Accuracy of Photo-Optical Timber Measurement Using Stereo Camera Technology

Harvested timber requires efficient and accurate measurements for timber trade. Recently, the amount of timber that is harvested for industrial purposes is growing and methods for round - wood measurements are under constant development. Some of these solutions are with certifi - cation confirming its accuracy, some of them are proposed for general use without certification. The aim of this paper was to select the best and most useful electronic solution for timber volume masurement and calculation, and to recommend the most effective and accurate solution for future timber trade. Three photo-optical systems were tested: LogStackPro, iFovea and Timbeter. Each system was used to measure 71 stacks of pine and oak roundwood, which amounted to 3481.15 cubic meter stacked. Timber volumes obtained from the manual measurements were used as reference. Volumes obtained from the photo-optical systems were larger in comparison with the volume from manual measurements, by 3.37, 8.07 and 9.08%, in LogStackPro, iFovea and Timbeter, respectively. It was concluded from the tested systems that, currently, the most recommended solution for timber measurement will be LogStack Pro, which also presented, in most cases, the smallest deviations from the volume obtained in manual measurements.


Introduction
Harvested timber requires measurement for trade purposes.According to FAO estimates, global round wood production, including wood fuel (WF) and in dustrial roundwood (IR), totalled 3966 million cubic metres (WF 1945 million m 3 , IR 2021 million m 3 ) in 2020 (FAO 2020).Compared to 2000, global timber production has therefore increased by approximately 24% (FAO 2000).In Poland, in 2020, roundwood tim ber and logs accounted for more than 60% of harvest ed wood, mainly prepared using cut-to-length tech nology (Mederski et al. 2022).Applications enabling the use of photo-optical methods in forestry are com ing onto the market, and these can facilitate and speed up the measurement of stacked wood.
As a result of the processing, the application iden tifies the front of the stack and, depending on the se lected measurement method, determines the diameter of the individual logs and their number, or the area of the whole stack based on its outline (polygon) and the reference marked on the stack.Both, the diameters of the individual logs and the stack shape (polygon) can be adjusted directly during stack processing or after the stack has been saved.Importantly, measurements can be performed without internet access.In case of software manufacturers who use external devices, the measurments are made by recording the video of the timber stack.Moreover, in case of Dralle and LogStackPro systems, the reference is defined as dis tance between centeres of two cameras located on the device.The only photo-optical stereoscopic (hereafter simply called stereoscopic) handheld device without a reference is LogStackPro, created by HD Silva.The sScale system, also based on a no-reference method, was created by Dralle, but in this case the device needs to operate on the roof of a car, and thus requires Croat.j. for.eng.45(2024)1 locations with good road systems alongside the timber stacks.Moreover, the Dralle sScale system is recom mended for large stacks (Mederski et al. 2018).The LogStackPro and Dralle sScale photo-optical measure ment systems both have certified accuracy.
The solutions proposed by manufacturers are quite universal; however, all ultimately require adaptation to the rules for measurement and conversion methods applied in each country.These methods are already used in countries such as Germany, Finland, Estonia, Japan and Brazil (HD LogSystem, SDP Digitale Produkte GmbH, Dralle, Timbeter, Trestima).In Poland, the use of photo-optical methods for wood measure ment is recommended in case of natural disasters (State Forests -document no.51/2019), and the Dralle measuring system was used in 2018 (Mederski et al. 2018) after a derecho windstorm that damaged ca. 8 million m 3 of timber (Sanginés de Cárcer et al. 2021, Schweier et al. 2024).Currently, smartphone apps cannot be certified due to the lack of control over updates to the operating system of the device on which the photo-optical software is installed.
