the development of said approach and contains data derived from simple experiments to gain answers as to whether this proof-of-concept application is a viable method for On-Tree fruit sizing. The results show a positive precedent as the On-Tree apple experiment as well as the cherry bunch experiments demonstrated error range percentages within 6% between 0.5m and 2m away from the apple. Furthermore, the post-modification apple experiment was conducted on the same apple data-set produced error range percentages within 0.5% between 0.5m and 2m. The post-modification experiment purpose highlights the requirement for a dynamic approach to the exponential regression best-fit parameters applied. An alternative experiment was done to test the long-range capabilities of the proof-of-concept as well as determine the reliance on accurate pixel width measurements. The results of the said experiment are promising, showing an error range percentage within 4% between 0.5m and 5m. As the machine learning model is derivative based on [1] provided by Hectre, the results for the detection count experiment were lackluster demonstrating odd values such as 29/52 cherries detected. The limitations of the project are the heavy reliance on an accurate calibration of the parameters as well as an accurate pixel width of fruit (aka bounding box measurements from the machine learning model). Overall, the benefits of the proof-of-concept include lightweight accessibility, high sample rate, cost-efficient as well as competitive with alternative fruit sizing systems and far exceeds that of manual intervention-based On-Tree fruit sizing.