Several studies have been carried out on photooptical measurement methods using smartphones (Boberg and Lilja 2016, Berendt et al. 2021, Cremer et al. 2021, Borz et al. 2022, Borz and Proto 2022), har vester built-in devices (Hohmann et al. 2017, Mederski et al. 2018), photogrammetry, 2D and 3D image pro cessing (Janak 2007, Herbon et al. 2015, Acuna and Sosa 2019, Nuhlíček et al. 2020), and other appliances (Gutzeit et al. 2011).Photo-optical systems have been the subject of recent scientific research.Studies by Cremer et al. (2021) and Berendt et al. (2021) were focused on iFovea Pro and manual hardwood mea surements.Borz et al. (2022) studied the potential of LiDAR for use in estimating log biometrics and com pared the results with traditional methods of measur ing logs using tape and a caliper in the sawmill envi ronment.Kärhä et al. (2019) tested the Trestima Stack app in two different conditions -in large terminal yards and at smaller roadside landings -and reported the higher accuracy of photo-optical measurements in terminal yards, while manual measurement accuracy was similar in both locations.Available published studies on the photo-optical measurement of harvested wood using smartphones indicate the potential of this method as an alternative to traditional timber measure ment.All smartphone applications based on photooptical systems require a length reference, however, and no data are available concerning the use of a stereo camera technology that can be used for measurement without a reference.The stereo camera concept was used in studies carried out at storage sites and depots (Knyaz and Maksimov 2014), where the researchers applied a photogrammetric technique for measuring stack volumes at a timber depot, using a notebook computer and two synchronized digital Canon EOS 1100D cameras installed on an 800 mm baseline.
Electronic systems for timber measurement are par ticularly useful in large-scale operations.In Poland, the conditions of the State Forests are suitable for the use of photo-optical measurements: the enterprise manages forests totalling over 7 million ha in area, with an an nual harvest amounting to 40 million m 3 of timber, of which over 90% is industrial timber (Statistical Year book of Forestry in Poland 2021).In the last decade, the State Forests have tested several measurement systems and applications, including Timbeter, iFovea Pro and sScale.These tests were carried out in order to recom mend the best solution for use by the organization.
Manual wood measurement is reported to be one of the most time-consuming jobs of a forester during a working day (Grzywiński et al. 2019).According to Jodłowski et al. (2016), manual measurement of raw wood takes more than 0.19 min/m 3 , while photo-optical applications take about 0.10 min/m 3 .Manual measure ment takes from 5 to 8 minutes longer than photooptical methods.Similar results for time consumption in timber measurements were reported by Borz and Proto (2022).According to Kärhä et al. (2019), measure ment with a mobile photo-optical app was faster by 3.1 s/m 3 than a conventional stacked wood measure ment method.
Studies on the accuracy of remote timber measure ment systems usually compare the results obtained from a device/software with timber volumes mea sured manually (VMM).A VMM is only an approxi mation to the real volume (RV).The RV can be ob tained by submerging the timber in water; however, this is a very time-consuming and expensive method.In the present research, it was assumed that values close to the real timber volume could be obtained for a large number of timber stacks, which were measured in practice for purposes of trade in the forestry sector.The hypothesis was that manual measurements are subject to errors, but some of them will lead to overes timates and others to underestimates so that eventu ally a large number of stacks will give results statisti cally close to the real volume.The aim of this study was to select the best and most useful electronic solu tion for timber volume measurement and calculation, and to recommend the most effective and accurate solution for future use in the timber trade.Therefore, different systems were compared: one based on stereo camera technology without a length reference, and two smartphone applications requiring a length reference.All selected methods were compared with the results obtained from manual measurements.age width, and height were measured according to the following guidelines:

Log length
The length of a stack was considered to be the nominal length of the logs stacked in it.The length was determined to the nearest 1 cm.

Average width
The width of the stack was measured parallel to the lower edge of the stack face, at its widest point, to the nearest 1 cm.

Average height
The height of the stack was measured perpendicu larly from the bottom edge of the stack to the top edge, to the nearest 1 cm.The height of the stack was deter mined as the arithmetic mean of at least four measure ments of the front side of the stack.The measuring points were evenly spaced along the width of the stack and permanently marked with paint.The distances between measurements (n) were not greater than 1 m (for stacks up to 10 m wide) and 2 m (for stacks wider than 10 m).The first height measurement was made at the left end of the stack, in a place where there were at least two logs on top of each other.The last height measurement was made in a place where there were at least two logs on top of each other (Fig. 2).

Photo-Optical Measurements
To test photo-optical measurement methods, two smartphone apps and one stereo camera technology were used.A Samsung Galaxy Tab Active 3 tablet was used for wood measurements performed with the two Android applications (iFovea and Timbeter).For measurements with LogStackPro, a Mobile Demand tablet was used.This is a device with a Linux operating system, designed by HD LogSystem, with two integrated stereoscopic cameras.In the case of mobile apps, measurements were made by taking photographs of the front of the wood stacks; if there was more than one photograph, the app would com bine them into one image (panorama picture) (Fig. 3).In the case of LogStackPro, the wood stacks were measured by recording a video of the front of the log

Study Design
The study was carried out from November 2020 to May 2021 at the Forest Experimental Station, Poznan ´ University of Life Sciences.During the study, 71 stacks roundwood were measured, of which 48 were pine and 23 were oak.The total cubic meter stacked (CMS) volume of the measured wood was 3481.15 m 3 (Table 1).Each stack was measured by the standard method used for timber trading in the local market, and by three photo-optical methods: Log StackPro (stereo camera technology), iFovea Pro (smartphone app), and Timbeter (smartphone app).All logs in measured stacks had the same length -250 cm and had similar quality, which enabled them to be classified into the same assortment group (Fig. 1).All measurements were taken by 4 foresters, who have been trained by the software manufacturers.All stacks were measured under similar weather condi tions.If there were any errors in the measurement, e.g., stitching of the image, the measurement was repeated so the final result was based only on correct trials.Next, based on the volume, each stack was as signed to one of the following size classes (CMS in m 3 ): 1) 0-10; 2) >10-20; 3) >20-40; 4) >40-60; 5) >60-80; 6) >80-100; 7) >100-150; 8) >150 (Table 2).

Manual Measurements
Manual measurements were made according to the standards laid down for volume calculation and mea surement of stacked timber for trading purposes (the State Forests, Poland; document no.51/2019).To esti mate the volume of a wood stack, the log length, aver  stacks.As a reference for the mobile apps, the width of the log stack was measured manually, using a measuring tape with 1 cm accuracy.In the LogStack Pro system, the reference was the constant distance between the stereo cameras.Each measurement was performed at a distance of approximately 2-4 metres from the wood stack, in accordance with the instruc tions.In the next step, post-processing editing of the detected contour was carried out (Table 3).
Finally, the best digital measuring system was se lected based on mathematical criteria: the best solu tion was considered to be the one that gave a mean timber volume closest to the mean volume obtained from manual measurements.The second criterion was standard deviation between the volume obtained by the electronic measurement and the volume mea sured manually.

Statistical Analysis
To verify the distribution of the data, the Shapiro-Wilk test was performed.The data did not meet the requirements of normality, which led to the rejection of the normal distribution hypothesis.To compare data obtained from the analysed methods of measure ment, the non-parametric Mann-Whitney Utest was performed.Statistical analysis was performed at the significance level α = 0.05.The program Statistica 13.1 (TIBCO Software Inc., Palo Alto, CA, USA) and the R package (RCore Team 2022) were used for the calcula tions and for visualization.

Results
The average photo-optically measured volume of the stacks was approximately 49.8 m 3 CMS, while the average result obtained by the manual method was 46.6 m 3 (Table 4).The largest wood stack volume mea sured manually was approximately 221 m 3 , and the smallest was 8 m 3 .The most comparable results were obtained by LogStackPro, which gave a volume only 3.4% higher.The mean volumes as measured by iFo vea Pro and Timbeter were quite similar; compared with the manual measurements, they were 8% and 9% higher (Table 4).The highest differences between pho to-optical measurements of the examined software were noticed between LogStackPro and Timbeter, and then between LogStackPro and iFovea Pro.Mean while the differences in obtained CMS volume be tween iFovea Pro and Timbeter were below 1% (Table 5).In the case of pine wood, the average volume ob tained by the manual method was 61.3 m 3 .The highest measurement accuracy was achieved by LogStackPro (+3.5%), while the differences for iFovea Pro and Tim beter were around +8.0% and +9.0%, respectively.For oak wood, LogStackPro again gave the most accurate results (2.7%); for both iFovea Pro and Timbeter vol ume was 6.6 greater (Fig. 4).Based on the Mann-Whit ney U-test, there were no statistically significant dif ferences between the examined methods, either for all measured stacks or for specific wood species.When different size classes of wood stacks were considered, the most accurate method of photo-optical measurement was LogStackPro.The results obtained with the mobile apps show Timbeter to be more ac curate for only two classes of stack volumes, 10-20 m 3 and 100-150 m 3 , while iFovea Pro was more accurate in most of the classes (Table 6).In the case of pine wood (where the smallest stacks were assigned to the 10-20 m 3 class), iFovea Pro made more accurate mea surements than Timbeter in every class except for 100-150 m 3 and 10-20 m 3 (Table 7).The findings for oak wood were similar to those for pine wood, al though the measurements included wood stacks from only three size classes (Table 8).The iFovea Pro app achieved greater accuracy only in the case of the 10-20 m 3 class.Statistically significant differences were found only in the case of 80-100 m 3 pine wood stacks measured using iFovea Pro and the manual method.

Photo-Optical Measurements
Most published studies provide only data on mea surement accuracy between one selected app and a manual method (Kärhä et al. 2019, Berendt et al. 2021, Cremer et al. 2021) or present general information about photo-optical technology (Pachuta and Table 6 Differences in volumes obtained with photo-optical methods -all compared with volumes obtained by manual method (defined as 100%); pine and oak -indicate that the volume obtained by photo-optical method was higher or lower, respectively, than the volume obtained by manual measurement.All values in %   (Herbon et al. 2015, Kärhä et al. 2019, Berendt et al. 2021, Cremer et al. 2021).
Almost all of the average photo-optical results for timber volume obtained in the present study were higher than the volumes obtained from manual mea surements.In this study, the LogStackPro measure ment system achieved the best accuracy (+3.4%); the CMS volume results from the mobile apps were ap proximately +8.0% (iFovea Pro) and +9.0%(Timbeter) higher than VMM.Measurements of stacked oak tim ber were more accurate than measurements of stacked pine timber.In the case of oak timber, the measure ments performed using the mobile apps had similar accuracy.However, the pine timber measurements obtained with iFovea Pro had better accuracy than those of Timbeter.The reason for this difference may be the smaller range of stack volume classes in the case of oak.The results of this study are not in very good agreement with those of Berendt et al. (2021), who ex amined the accuracy of broadleaved wood measure ments performed using only the iFovea Pro mobile app.That study found average deviations between -5.21% and +0.53% when oak timber volumes were measured using iFovea Pro, while in our study the deviations ranged between +5.14% and +8.09%.
The results obtained using stereo camera LogStack Pro measurement technology exhibited the best ac curacy in every analysed stack volume class.The vol umes obtained with the mobile apps were higher than those given by LogStackPro.Moreover, the results obtained from the two apps within the same volume classes were not similar.In most volume classes the iFovea Pro app provided better accuracy than Timbeter.No statistically significant differences were  wind causes standing trees in the background to sway, or when the timber stack covers less than 60% of the photograph.A very important aspect of correct photooptical measurement is the quality of wood stacking.
Stacks should be piled as evenly as possible to avoid errors during image recording.In addition, photo-op tical measurement requires a high degree of accuracy on the part of the person using the device.The user should carry out the measurement according to the manufacturer's guidelines and must determine the ref erence length very accurately.It is also important to walk alongside the stack parallel to it, maintaining the same distance from the front of the stack.Failure to observe the principles of correct measurement may re sult in incorrect stitching of the series of photographs and can lead to a distorted final orthophotomosaic im age, on the basis of which the final stack volume result is calculated.Photo-optical measurements using length references can also be subject to errors associated with reference determination.The reference must first be determined on the stack to an accuracy of 1 cm.It should then be plotted on the orthophotomosaic in the post-production screen.Any deviation from the starting value of the reference results in an erroneous volume result for the respective stack.Similar errors can also occur during manual measurement, which has a rela tively low repeatability.Meanwhile, many application manufacturers claim that photo-optical measurements can be performed with an accuracy of ±3%.However, only LogStackPro and Dralle have certified accuracy.Therefore, for the time being, the best photo-optical so lution seems to be the one that does not require the user to take additional measurements at the front of the stack.

Study Perspectives
Looking to the future, considering the develop ment of technology in the mobile device market, which will favour the introduction of digital tech niques in forestry, an interesting solution for the mea surement of stacked timber may be the use of LiDAR.This is already being used on devices with the iOS system; for example, for measurements related to for est inventory (Balzter et al. 2007, Magnussen et al. 2018, Pyörälä et al. 2019, Gollob et al. 2021, Xu et al. 2021) and single log volume (Borz et al. 2022, Borz andProto 2022).

Conclusions
As technology advances, the devices and applica tions used in forestry can provide increasingly accu rate measurement data.According to the results of this observed between the manually and photo-optically measured timber volumes, or between the photo-op tical methods.
From a statistical point of view, the results show that there is no difference between the compared pho to-optical methods for timber measurement.Howev er, from a practical point of view, every cubic metre by which the measured volume falls below the actual volume represents a loss to the forest owner.This has also been pointed out by other authors (Pásztory et al. 2019, Berendt et al. 2021).The mean price of 1 m 3 of timber in Poland in 2021 was approximately 47 EUR (1 EUR = 4.81 PLN according to data from the Nation al Bank of Poland as at 5 October 2022).Therefore, an error of ±1% in calculating wood volume results in a gain or loss to the forest owner of 4.7 EUR per 1 m 3 .In the case of photo-optical measurements, manufacturers claim that an experienced worker can repeat the mea surement with an accuracy of ±3÷5%.However, no sci entific studies have yet been conducted to confirm the effect of experience on measurement repeatability.
Another very important aspect of the measurement procedure that affects its accuracy is the length refer ence.Depending on the application used, length refer ences from 1 m up to the full width of the wood stack are recommended.In this study, measurements made with the iFovea Pro application, for which the refer ence segment covered the full stack width, were found to be more accurate than those of the Timbeter applica tion, where the length of the reference segment was 75% of the stack width.This may suggest that the size of the reference segment can affect the accuracy of the measurement.However, this hypothesis needs to be confirmed by investigating the effect of reference seg ment size on measurement accuracy within a single application.

Study Limitations
Despite their good accuracy, photo-optical measure ment methods may also be subject to limitations.Smart phone photo-optical applications place limits on the size of stacks that can be measured; for iFovea Pro the maximum width is 30 m and the maximum height ap proximately 4.5 m.Timbeter does not define such lim its.If the width of the stack exceeds the app's maximum measurement value, it is advisable to divide it into sev eral separate sections and to measure each section sep arately.Another limitation of the measurements may be excessive light.If there is too much light in the field, it is recommended to shade the camera using one's hand, although this is not a very convenient or effective solution.Problems with measurement can also occur if the stack is situated near the logging site and a strong study, it can be concluded that, at present, the use of stereo camera technology (LogStackPro) is the most promising and accurate solution to replace the manu al measurement of timber volumes for trade purposes.However, from a statistical point of view, the differ ences between all of the tested methods of stacked timber measurement were not significant.From a practical point of view, every cubic metre by which the measured volume falls below the actual volume (man ually measured) represents a loss to the forest owner.The length reference solution seems not to help very much in avoiding errors in photo-optical wood mea surements.LogStackPro gives good accuracy of tim ber volume, it is easy to use even without length refer ence, and seems to be very practical in many aspects.A valuable feature of the whole measurement process when using LogStackPro is the fact that, when acti vated, all data collection is taken in one go without stopping (creating a short film), rather than taking single pictures.Considering all results and practical issues, LogStackPro is a good candidate for future stack timber measurement and trade.

Fig. 2
Fig. 2 Standards for measuring timber stacks: points of height measurement and determination of the top edge of a pile.If the measurement fall between two logs (h 2 ), the height for the lower adjacent log was taken

Fig. 3
Fig. 3 Devices used for LogStackPro (a) and mobile apps (b)

Table 1
Number and volume of manually measured wood stacks, by species

Table 2
Number of manually measured wood stacks and all volume measurements in stack size classes

Table 3
Hardware and software used for photo-optical measurements

Table 4
Comparison of volumes obtained by photo-optical and manual measurement for all timber stacks

Table 5
Lepoglavec et al. 2019, Pásztory et al. 2019red stacks between photo-optical methods of measurements, % Chojnacki 2018,Lepoglavec et al. 2019, Pásztory et al. 2019, Tomczak and Naskrent 2022).The range of wood volumes used in this study represents the typical dis tribution of CMS volumes of stacks of roundwood of the examined species.The higher proportion of pine roundwood in the study was a consequence of the pro portion of that species in the total timber harvest in the experimental area.Overall, 14,054,000 m 3 of industrial conifer roundwood and 4,288,000 m 3 of hardwood roundwood were harvested in Poland in 2020 (Statis tical Yearbook of Forestry in Poland 2021).Similar studies concerning the accuracy of photo-optical wood measurements in real forest conditions have been car ried out by other authors in bold -base value Fig. 4 Range of accuracy obtained by photo-optical and manual measurement with respect to species.Whiskers correspond to minimum and maximum values, boxes represent the 1 st and 3 rd quartile values, midlines indicate the median

Table 7
Differences in cubic meter stacked volumes obtained by photo-optical measurement methods compared with manual method for pine wood stacks in different stack volumes

Table 8
Differences in values obtained by photo-optical measurement methods compared with manual method for oak wood stacks in different stack volumes